Eva Dou is the author of The House of Huawei, an excellent book covering the personal, economic, and geopolitical arc of Huawei, China’s most important company.

We discuss…

• The life of Huawei’s founder, Ren Zhengfei, who rose from Cultural Revolution disgrace to become one of China’s richest businessmen,

• How Ren built Huawei, and what makes their corporate culture unique,

• Huawei’s strategic entry into developing and high-risk markets like Libya, Iraq, and Iran, and whether the controversial deal with the UK is a threat to national security,

• How Huawei outcompeted Chinese state-owned telecom companies and eventually achieved national champion status,

• How Ren’s personal interest in foreign art, music, and architecture advances Huawei’s market share.

Co-hosting today is Kyle Chan, a postdoc at Princeton and author of the High Capacity Substack.

Have a listen on Apple Podcasts, Spotify, or your favorite podcast app.

Forged in Fire 浴火重生

Jordan Schneider: Eva, let’s start with the Cultural Revolution. How did it impact the future founder of Huawei, Ren Zhengfei, in his early years?

Eva Dou: Huawei’s founder, Ren Zhengfei, comes from quite a humble background. He grew up in rural Guizhou Province, where both his father and mother worked as schoolteachers. His father eventually became a principal, and they instilled in him the importance of learning and a love of reading that continued throughout his life.

At the same time, that academic background made them a target when the Cultural Revolution came around during Ren Zhengfei’s teenage years. His father had worked briefly for the Nationalists earlier in his career, making him one of the many people targeted during the Cultural Revolution. He was struggled against at his school, criticized by students, and eventually put in a labor camp.

This affected his entire family — Ren Zhengfei and all his siblings. It impacted their work prospects throughout his early career. He would later complain that he was unable to get the promotions that his peers received, no matter how hard he worked. His company Huawei is now considered case number one of a national champion in China, but he really started out on the outside of the system, looking in. That’s part of what makes his story arc so dramatic — how far he came.

Jordan Schneider: There are a number of passages that were really evocative for me. There’s this theme of suicide that runs through the book. Ren Zhengfei is an incredible workaholic with multiple divorces who at one point said he was excited to use Chinese national holidays to travel abroad and get more work done.

This connects back to his father’s experience during the Cultural Revolution. You have this line that said Ren Moxun 任摩逊, his dad, also considered ending his life, but he didn’t want to die before his name was cleared, as it would leave his wife and children with a cloud over them. Ren Zhengfei explained that if his father died, his children would have to carry this political burden. “He endured 100 tortures, but would not kill himself.”

Later, Ren Zhengfei said, “The Cultural Revolution was a disaster for the nation, but for us, it was a baptism. It made me politically mature so that I wasn’t a simple bookworm”

When you look at Huawei from the outside and see the handsets, AI accelerators, and base stations, at its core is this stupendously driven human being who was shattered to his core. There are many business stories in China where people come out of the Cultural Revolution — Xi Jinping as well — forged into these super-driven individuals. But obviously tens of millions of people didn’t turn out that way. Even the ones who did end up succeeding beyond their wildest teenage imaginations carry very deep scars that play out over the subsequent decades of their lives.

Eva Dou: Many people have pointed out that Ren Zhengfei, as an entrepreneur and business person, carries this pessimism with him that some attribute as one reason for Huawei’s success. He’s always looking at the worst-case scenario that could happen to his company. Even in the years when business was flush, he would warn his employees that something bad could always happen — they could go bankrupt at any moment and had to be prepared for anything.

These early experiences of facing life-and-death stakes and seeing how fickle the political environment and business can be really did inform him. In many ways, it helped make Huawei a company that could survive unexpected things, which we’ve seen through two Trump administrations and many unexpected developments.

Jordan Schneider: There’s a bit you have later where you talk about how he painted black swans everywhere once he built this giant campus. The Cultural Revolution was the ultimate black swan. Once you live through that, what’s a couple of sanctions violations?

Another illustration of the psychological jumps you have to make to process and move forward from living through that time comes from this incredible quote from his dad when reflecting about his experience being struggled against in middle school. You write: “One day, one of his old students visited him and recalled the struggle session where a classmate had beaten him with a wooden stick until it broke. Ren’s father smiled grimly. ‘I have to thank that piece of wood. Had it been sturdier, I surely would have been beaten until something was wrong with me.’”

It’s a lot. Kyle, are there other business stories that come to mind — aside from Xi Jinping and Ren Zhengfei, which we’ll be exploring in an upcoming episode with Joseph Torigian — about this kind of dramatic arc for Chinese business people?

Kyle Chan: What’s really interesting is that there’s a strong parallel between Ren Zhengfei’s worldview and the view of many Chinese entrepreneurs during these tougher years, as well as Chinese political leaders in their attitudes towards China’s development. This idea especially that we are operating in a very tough world, that we have to rely on ourselves, that there’s danger at every corner and risk, and what to do about that preemptively.

Jordan Schneider: On the arc towards being a national champion, Ren works for the PLA as an engineer for a few years, finally makes it into the party — which was a dramatic arc in and of itself — and then gets invited to this famous Deng National Science Conference in 1978. There, Deng Xiaoping essentially blessed everyone, saying it’s okay to start businesses and do science — you don’t necessarily have to live inside the system to serve the country. That was a big psychological unlock, almost in the way you describe how he reflects on that. Can you talk about these years? Let’s bring the timeline up through 1978.

Eva Dou: The first part of Ren Zhengfei’s career was as an engineer in the engineering corps for China’s military. During this period, he was largely low-key. Occasionally he would gain kudos from superiors for his work, but he wasn’t anyone famous by any means. He probably thought that’s where he was going to stay for his entire career, his entire life.

This abruptly changed because Deng Xiaoping decided to make this switch to begin the capitalist experiment in China. This was coupled with a dramatic downsizing of the military, and he was laid off from the military and sent to Shenzhen to work in this burgeoning private sector.

He’s talked about feeling very disoriented at the time. This was similar to the experience of many people in China who had been in a system that was only a planned economy and suddenly it was all different. They had new ideas, new fashion styles, and new trends coming in from Hong Kong over the border to Shenzhen. He had this sense that he was really far behind — he was already almost 40 at that time — in trying to adapt to this totally new world.

Jordan Schneider: You have this quote where Ren would later tell his colleagues that he’d wasted time during those years when he was outside the establishment, not as a party member. You can read this quote two ways: “I was a soldier for all those years I didn’t join the party. My life was full of adversity. When I think of all that wasted time, I wonder how could I have been so naive and ridiculous that I didn’t understand it all being about compromise and shades of gray."

On one hand, I see that as maybe this guy’s still kind of pissed that the party ruined his parents’ life. But also, there was no way he was going to be accepted as a party member, even though he was an engineer for the PLA, because of his class background — his nationalist background. What’s the right way to understand that quote, Eva?

Eva Dou: I agree, it’s a bit ambiguous. He does give this kind of counsel to his staff repeatedly through the years, advising them basically that politics is not fair. He’s told them straight out that life isn’t fair and you have to keep in line with the political system that we live in. He said Huawei is a Chinese company and it’s a requirement to be patriotic to be part of the company.

That’s informed partly by his earlier experience with the Cultural Revolution. But there’s another part that isn’t talked about much, which is the Tiananmen democracy protests in the late 1980s. It’s remembered largely as an event in Beijing at Tiananmen Square, but these protests were going on across the country. Some of the largest ones were in Shenzhen, where the young people there were very progressive, international, liberal, and bold. They held some of the largest protests.

That was at the time that Ren’s daughter, Meng Wanzhou, was in high school and about to go to college in Shenzhen. Those protests were quashed in Shenzhen as well as Beijing. That also informs the approach and worldview of Ren Zhengfei and other business people of his generation. They remember these things.

Jordan Schneider: It comes back to the thing that Kyle brought up earlier: to what extent does he really believe the nationalist stuff versus just being a business person who wants to do cool science and compete with Ericsson because that’s cool and will make you money and make you remembered?

You have all these quotes where they’re laying it on too thick at times with how much they’re using the nationalist angle and the national strengthening angle to motivate employees. But for anyone in that generation, there has to be some kind of internal narrative that not all is right in the state of Denmark. I don’t really know where to go with that.

Eva Dou: There is one comparison that I look at in the book. Before Huawei in the 1980s, China’s most promising tech company was this company called Stone Group. The founder of that company was bold and radical in his political beliefs. He actually supported the pro-democracy protests at Tiananmen, and it basically ended up with his company being wiped off the map in China.

He ended up having to flee the country shortly after the Tiananmen crackdown. He was never able to return to China again and has lived as an exile ever since. Ren Zhengfei’s generation of entrepreneurs grew up revering this company and these entrepreneurs, and they saw what happened there. Since then, it’s been taken for granted that you have to bend the knee to the political line to build a successful business in China.

Jordan Schneider: One of the interesting leitmotifs of this book is Huawei’s relationship to the state. The story that you talk about, and that Doug Fuller goes into more deeply in his book, is the fact that Ren actually kept the state at much more of an arm’s length than many of his competitors in the 90s and 2000s. He focused on export discipline and investing in R&D as opposed to the cheap, easy money that you can get from government contracts. That was the way he kept Huawei advancing at the technological frontier.

Let’s give a little sense of what the industrial upgrading arc that Huawei went on in the 80s and 90s looked like, and what decisions they made different from competitors who ultimately weren’t able to achieve the levels of greatness that Huawei did.

Kyle Chan: Just to jump in with some broader context here: with China’s effort to develop its own telecom equipment industry and its broader tech industry, there’s this really interesting connection between the role of the state and the role of these different businesses. At various times you have other companies, other state-owned enterprises like Great Dragon that were the real national champions originally. There’s this idea that some companies were seeking to form connections with the state, while others were actively presented as the spearhead for China’s technological and economic development versus Huawei at this time.

Now we know Huawei as playing that role much more prominently. But back in those early days, Huawei was one of many competitors and was actually the underdog. To build on what Jordan was saying, Eva, you had a great passage about how Ren Zhengfei was even worried about taking loans from state banks, wary about becoming too entwined with state interests and all the strings that are attached. Could you elaborate on some of those early years and eventually reaching that status that we now associate with Huawei of dominating the industry? Really, in those early days, that was a story that was still unfolding.

Eva Dou: Especially in those early days, China’s state sector — these state-owned enterprises — had a reputation of being very sluggish and uncompetitive, basically a place where people knew they would stay employed whether they worked or not. This was something that Ren Zhengfei knew meant the death of his company if they became like that. He was quite wary of being too intertwined with the state, especially in Shenzhen, which is where this capitalist experiment was beginning in China.

They saw themselves as different, as forging a new kind of business system than the rest of China. They would call the rest of China “the inland.” Huawei was very aggressive, very vicious in how they went about doing business and also in their expectations for their own employees. They were notorious for firing people summarily if they didn’t meet expectations.

For a while they were doing joint ventures with state-owned companies across provinces in China to try to expand their footprint. That was a big culture shock for the state-owned employees. You have these stories that they were used to taking their daily naps during the day when they got sleepy. Then the Huawei management would come in and clean house, and things would be totally different.

Jordan Schneider: I love the part where, when they start winning, the state-owned company is like, “Hey, come on, this is a private company. Why are we letting them win these contracts?” Can you tell that story? I really liked it.

Eva Dou: It was one of the local officials in Shenzhen who had written a memoir after he retired with some of his recollections of dealing with Huawei and these other companies. He talked about how the state-owned companies after a while started coming to him and complaining that he couldn’t just be supporting Huawei, that he should be supporting them too. He put them in their place and said, “Well, you guys have been getting these subsidies and all these benefits for so long. If you can’t compete with Huawei, you have to look at yourself and see why that is."

Jordan Schneider: They sell cheaply to get market share. Someone at Great Dragon complained to Shenzhen’s Science and Technology Bureau that the local government shouldn’t just support privately owned enterprises. I mean, sorry man, sucks to suck. You could have been Huawei in the 80s, but instead you let Ren do the thing.

We’re in the 90s now. Huawei is a going concern, on the up and up. There is this fascinating blend of Western culture and PLA culture that Ren tries to inject into the lifeblood of how Huawei operates. Can we talk about the different influences that he tries to incorporate and put into his company and how it does business?

Eva Dou: He and his deputy team really felt that they were forging a new kind of business model in China — one that was not the state-owned enterprise and also not purely the Western style of private enterprise. He was seeking wisdom from all different places in the mid-90s and late 90s when he was working on this.

Famously, IBM was the Western company that they hired as a consultant to teach them how to transform from a startup into a multinational company. There were just many things that Ren Zhengfei knew would take them too long to figure out on their own: How do you meet production schedules year after year? How do you plan product launches to remain on schedule? How do you manage your supply chains? They went to IBM and some other Western companies to try to learn this as quickly as possible. At the same time, he was also meeting with leading Chinese entrepreneurs to try to understand what the relationship to the state should be at that time.

Subscribe now

Jordan Schneider: In 2017, I was at PKU in grad school and they have all these recruiting events. Of course I’m going to go to the Huawei one, and they hand out this book that they put out, “The Huawei Way,” where it’s these chapter-long essays of employees reflecting about their experience in the culture. I’m sure you’ve read it.

It’s funny, Eva, because usually when you see business people and they’re defining a new corporate culture, it’s bullshit and they just want to look back on themselves as intellectuals and business innovators. But in all of my reading about Huawei and diving into this book, I mean it is a weird artifact, right? This company and the way it goes about things.

His openness to Western management practices, the eagerness to go overseas and localize while at the same time the wolf warrior, hardcore zaibatsu energy but also up-and-out and “we’re going to fire you if you don’t deliver" — all of that is not something that you see anywhere else and only comes from Ren personally pushing to synthesize his personal experiences and his drive to synthesize a new way of being from a corporate perspective. That’s one of the most fascinating parts of his story and your book in particular.

Eva Dou: In most countries, there are older businesses, businesses that have been around for many years to look back to. In China, just because of the Cultural Revolution, that was all wiped out and there was a sense they’re starting again and have both the privilege and the responsibility to build a new model. Ren has said many times that he wanted to build a company that could last for 100 years. It’s been very much an open question how you do that. It’s only been a few decades since the end of the Cultural Revolution really, so it’s still an ongoing experiment of building a private enterprise in China that can last for a century.

Jordan Schneider: Kyle, do you want to do the East Asia comparison? I think Samsung is kind of the closest analog from a crazy founder, hardcore founder perspective. What other parallels do you see?

Kyle Chan: There are other ones too, like the founder of Honda and also the founder of Toyota, breaking the rules in the early days, not being allowed actually to have a license to produce cars but defying that and then eventually becoming the star player in the industry. You see this again and again, and actually another strong parallel is just the expectations — these sky-high expectations that you get now in Silicon Valley.

I jokingly have talked to people about this book and said, “Wow, Elon Musk would love this approach to just hard charging, ‘Go to Mars’ mentality that Huawei seems to instill in its work culture.” Some of this is nuts and bolts of how you run an organization, but some of this really seems to be, “We can build, we can defy the EUV lithography export controls” — which we’ll get into later. But all those things are just mere speed bumps along the way to these greater global ambitions.

Jordan Schneider: When you read the early Apple stories of just the intensity and the near-death experiences, you get a bit from Steve Jobs at his most unleashed peak. But I really think Samsung is the closest comparison. They have a similar story in their arc to the one that you wrote about, Eva, where everyone signs their resignations and says, “I will sacrifice.” The importance to the country arc that ends up developing is not quite as present in a Tesla or a Honda as it is in a Samsung, which modernized an entire country for better or for worse.

Eva Dou: That’s interesting, the comparison to Samsung. Huawei’s mission has shifted over the years in that in its early days it was not trying to be a company like Samsung. In fact, Ren was very adamant that they were going to be very narrow in what they did — that they were just going to do telecommunications equipment. They didn’t want to be this sprawling multi-armed conglomerate because they felt they would end up not being good at anything.

It’s been a process where they’ve gradually expanded. During the 2000s they began making mobile phones and then smartphones. That was something that generated a lot of debate inside the company. Ren at the beginning was against it. He thought they were diversifying too much and was eventually won over by some of his deputies that this was the way to go. More recently, because of sanctions, they’ve been forced to make more of the nuts and bolts themselves, especially chips.

By how this worked out, they’re now a much more direct comparison to a company like Samsung — they have so many different lines of business. What’s interesting is they didn’t start out that way, and for many years that’s not what they wanted to become.

Jordan Schneider: It’s interesting also because this happens over the course of Ren getting older and older. He has this great line where you said he suspected that smartphones were overhyped. “We believe that the Internet has not changed the essence of things,” he told staff. “A car must be a car and tofu must be tofu."

But if he succeeded, which he did, in hiring all of these hard chargers who are going to want to spin up new industry verticals or whatever — just having them all fight for the same three slots running different verticals in telecom — is actually not sustainable. One of the remarkable things is that he was not stubborn enough to shut down these new business lines. His commitment to investing — what’s the number? Like 30% of annual revenue into R&D — some crazy stat which basically no other Chinese company did in the 90s and 2000s and even into the 2010s — is something where if you’re going to end up doing all that, then yeah, you’re going to end up wanting to branch out into new businesses and grow in different directions.

Kyle Chan: Speaking of branching out to new businesses, I was wondering if you could say more about this moment in Huawei’s rise where they’re getting into a whole bunch of different areas. While they began with switches and telecom equipment, they really started to expand into everything from undersea cables, and then eventually EVs, AI, and semiconductors. Actually, your book points out that Huawei was working on semiconductors from a very early stage and was quite important in China’s broader push into semiconductors. Could you say more about that period and what drove some of that expansion? What was that like for a company that began with such humble beginnings? This is part of the arc leading up to a $100 billion per year revenue business.

Eva Dou: One way to look at it is that their customers — their international customers — are often buying all these different things. From that sense, it makes a certain logic. Smartphones and telecom gear seem so different — one’s very specialized, one is for the consumer — but it actually ended up being a savvy way for them to get their foot in the door into markets where it was quite difficult to sell their telecom gear, such as Europe, because smartphones were seen as much less sensitive technology.

In countries where they weren’t able to sell telecom gear directly at first, they were able to sell small, cheap smartphones or mobile phones. Actually, the customer is the same — it’s these telecom operators who run these mobile shops, where, as a consumer, you would go in and buy a phone. That allowed Huawei to develop these relationships with these companies. In some cases, once they were selling phones, they were eventually able to sell the telecom gear.

Chips are a little different. It started out partly for cost reasons. For some cheap chips, they found they were able to save costs significantly if they were able to produce them themselves instead of relying on foreign suppliers. There’s probably a security component to that too. He said from an early time he’s pointed out the national security implications of telecom gear, and it probably helped assure Chinese officials if they were making some of their own chips. In fact, their early chips were often used for surveillance cameras. That was one of the early lines of business for their chips.

They’ve gradually found that one line of business can help the other line of business. They’ve expanded probably beyond what Ren would have been comfortable with back in the 1990s.

Spare Tires and War Zones

Jordan Schneider: I love this line from the head of HiSilicon in 2019, which is Huawei’s chip development arm, who said, “For years, the unit had worked on spare tires 备胎. But now, today is the decision of history. After Trump’s export controls on the company overnight, all the spare tires we built have become the main tires. Our years of blood, sweat, and tears have been cashed in overnight to help the company fulfill its commitment to keep serving customers."

It’s very emotional working at this place. On the one hand, you have these stories of people — one of the executives talks about how his annoying family will never forgive him for missing the birth of his two children. But on the other hand, it’s clear that for these employees this is not just a paycheck. There is a lot of life fulfillment and meaning that these folks derive from this work. Maybe I was a little too dismissive earlier about them laying it on too thick because, look, we’re in a country where religion isn’t really a thing and you’ve got to find it from somewhere to get you motivated enough to work 80, 100-hour weeks past the point where you’ve already met your creature comforts.

Eva, what are your thoughts on Huawei as a spiritual totem for its founder and employees?

Eva Dou: Definitely, part of Huawei’s success has been very much its founder’s charisma and his ability to motivate his employees to work incredibly hard and sometimes take enormous personal risks in working for this company. There is a spiritual component to it where he is telling them they can find meaning in helping their country become technologically sufficient in key technologies. The United States has been this bugaboo throughout its history that’s held up as a country that is hostile to China’s development, that wants to hold the country back and that they at Huawei would have to surmount.

Through the years some of that has sounded melodramatic and overblown. When the sanctions came down during the first Trump presidency, that made people believe that Ren wasn’t just overly scared, that he had indeed been preparing for what was going to happen.

Jordan Schneider: Another Elon parallel, right. The American company that gives the most spiritual energy to its employees to work really, really hard is SpaceX, with this incredible vision of making us an interplanetary species.

My favorite illustration of the truly devoted service to customers — not just the Chinese government — is during the Libya civil war. Literally everyone and their mother was pulling out of the country. The Chinese diplomats pulled out. Same with Bahrain when they were going through their civil disturbances. But Ren Zhengfei and Huawei were like, “No, we’re going to keep our engineers here. We signed some contracts. We’ll just have two businesses, one for the government, one for the rebels."

You have this great interview with the wife of one of these engineers saying, “Yeah, every time we call him, we hear gunshots in the background.” Libya is not this huge growth market, but it is a fascinating illustration of just how committed this company is to expanding and succeeding.

Kyle Chan: This is a really interesting segue because so much of this book is looking at Huawei’s trajectory over time in parallel with China’s. Some of these twists and turns and even geopolitical strategy at the company level mirrors that of China’s approach. Huawei, especially in the early days when it was really trying to establish itself internationally, was entering these markets that were a bit too risky or too volatile perhaps for some of the more established players. It gave Huawei an opportunity to start working outside China.

It parallels China’s efforts to develop relations with different parts of the Global South — Southeast Asia, Africa, Latin America, Middle East — while gradually seeing ties with the West as being a longer-term project, but one that would be much harder to tap into early on.

Could you talk more about this opportunistic phase in Huawei’s trajectory in terms of its global expansion? Later on, when you see one of these maps of where Huawei ended up, it’s all red Huawei or ZTE 5G equipment all over the world. That’s the later endpoint, but along the way, what was Ren Zhengfei trying to do and how were they trying to make inroads piece by piece into the international market, culminating with the British Telecom contract, which gave them that kind of legitimacy that parallels China’s quest for legitimacy on the international stage?

Eva Dou: Certainly. Huawei’s international expansion can very much be seen as a proxy of China’s broader international relations effort over the decades. The company was founded in 1987, and at the beginning they only focused on the domestic market. Then in the mid-1990s, they started looking overseas.

At the very start, Russia was one of the first countries they looked at. That was partly through diplomatic help — the Chinese Foreign Ministry helped them go over for a trade show and helped smooth the way to their first very small contract there. At the same time, they were looking to countries in the Middle East and Africa in the early days.

There is some history of their eventual run-in with sanctions issues with the United States, which is that some of their earliest customers were countries where the major Western companies were a little more reluctant to go for whatever reason. That included Iraq, Iran, Afghanistan, North Korea.

Then you get to the point where China is entering the World Trade Organization in the early 2000s. Huawei jumped on this trend of internationalizing. That was when you saw them start to push into developed markets, into Western markets more aggressively. Of course, it took many years for them to be established in Europe. But if you look today, there’s hardly any country in the world that doesn’t use Huawei telecommunications equipment to a greater or lesser extent.

Kyle Chan: In parallel with this rise, what’s really interesting is that as Huawei is expanding globally and as Huawei’s star is rising in China as a symbol, you have growing suspicions from the West, especially from the US, about what Huawei is doing and what its real relationship with the Chinese government is. There are concerns about surveillance technology. You have the rollout of smart cities and safe cities programs from Huawei. This ties into the very markets that Huawei is deciding to enter.

In the early years, as you mentioned, Iran and Iraq were some of the earliest cases of Huawei coming on the map in terms of US concerns about security. How did you see this reaction to Huawei’s expansion, especially from the US? You have congressional hearings later on, which is a high point of the book, a point of high drama, where Huawei and ZTE executives are brought in front of Congress to testify and answer basically a pretty tough interrogation session. Something similar happens with British Parliament to a certain extent.

There are growing suspicions rising almost right alongside Huawei’s growing prominence on the world stage. Could you describe some of that and where it’s coming from? Obviously Huawei’s own activities within China contribute to this. It’s not just what Huawei’s doing abroad, but later on, you were one of the earliest reporters on use of Huawei surveillance equipment in Xinjiang, for example. Could you tie all those different pieces together?

Eva Dou: The first incident where Huawei really got on the map for the US government was in Iraq in the early 2000s. Huawei was helping the local government and military build a fiber optic telecommunications network. If you think about it, these older systems where they’re using radio communications — that’s something that is more easily surveyed or hacked. That’s what the US government had been doing. They’ve been using it to keep track of what Iraq’s military was doing. Suddenly those conversations were going underground. Huawei was helping them build these fiber optic lines that ran underground and that were much more difficult to tap.

That was when George W. Bush gave the order to bomb those installations, which would force Iraq to use radio again. This was later explained by some US officials. That was something Huawei was doing for governments around the world, helping them build these more secure communications networks that were harder to survey for Western governments. That was how they came into conflict with the US at the start as a national security concern for the United States.

Later on this became other things. You mentioned Xinjiang and surveillance. Their surveillance systems became a significant line of business for them, both domestically and in countries around the world. These are the modern surveillance systems for an entire city. There are video cameras, but more importantly, there’s the software on the back — there’s facial recognition, there are AI algorithms to help track trends.

That is a more recent iteration, and what sparked the sanctions was the 5G generation of networks being laid out around the world and Huawei seeming poised to win an enormous chunk of those orders. That set off alarm bells in Washington at the time.

Kyle Chan: This gets to a whole question about the role of technology today and to what extent can we separate some of these core functions of what telecom equipment is supposed to do versus issues of data privacy and national security. I wanted to read a quote from the head of global cybersecurity for Huawei at a UK parliamentary hearing. He’s asked about whether Huawei’s telecom equipment could be used for surveillance and data collection and sending that back to China. His answer is: “It therefore does know where you are because it knows where the information is coming from. In that context, telecommunications networks from all vendors know where you are so as to connect you to those networks. Huawei’s equipment is no different from anyone else’s equipment."

This was just a question about whether Huawei’s telecom equipment can track you. His answer seemed to be, “Well, in order for a cell tower to work, that’s really what it does. It has to triangulate your position and make sure you receive and get the signal."

How much of this seems to become only a bigger issue over time? Now we have a whole bunch of questions about not just telecom equipment, but this has entered into many different domains. Everything from TikTok to connected vehicles, DJI drones. The latest one was TP-Link routers. It’s this growing concern from countries like the US about what will happen with data that’s passed through or collected through these systems.

Is there anything that Huawei could do to reassure the outside world, especially in the West? There was a really great example where Ren Zhengfei or a Huawei executive offered to share source code or do a tech licensing deal to just put it out there and say you can look through our source code and see what’s going on, if that’s any kind of reassurance. There are different ways of trying to deal with this issue. But to what extent is it just something like a Gordian knot that can’t be untied?

Eva Dou: The comparison to how US and UK policymakers have looked at this issue is pretty interesting, because until recently, they’ve taken very different approaches. The UK has taken the approach that these security risks can be mitigated, and we are comfortable with it. The US has taken the approach that this is an intractable problem — these security issues cannot be mitigated.

Cybersecurity experts in both countries receive very similar training. Ultimately, it is kind of a political question and a political answer of what kind of relationship you’re willing to have with China, what level of “risk” you’re willing to take on, and to what degree do you see China’s government as an enemy that needs to be blocked by all means necessary?

Until recently, the UK approach had been that they have this center where Huawei has its source code for its products and where UK officials, including UK intel officers, can go take a look at the back end and comb through it and resolve any cybersecurity concerns that they feel they have with it, and they were okay with it. Actually, today, a number of other countries do use this model, and they feel that mitigates whatever hacking risks that they feel there are.

Subscribe now

The US has been on the extreme end of the spectrum compared to other countries in feeling that these are unmitigatable risks that they cannot — that we just can’t use this equipment. That speaks partly to just how the US and Chinese governments see each other, that they see each other as direct rivals, whereas for other countries, it’s not the case.

Jordan Schneider: It’s funny, because there is kind of this spectrum, right, where you have these African countries and Malaysia being like, “Whatever, spy on me. What do I care? It’s not like I could stop them anyway. Might as well save the money with the Huawei kit.” Then we have the UK kind of in between. You’ve got this great back and forth between John Bolton and a senior UK security official saying, “What I don’t understand is why out of all the things — chips, AI, rare earth minerals, whatever it is — your administration has decided that a modest amount of base stations on hilltops in England is the epicenter of your new declared war on China. Why?"

"You got to pick something,” John Bolton replied, but that feels like a facetious response. The deeper one, which is the one you alluded to earlier, Eva, is that losing this entire industry of telecom broadly and base stations in particular to a Chinese player is something that the US couldn’t really countenance as the country that feels like it has more to lose from being spied on by China than even folks who maybe aren’t all the way at the side of Malaysia, but the UK that feels like they can kind of deal with it. Even if they can’t, how bad is it going to be? It’s not like they’re the ones upholding the industrial base of the liberal democratic order or what have you. Kyle?

Kyle Chan: It does tie into this question from the US side too of what are the motivations for each of these things? I see a lot of parallels for the EV industry. There’s a lot of debate about to what extent were the Biden era tariffs on Chinese EVs and then later on the connected vehicles ban motivated by security issues and to what extent was it motivated by economic competition issues about worrying about what will happen to American automakers and US auto jobs along the way.

That just puts the US in a different position than other countries when it comes to its relationship with China and Huawei and a lot of these other rising Chinese tech firms that seem to be now entering into spaces that US companies were comfortable being dominant in for a long time. There’s also that parallel with Europe too, taking a different approach, again with EVs. They have similar concerns, but there seems to be at least a greater interest in investment and production of Chinese EVs within Europe. All of that is just to say it’s a lot messier than merely “let’s block everything Huawei” or whatever, “anything goes.” There’s a full range of different approaches out there across the world.

Jordan Schneider: I want to talk a little bit about the organizational, senior management structure and how it pattern matches to the way the Communist Party structures itself. Can you talk a little about that, Eva?

Eva Dou: Huawei has for years adopted this kind of collective leadership model, which in many ways is similar to how China’s government is run. There’s a senior group of officials, both explicit and also unofficial, semi-retired officials who all have a role in deciding the direction of this company. On purpose, it is a little vague exactly where those lines are drawn of who’s in charge of what.

For now, Ren Zhengfei is still the top guy. He’s talked for years that he’s going to be retiring one day and now he’s in his 80s, and it’s still unclear when he’s going to be retiring. That’s going to be the big test for Huawei, of course, for any company — that initial, the first handover from the founder to the next generation of leadership and if they can keep the company running to the same degree of success.

Jordan Schneider: I want to close on the human arc a little bit. A lot of Western media makes fun of the Huawei campus. The fact that it’s kind of this Disneyland — it has a Versailles, it has a Kremlin, it has all these international styles. This moment that you write about, about him moving villages in middle school, you write that he was astonished when they moved to the county seat and saw a department store. It was the first time he saw a two-story building.

Then, a few decades later, he goes to the US for the first time, sees Las Vegas and thinks it’s the most beautiful city he’s ever seen. These giant buildings and the pyramids and whatnot. Then he builds this incredible company and decides that he wants to replicate all of these architectural wonders from the West. Not really from China, but from the West.

Just that arc of this boy whose dad was principal and was beaten to within an inch of committing suicide during the Cultural Revolution to go from having this really complicated, admiring, but also rivalrous relationship with the rest of the world and looking up to so much that the West has brought, but also really wanting to be able to compete and strive to stand at the same level, if not necessarily playing by all the rules that the Western countries and companies would want them to — that’s just a remarkable arc. That is probably the thing that’s going to stick with me most from this book. When Ren, who’s pretty old now, does end up passing from the scene.

Kyle Chan: This image of him from his early childhood days, really the rural boy from out in the provinces to then later on being able to be this savvy political operator at the national and international stage and basically getting on side with multiple Chinese leaders over time. You write about how he is going out doing deals with, from Jiang Zemin, Hu Jintao, and now Xi Jinping.

We recently had that big meeting with the tech executives, Chinese tech executives meeting with Xi Jinping. Who was there, sitting in the very front and center, was Ren Zhengfei himself. Who could have imagined, probably not him from his early days, that he would be at the center of so much of the Chinese leadership’s ambitions, US-China competition tensions, and part of this story of China’s rise over time.

Eva Dou: It is kind of staggering, the distance he’s traveled since his childhood to where he is today, both personally, his company, and also China, that China is now a true technological competitor on the world stage. That’s part of what interested me in this project — just understanding how much has changed and how it changed.

As far as the palaces that you alluded to that Huawei has built on campus, Ren has said — it’s not only the Western media that critiques it, sometimes his own employees do. He’s talked about this and said, “You guys can gripe about these palaces, but we build them for the customers."

It’s not just him himself. This has been China’s experience, broadly, this transformation. When they bring government officials and telecom officials from across the nation to the campus, this is the same transformation that they are experiencing themselves. It is appealing to a certain part of his customer base, all the glitz and glamour. Most of his international base is the developing world. Huawei is famous for bringing foreign officials on these junkets and wowing them with both the campus and the food and the hospitality.

Jordan Schneider: Eva, are there any stories you want to share from the reporting of this book? In particular, I’d be curious about, in reading a lot of these memoirs and Chinese coverage of the company, what are the narratives that get a lot of play in mainland China that you don’t end up seeing written up much in the Washington Post?

Eva Dou: The most fun part of this was doing archival research for me and trying to find sources that even in China are a bit obscure. You do have the easy narratives, both the international ones about Huawei being a national security threat and within China, Huawei being this great company that’s getting better and better. Part of what I try to bring to this is the smaller voices.

I really enjoyed reading some local government officials’ memoirs in China. As Huawei has interacted with so many officials both within China and around the world over the years, I wanted to bring some of their recollections of the experience of meeting with Huawei, both good and bad. In China, as you know, there’s pretty strict censorship. In memoirs, when people are late in their life and no longer worried about keeping their job, they are a little franker sometimes in their recollections. That’s been an enjoyable part of this process.

Kyle Chan: I was astonished looking through the sources that you cite and the amount of material that you had to go through. I’m sure the sources that you cited are just the tip of the iceberg for everything that you were wading through and trying to sift through. You even chased down a Harvard Business School classmate of a Huawei executive and interviewed them. I was astonished by the lengths you went to really get to all the details of this story, including a ton that just has not been talked about before.

Subscribe now

Eva Dou: Thanks so much. The process of digitization of archival material around the world was really helpful in this project. So many times during the process, I thought about how if nothing was digitized and I had to go in person and just read each page one by one, how long would this project take? It would have taken years and years more than it did. It really gave me an appreciation of the earlier generation of researchers who were doing everything in analog, in person, archival research.

Jordan Schneider: But they didn’t have Twitter to distract them. Eva, how long do you think you’d last at Huawei?

Eva Dou: Oh my goodness.

Jordan Schneider: Say your 24-year-old self. We’ll give you a little boost of energy.

Eva Dou: Well then good. I probably could hack it a couple years, but they are ferocious in weeding people out.

Jordan Schneider: They’d catch on to you eventually.

Eva Dou: Yeah, probably.

Jordan Schneider: Is there a moment or a meeting that you wish you could have been a fly on the wall for?

Eva Dou: There are a few points in their early history where they changed the company’s structure or ownership and I’ve always been pretty curious about what exactly those conversations were. They started out as a private startup — it was a pilot program for private startups. No one knew how long that would last. At some point they shifted to where they were almost more state-owned for a few years and then they went back to being a private corporation.

Kyle Chan: One funny thing that stuck out to me is that Eva, you have pointed out the pronunciation of Huawei among Americans. So many people say “Wawei” instead of “Huawei.” Once you pointed that out, I can’t stop hearing that. People do it all the time and I don’t know where it came from.

Eva Dou: Part of it is the company itself. At some point their PR team was going around teaching people to say the company name this way.

Kyle Chan: That’s so interesting.

Eva Dou: Which is a curious thing.

Kyle Chan: Wow. It reminds me of Hyundai trying to be like, “We’re not Honda, we’re not ’Hi-yun-dai,’ it rhymes with Sunday.” It’s tough.

Eva Dou: They thought about changing the name in the early days when they realized it was so hard to pronounce for international audiences and they batted a few different things around. But in the end it seemed too difficult to change, and we’re stuck with Huawei.

Jordan Schneider: One of the daughters is a singer, right? We have to do that as our outro music. But are there — did Ren ever talk about liking music? Did he have any favorite songs or genres or anything?

Eva Dou: Interesting. I don’t know off the top of my head about music, but he’s always been an admirer of the arts and that comes from his parents and this desire to be a cultured renaissance man. That’s something he’s really encouraged with his employees — for them to spend their spare time cultivating themselves, listening to different types of music, looking at fine art, things like that.

Jordan Schneider: Cool. Aside from the architecture thing, which people are clear on, are there painters or writers? Give me a sense of the Ren Zhengfei cultural constellation.

Eva Dou: Well, classical style art is something he quite appreciates. There are stories of him collecting paintings from around the world and having his employees help him bring oil paintings over from abroad. He’s talked about his family quite liking Europe, liking to go to Europe on vacations. His youngest daughter grew up largely between the UK and China. He often talks about these cultural things to make a point, as part of his outreach to different markets. He talks about international culture that he likes probably more often than domestic.

不要复制，而要超越。

——陌生女人1号 兔姐

亲爱的媎妹：

见字如面！

前段时间我写了一封信批判部分互联网“激进”女权是假激进真保守、假解放真极权，发布之后引起了不小的争议。不过除了上来就扣帽子辱骂我们的评论以外，其实还是有一些有益的批评的（见《从还有明天出发，谈谈什么是真正的激进》）。

于是我就在想，到底是什么引起了这种分歧，为什么有些人如此激烈地表示反对？后来想到，可能还是由于我们这里的性别平等状况实在太差了。所以我发文说“不要指责和审判女人”、“女人不是只有变强才配获得自由，自由应该是天赋人权”等等，很多人才会警铃大作。或者说，我们国家的女性真的太容易滑落了，甚至只要踏错一步就可能万劫不复，所以大家才对“尊重女人的决定/相信女性能自主选择”这种话十分警惕、很害怕我是在鼓吹“向下的自由”，因为真的有太多男人以「自由」之名引诱女人走进陷阱，然后敲骨吸髓了。

（我之前在豆瓣发过一个讲《芭比》为什么在国内水土不服的帖子也有讨论到这一点）。

而很多本土女权主义者如此急切地催促女性用个体行动反抗结构性压迫（主要是通过“非暴力不合作”的方式进行消极抵抗），大概也是因为现实确实水深火热，而制度性的改变也实在有点无望。也正因如此，她们可能把我的观点——不要把责任归咎于个体、忽视结构压迫——理解成了“鼓励个体彻底躺平”，所以才这么愤怒？......这样看来，我那篇文章激起的情绪/观点其实也是时代和环境的产物。

不过呢，尽管一定程度上理解了反对我观点的人（可能）的立场，且在批判美役和异性恋方面我们的观点有所重合，但我依然无法认同她们解决问题的方式。我认为，「强迫个体对抗结构」的解法最多只能让女性暂时摆脱部分压迫，却终究无法带来系统性的变革。

而且，在“逼迫”其她女人用自己认为正确的方式反抗时，一些本土女权主义者开始分裂女性群体并实施“多数人”的暴政，而某些看似反叛的路其实是不过对父权的镜像复制。这类做法与其说是Radical（彻底的、根治的），倒不如说是Aggressive（有攻击性的）——判断一个人是否激进的标准应该在于她和权力的关系，而不是骂人（包括骂“婚女”、“弱女”）时言辞激烈的程度。

更重要的是，现实不是非黑即白的，我不认可某些自称“激女”的人的行为，并不代表我就要给男人“接三胎男宝”了。。🚬（这是默认“激女”必然最正确，反对者就一定是男权了？另外，女性应该拥有生育自主权，所以不应该用这种话侮辱女人，不过大家明白我要表达的意思就好...）

换言之，在我看来，在「放任女性滑落并成为彻头彻尾的父权受害者」和「鉴定伥鬼和媚男女、审判女人的行为是否女权，以及通过辱骂“弱女”逼迫女人向上」之外，还有第三条路可选——也就是我的女权主义十要、十不要。

1. 要诚实揭露异性恋父权对女性的剥削，但反对用“婚驴”、“虜”、“伥鬼”等词语分化和羞辱女性。

异性婚恋对女性的制度性剥削是一个不争的事实。我们也曾写过很多文章批判异性恋父权，这里就不赘述了（见《婚礼如何体现性别不平等？》、《彩礼是物化女性吗？》、《恋爱脑根治指南》等）。

的确，恋爱脑的女性更容易对男人抱有不切实际的期待，在受害时依旧自我欺骗、下意识美化异性恋浪漫关系，最终结果则很可能是掉进一个又一个男权陷阱。

可是，常常感到“缺爱”并非女性的错，而是系统性失权的结果——女性自出生起就在家庭和社会遭到全方位的打压、极度缺乏自信和安全感，而浪漫爱在此刻趁虚而入，让女人误以为自己能从男人那里获得自己渴望已久的温情。正是因为长期缺乏支持和尊重，女性才会为了异性恋里的一点点“蜜糖”吞下砒霜。

因此，要想破除异性恋父权对女性的剥削，我们必须从根本上重建女人的支持系统，而不是辱骂由于各种原因顺应了社会主流结构的女性，或是简单粗暴地将她们开除女籍，通过「和受害女性割席」完成自保（这和理性探讨已婚和单身女性之间客观存在的矛盾是两码事）。

例如前段时间发生的小谢被家暴事件，很多女性在网上表示“婚女”的事情与自己无关，认为单女没必要“淌混水”、替“叛徒”争权。更有甚者用“婚驴”、“伥鬼”等字眼辱骂受害者。她们好像认为，只要不沾男➕和已婚朋友绝交➕对婚女权益不管不问，女人就不会受害。可是现实真是如此吗？

在我看来，与「在父权结构中陷得更深的的女性」割席并非女权主义，而是“独善其身主义”。更可悲的是，这还是一条注定失败的“独善其身”之路。

“单女”无法通过和婚女割席保全自己。这是因为女性利益是一个“一荣俱荣、一损俱损”的整体。除非离群索居，否则没有任何一个女人能够逃脱父权社会的伤害。

女人当然有独善其身的权利，但我反对将“独善其身”等同于“女性主义”。女权主义者不应该将任何一个女人开除女籍，剥夺她受到共情和帮助的机会。

对同伴变得宽容，就是从意识到我们命运相连的那一刻开始。无尽的远方，无数的女人，都与我有关。我们是你中有我、我中有你的共同体。

当我看到被家暴的妻子，我看到的不是投敌的“叛徒”，而是和我一样曾遭受过性别暴力的女人。

当我看到被家人“吸血”却不愿断亲的女大学生，我看到的不是活该被欺负的“软柿子”，而是系统性厌女下渴望温情的孩子。

当我看到带年幼的男孩进卫生间的女人，我看到的不是“挤压单女生存空间”的“伥鬼”，而是因丈夫缺位、丧偶式育儿无比疲惫却依然在努力生活的母亲。

还有，我不会和男人结婚，但我和姐姐都是辛苦工作才能赚一点工资的女性劳工；我不想生育，可我和妈妈都是在父系家庭中受过轻视的个体；我不会做家庭主妇应聘“妈妈岗”，但我也一样要面对职业零工化、职场女性边缘化的严峻现实。我们都是在父权的围追堵截中依旧活出了自己人生的战士。我生活里的任何一点不如意，都不是另一个女人的错。

女权从来都是集体的运动，她不应成为个人主义的极致发挥。在为全体女性争取权利的过程中，我们要弥合而非撕裂，尽量求同而非排异。毕竟，女人间的不同不应成为加剧对立的借口，而应该是无尽创造力的来源、是伟大变革的起点。

P.s. 最后想说一点，我理解强调“单女”是为了反抗“单身女性的污名化”，有一定进步意义，但我依然不喜欢这个词。因为无论我们是否已婚/在恋爱，我们怎么可能是“单”的呢？人永远处于各种各样的关系中。脱离了和世界的关系，“我”将不复存在。所以我认为“单女/婚女”的划分依然是在加固异性恋父权以婚姻为中心的社会结构，且再一次贯彻了「分类/分化女性、挑起内部争端，从而让真正的压迫者美美隐身」的父权逻辑。私以为，真正具有颠覆性的行动不是在结构内和压迫者“对打”，而是彻底打碎这个以「婚否」为基础的分类系统。

由于篇幅限制，我在这封信中只阐述了《十要十不要》的第一条，余下内容及结语会在下封信里和大家分享。感谢姐妹们的关注！☺️

陌生女人1号 兔姐

二〇二五年六月十六日

Read more

Renaissance Philanthropy — in my opinion, the most exciting philanthropic venture in the US — is getting a one-year check-in. Kumar Garg first appeared on the show right before I went on paternity leave, and now we’re back for round two. Before founding Renaissance Philanthropy, Kumar worked in the Obama Office of Science and Technology Policy and spent time at Schmidt Futures.

We discuss…

• How Renaissance catalyzed over $200 million in philanthropic funding in its first year,

• The goals of the organization and how it has responded to Trump’s S&T funding cuts,

• What sets Renaissance apart from traditional philanthropic organizations, and lessons for China-focused research foundations,

• AI applications in education, from tutoring to dyslexia screening,

• Donor psychology, “portfolio regret,” and how to build trust within a philanthropic network.

Listen on Spotify, iTunes, or your favorite podcast app.

The Hedge Fund Model of Giving

Kumar Garg: I like that this is becoming an annual tradition.

Jordan Schneider: Yeah, we've got to set goals this year, and we can hold you to them in 2026.

Can you start off with the 101 of Renaissance philanthropy, and explain how the thesis has played out over the past year?

Kumar Garg: I’m grading myself here, so this is a biased view, but it’s been a very strong year. When we were launching the organization, we were trying to do something different.

Most philanthropic organizations exist in a single model — they work for a single donor. That donor has resources, whether they sit in a foundation, in their DAF, or as personal wealth. The organization works for them, asking how much money they want to give and on what topics, then runs their philanthropic giving.

There’s another class of organizations that are basically the people spending the money — researchers running labs and doing high-quality research. The philanthropic system has mostly operated with givers and takers — folks operating these organizations and folks doing high-quality work.

The idea behind Renaissance Philanthropy was to sit in the middle and style ourselves more like an investment fund — more like what happens in the world of finance. The folks who are the holders of capital, who have the money, mostly don’t spend their time trying to directly deploy that money.

If you work as an LP for a family office, you might have a team of 10, 20, or 30 people, and you’ve got billions of dollars to deploy. What do you do? You go out there and find intermediaries — private equity funds, hedge funds, venture capital funds, or other experts in particular sectors and areas. You give them the money, and they deploy it on your behalf to help you earn a return.

Philanthropy has mostly operated differently. It’s odd, but it’s historically contingent. The investment world moved toward specialization from the ’70s onward, while philanthropy went in the direction of direct giving. You have really large philanthropic organizations, often well-staffed by experts, that do the giving.

The challenge is that there’s a subset of donors who want to build large organizations, and there’s a large set of donors who don’t. The ones who don’t have been sitting on the sidelines. What ends up happening is maybe when they retire, they build an organization, or when they die, they bequest it to a nonprofit or university. That leaves a lot of value on the table.

The idea of Renaissance was, on various science and tech topics, can we do what an investment fund does? We write down a thesis for three years, five years — we want to achieve this goal. We recruit a field leader to run that fund, then treat the donors almost like LPs in a philanthropic fund. We’re not giving them a return back, but they’re putting money to work against that strategy.

A year ago, when I told this story to people — “I’m going to create an organization that does this” — the operative advice was, “Good luck.” You’re going to cover the waterfront across AI, climate, and economic social mobility. You’re going to take on this massive fundraising goal. That seems like a very hard way to operate. You have no natural advantages — you’re not spending one person’s resources. You have to raise the money and deploy it. It seems doubly hard.

What I was interested in was growing the pie — can we use this model to bring new donors in?

A year in, the early grade is strong. We’ve been able to stand up multiple philanthropic funds. We have a fund using AI to accelerate the pace of math research. We have another fund using AI to deliver public benefits better. We recently launched work on climate emergencies — can we solve for runaway climate risks and increase the technology readiness level of various climate technologies?

We have different funds in various areas. Each has this basic structure — they have a thesis they’re driving against, a field leader running it, and we’re recruiting donor money against that strategy.

What I’m hoping for is that this starts to become — not the only way philanthropic giving happens — but a much more credible path. This allows more donors to be active without necessarily having to take on all the operational load themselves.

Jordan Schneider: You’ve launched this in a particularly precarious time for the future of science and research in America. We’ll get to your takes on the policies in a second. But I’m curious from a donor appetite perspective — what has all the tumult in universities and government funding done for those billionaires sitting on the sidelines, giving just 1.2% of their assets annually to philanthropy?

Kumar Garg: That’s a great question. I don’t have one system-level answer — it’s a frequent question I get about how donors are interacting with the environment. They’re interacting the way most people are — there’s an incredible amount of chaos and news every day, leaving many frozen in place.

It’s relative. Government has pulled back on research funding in the short term, causing significant churn. Industry is also holding back as companies figure out what’s happening with tariffs and everything else. Philanthropy, comparatively, is cross-pressured but hasn’t engaged in the same pullback.

There are donors we interact with who are certainly reformulating their strategies. There are others who, as I mentioned, are interested in compelling ideas and looking for those just like anything else. I haven’t seen an overall pullback — just more of a sense of “Is this idea good in itself, even if government didn’t help at all?"

Jordan Schneider: Can you put in order of magnitude the hope of the new model you’re trying to manifest against, I don’t know, NIH budgets being cut by a third?

Kumar Garg: There’s no world in which philanthropy fills the gap. If you step back and ask how the US built its lead — well, the US spends on the order of $200 billion a year on R&D. Once you include basic and applied research across DoD and civilian agencies, that’s an order of magnitude more than philanthropy spends on research.

The place where these new models will get traction is that how you organize scientific organizations has suddenly become much more of a jump ball. It used to be that the academic bundle — being at a top university — had everything stacked on top of itself. You could get really good talent that way — graduate and post-graduate talent, great students. You build your lab there, do cutting-edge work. Usually, the university gives you flexibility to do many things on top of it. If you’re an academic doing well at cutting-edge research, you could do that within the four walls of a university.

Some researchers have left universities and built what are basically academic research labs outside the university. You’ve got the work that Patrick Collison is supporting around the ARC Institute, the Flatiron Institute that Simons supports. You’ve seen the FROs that Convergent Research proposes. For a long time, that’s been a very alternative path — rare to do, often requiring you to figure out what happens to your university affiliation and how it changes your career path.

If you’re a researcher who’s ambitious and wants to do big projects, whether you’re doing them within the four corners of the university or in your own nonprofit research lab and partnering with universities becomes more of an open question — especially in a world where university funding might fluctuate based on political developments.

I don’t know how that will play out over time, but we’re three months into a deeper shift in how institutional financing will happen. That could have big implications generally. On net, if the federal government doesn’t play its important role in funding research, it’s all a net negative. If federal funding returns to a healthy level, researchers will still take this as a wake-up call to think about structuring their research organizations to be more resilient against systemic shocks.

This episode is brought to you by ElevenLabs. I’ve been on the hunt for years for the perfect reader app that puts AI audio at the center of its design. Over the past few months, the ElevenReader app has earned a spot on my iPhone's home screen and now gets about 30 minutes of use every day. I plow through articles using Eleven Reader’s beautiful voices and love having Richard Feynman read me AI news stories — as well as, you know, Matilda every once in a while, too.

I’m also a power user of its bookmark feature, which the ElevenReader team added after I requested it on Twitter. ChinaTalk’s newsletter content even comes preloaded in the feed.

Check out the ElevenReader app if you’re looking for the best mobile reader on the market. Oh, and by the way — if you ever need to transcribe anything, ElevenLabs’ Scribe model has transformed our workflow for getting transcripts out to you on the newsletter. It’s crossed the threshold from “95% good” to “99.5% amazing,” saving our production team hours every week. Check it out the next time you need something transcribed.

Jordan Schneider: I observed the EAs being very excited about how many lives they save based on the bed nets they bought. Then you net that out against USAID no longer existing and all the human suffering that’s going to come out of that. The correct calculation may have been to spend all your money lobbying Congress to get people to focus on this.

I think both of us are pretty aligned — we’ve done other shows on immigration policy, university funding, and what’s happening to NSF and NIH budgets going forward. But why do Renaissance philanthropy when Kumar could be spending 100% of his time in D.C. banging on doors and trying to make it 5% more likely that we get an extra $10 billion a year for this stuff?

Kumar Garg: That’s a great question. Being policy-adjacent is generally very high ROI. No matter how you run the numbers, policy advocacy — especially on science and tech topics — punches above its weight, regardless of what you’re doing. It’s probably why I spent time in government as a policy staffer. It’s partly why, no matter what I’m doing, I’m constantly interacting with policymakers and making the case. It’s also why, when funders ask to what extent advocating on behalf of the research community should be part of their work, I’m strongly supportive.

The reason we structured Renaissance this way is that I wanted to specifically think about growing the pie of philanthropic funding because I thought no one was doing it. There are organizations working on policy advocacy. Very few organizations were trying to bring new donors into the mix.

We would be failing as an organization if we weren’t constantly thinking about how our work could impact shaping the debate on the future of R&D funding. We try to be in conversations with both Congress and the administration, as well as policymakers up and down the ladder, to say, “Here’s why this work matters, here’s why the future matters.”

Part of the new models we’re funding — whether it’s things like FROs or AI accelerating science — is to make the case for why investment should happen. Many of the ideas I’ve funded over the years, you can see echoes of in the new Heinrich legislation around accelerating science through AI, where they’re talking about ensuring these AI investments can actually accelerate the pace of science using new models.

Philanthropy, when done well, opens the aperture for what funding could do. Hopefully we’re playing that role. One area I’d like the conversation to reach is moving beyond this dialectic between “science is important” and “science needs to be dismantled because it made mistakes.” I’d like us to reach a place where we recognize there are important things we can do to help reform how we do science. We should bring more discipline to trying out new ideas, bringing in new funding methods and new voices, and reflecting on past mistakes — while also remembering that the investment agenda around science is critical for its utility. Hopefully we can be part of that dialogue.

In some ways, you’re pushing on something I think about all the time — I am a policymaker at heart. The deep utility of that shouldn’t be forgotten in my story.

Jordan Schneider: All right, I think the answer I would give to you is this is the federalist model of policymaking almost — as you said, the inventiveness that you guys can come up with from a form factor and discipline perspective when it comes to doing science and technology research is the type that’s weird enough that it’s not happening in government. But also, two, three, four years down the line, once you guys have some really awesome case studies, these are the sorts of things that can then get 10x or 100x in our gorgeous NSF circa 2027 that has been remade to fully align with the Kumar vision of how change gets made.

With that stance of optimism, let’s talk a little more in detail about some of the projects you guys have stood up. Take us on a little tour, Kumar. Where do you want to start?

Kumar Garg: I’ll go through a couple of the funds and projects we’ve launched. Just to give people a taste of how the model works, let’s start with our work in AI. Our operating theory in AI is that we’re living through a period of huge capability overhang. The idea is that the core technology is rapidly developing, but the number of people, projects, and overall work that actually applies these tools toward actual hard problems in society is really small.

I’ll give an example. We have an AI and education fund specifically focused on how AI can accelerate the pace of learning outcomes. If you follow social media and others, there are many people who write and talk about AI in education. It would give you the sense that a lot of people are working on AI in education. But if you actually dig into the space, the number of actual technical experts who have knowledge of both how education works and how AI works is still shockingly small.

We run something called the Learning Engineering Tools competition — an annual competition that invites tool developers to present cutting-edge ideas that use AI to actually advance learning outcomes. We’ve been running this competition for a couple years. I started it even before Renaissance and then brought it into Renaissance. That competition is the only large-scale ed tech competition in the world. It still blows my mind. No one is out there in a systematic way asking for sets of ideas from people who want to build AI for education.

We have another part of our AI education portfolio that specifically thinks about moonshots — what’s a really hard problem in education that AI could solve? We picked middle school math. It’s really important for advancing to future degrees, and students really struggle with it. We said, can you actually emulate the results of high-dosage tutoring, which the number of studies that J-PAL and others have done show can really double the rate of learning for students in math? Can you do it under $1,000 per kid — bring it under what would make it such that you could offer it to every kid?

We have that running as a program. We have seven teams in the program. We have two teams that are actually on track to potentially accomplish this goal.

Jordan Schneider: Which is wild, right?

Kumar Garg: When those teams are working on it and we ask them who they’re collecting lessons from, there’s not a big field they can go out to. When they go out and interview the AI labs — the ones that get written about every day — those AI labs talk about education, but they don’t have in-house education teams that can actually help these teams.

The biggest piece I would always say to people is that at the coal face, there’s tons of room to do work because when you actually start to work on it, you realize that the number of people who are actually working on it is shockingly small.

We’re now starting to explore our next moonshot area — should there be something that basically looks like the intersection of AI and early learning? Can we actually build a universal screener to best guess if a child is off track when it comes to early language development by having them speak into a device? There’s a bunch of interesting work happening in this area, but we don’t actually have a way to diagnose early learning challenges like dyslexia just by having a student speak into a device. It could vastly increase our ability to help them get to a speech pathologist, get back on track, and be reading by third grade, which is critical to future reading and learning.

That’s just one track — AI and education. That’s just one compelling thesis.

Jordan Schneider: Obviously AI is going to matter for education. Hard to find people to argue about that. Talk a little bit about finding the donors and finding the teams. What was the work that you guys had to do to make and launch this work?

Kumar Garg: What’s been interesting is it has been hard work for us to build out the team because the number of technical experts who actually know both things — AI and education — is small. We have slowly built out a team of ML experts who have educational backgrounds, basically. We call it a hub model. We basically have created an engineering hub and we recruit technical experts into it that specifically have this technical background.

I have somebody on my team, Ralph Abboud. He has a machine learning PhD, and he did his thesis on graph theory. He’s not an education expert, but we brought him onto the team. He has been working with a lot of these educational teams that we brought in. What’s interesting is that his ideas on what kind of language models they should be building are really good. It took him some time to level up on the education side, but now he is one of their highest value contributors, even though he sits on our team and he’s contributing there.

There is this transition where you can build up talent that sits across these two areas. But in AI and education we had to mostly build it. It was hard to immediately find directly. Now we have a constellation of these AI and education experts, some of which sit on our direct staff, some of which sit inside these teams that we’re betting on. It’s been great. Now we have a field team that can really go after more problems.

On the donor side, we’ve really lucked out. We found that our core donor for a lot of this work has been the Walton Family Foundation. They have a long history in funding education. What’s been interesting is that they’ve been interested in investing more in what they call their innovation portfolio, but didn’t know how to necessarily bridge that technical divide — if we’re going to do more in this area, who are going to be the technical experts who will actually do it? That had actually kept them more experimental. But their partnership with us has meant that they have become way more ambitious on how much investment they want to make on this technical AI and education lens.

That’s our core thesis — can we be the permission structure for donors to go much bigger on innovation? We’ve seen that in other areas. Slowly their support is causing other donors to come in as well. That’s basically whether you’ve been a long-standing donor but not active on science and tech topics, or you’re an early donor altogether.

Jordan Schneider: What’s the RenPhil management fee?

Kumar Garg: That’s a good question. We build our cost recovery into each fund. Usually the way that works is if we’re operating multiple funds, each fund has money going out the door for actual deployment grants, but then we’re building in our cost for the actual staff operating the funds, whatever services and technical support services we’re providing, the work we’re doing to partner with various funders, as well as our overall studio support. It varies fund to fund, but donors have found it — compared to having to try to do this themselves — much more actionable. For us, we want to build a thriving organization. We don’t want to cut corners. We want to build an organization that can both operate those funds and also be looking for the next ones.

Jordan Schneider: Does anyone complain about that?

Kumar Garg: The way it comes up is there’s a type of donor who actually has the answer in their mind. They’re thinking, “I think this needs to happen.” Really, what they’re looking for is an operating partner to just do that — “I want a conference, I want a workshop, I want to fund these three organizations.” Our model is we’re the product. You’re actually hiring us to go build out the strategy, recruit the team, deploy. If you actually have the answer in your head, we often tell them we’re way too fussy for that model. You should just — there are much simpler ways you can operate. That’s where the delta comes in. If you already have the answer in your head and you’re just looking for a partner to execute for you, we’re probably not the right fit.

Jordan Schneider: Yeah, you said this on another show. You were like, “We are for donors to take off the cognitive load.” The idea of being: yes, if I have $10 billion, maybe I’ll allocate $1 billion investing in stuff I know and think I have some subject matter expertise in, but I still have to put the other $9 billion somewhere — probably not cash. Yes, I am comfortable paying a hedge fund or financial advisor a management fee to do that.

Kumar Garg: A big part of it is opt-in. People don’t know what journey they’re on, but what they worry about is: am I going to feel stuck? A lot of folks end up not getting active philanthropically because the decision feels weighted by getting stuck. “Okay, if I hire somebody, and then six months from now, I decide maybe I want to change direction. Now I’m going to have to let someone go.” People hate that.

Or, “I met a researcher. I liked their research. I gave them one grant. But now they’ve reached out and said, ‘There’s so much happening in the world — I’ve lost funding from the government. Can you double the grant?’ I was just giving them a grant because I met them and I thought they were great. But now they’ve sent me a note that they might have to let go of postdoc students. Now I’m in this uncomfortable situation. If I say no, I feel I’m hurting them. If I say yes…”

People have all these experiences where they feel uncomfortable with the relationship they have around their resourcing. Rather than causing them to work through it, they actually hold back. One of the things we say to them is that our model is one where we’re the ones making the decisions. We’re going out there, finding researchers, finding projects, developing strategies. You can be as involved as you want. You want to be meeting the researchers? That’s great. You want to be learning from the strategy so you can do direct giving down the road? That’s great. But if you also take six months off and decide, “That was great, I learned for a few months, now I’m off doing something else,” — nothing will stop. We’re a fully operational organization that will execute on everything we said we were going to do, whether you’re involved or not.

It just takes the pressure off. You can opt in if you want to learn and be involved, but you can also choose not to. That actually frees them up to want to learn without the “Am I about to get stuck?” That sounds very psychological, but people forget how hard it is to get going on things — “I’m going to start to work out more. I’m going to start to do this.” Starting is hard. We want to make starting easy by saying you can provide a lot of value into the system without necessarily having to own all of that execution.

Subscribe now

Jordan Schneider: There are a lot of pieces of people’s jobs that seem like more and more AI can chip away at or enable or launch or whatever. It’s interesting because some of the things that you guys are doing — you have these seven playbooks, ways you can tackle problems — I would love to upload seven of those to ChatGPT and say, “Here’s my problem in the world,” and the AI can help me pick through which one. But talking someone who’s really rich, who’s feeling uncomfortable about giving money, into starting to donate philanthropically in a serious way for the first time seems like one of the more human things where there’s really going to need to be the friendly Kumar Garg, who now has a nice microphone he can do Zoom calls with to — what did Derek Thompson say? — whisper the dulcet tones of comfort and competence in their ear in order to get them on this path. I don’t know, it just seems like a very human thing that you’re engaging in on the donor engagement side. I’m curious for any reflections you have on that.

Kumar Garg: We are very curious about how much of our own internal processes we can automate. Why not? We sit next to AI, we should be thinking, we should be dogfooding. The place where we’ve seen it already provide some value is just what you would consider baseline automations. There’s a lot of grantee reporting that you should be able to do automations on. We’re definitely interested in: hey, we have a hunch around a thesis in this area. Can you do a research report and tell me what’s the relevant stuff to know? Scoping. We’ve even used it for, “Hey, we might do an RFP on this topic. Who are some researchers who should apply?” Sometimes we found some interesting suggestions for researchers that we should affirmatively reach out to.

I will say that we’re still far away from it actually helping on anything that we would consider high stakes. As you’re saying, a huge amount of what we’re doing is making something that feels like a trust fall. Hey, this is an important decision, but one where people who take their job very seriously and put their own personal legitimacy behind the work is an important part of it. When we screw up, it’s on us. We stand behind all of the work. People appreciate that these are serious people who stand behind the work they’re putting before them. They’re not some faceless intermediary. I don’t think maybe that will change, but that’s an important part. Even on the information you should know about various people and stuff, these current AI models are not that great.

The place where I — we have this intuition that there should be parts of being a program leader that you should be able to have an AI assistant for. Right? You take more and more of the task of being a program leader or fund leader and be able to say, “Okay, I want to do a workshop on this topic. Generate me an agenda for how you would run the day.” It generates based on — it takes a bunch of your past workshop flows and generates a sample workshop design. How much of that can we create so that we really could get to a point where a program leader or fund leader is basically able to operate without that much additional support? Obviously we need to create some cross-cutting support that I’m interested in. But the chance that we’re going to get to an AI advisor — we’ll have to wait.

Jordan Schneider: The trust fall works in multiple directions. You need researchers to give up their PhD programs or leave their current positions to spend half their time with you, while simultaneously needing donors to provide funding. Having a recognizable face with a proven track record and skin in the game on the other side of that equation is something that won’t disappear anytime soon.

Kumar Garg: One thing we debate internally is that much of my workflow relies on tacit knowledge. When I’m talking to somebody about their work, twenty minutes into the conversation, I’ll say, “Tell me more about that. Why is the field stuck on this point?” They start describing it, and I realize that if there were a canonical dataset with specific dimensionality, it might solve the problem. When I ask why that doesn’t exist, they explain it’s locked up somewhere.

Part of me constantly strives to figure out how we can make this process more explicit. When we recruit somebody new to the team, they ask if they can sit in on my calls and watch me work through problems with researchers. There’s something that feels wrong about just saying, “You develop this intuitive feeling for opportunities — just pull on that thread.” The more we can transition from tacit to explicit knowledge, the better. Right now, we operate on an apprenticeship model where people learn by doing and being embedded in these structures, but I don’t think that has to be the endpoint.

Jordan Schneider: Much of what you do involves human matching — putting people in touch with each other. While you could potentially feed all your past calls into an AI system, there’s an emotional and personality matching component that you’re handling. That remains very much a human process that current models aren’t quite ready for yet.

Kumar Garg: The matching capability changes over time, but I think what people really value when I connect them is that I took time out of my day to think the two of them should know each other. That’s the actual signaling value — that my time is precious.

Jordan Schneider: Slight tangent, but if people want to establish trust and rapport, the first thing they should do is spend $150 on a microphone for their Zoom calls. That’s my recommendation for everyone. When I do my calls, I sound the same as I do on my podcast, and people respond positively. You feel like an embodied person rather than a compressed, distant voice through AirPods. It’s advice for anyone who wants to make connections and raise money from billionaires on Zoom.

Kumar Garg: I’ll echo that point, though I haven’t practiced it myself. There’s an old political adage about microphone technology. If you look back at politicians historically, there was a time when microphones couldn’t pick up subtle intonations well — speakers were just projecting loud sound. Once microphones could capture subtle intonations, politicians who excelled at that style of speaking began to dominate.

People point to President Clinton as an example — he was exceptionally skilled at subtle microphone use. I remember reading a paper arguing that this was possible because the technology had improved to support that communication style. Politicians offer good lessons here because ultimately, communication is central to building trust with the electorate. [Here are papers that explore this]

Jordan Schneider: Absolutely. If you listen to old clips of Warren Harding or Teddy Roosevelt speaking, they’re basically screaming into microphones — which was necessary at the time. Teddy Roosevelt was exceptionally good at that style of projection. You needed to be very loud to stand on a soapbox and reach people twenty rows back. Now we have the dulcet tones that modern microphones enable.

Here’s another fun fact, Kumar — the microphone I’m using has been manufactured for sixty years. It’s remarkable that microphone technology for voice pickup has essentially reached its peak — we’ve basically maxed out the capability.

Kumar Garg: I should try to find that paper I mentioned. I wonder if it’s about mobile situations — being in some random union hall where you need to set up a handheld mic in front of a politician. Perhaps that’s why microphone technology improvements became so important. That’s an interesting angle.

Jordan Schneider: I’m curious about that. Alright, shifting topics — Yascha Mounk recently wrote on Substack about attending gatherings, conferences, and dinners where leaders of America’s biggest foundations have been strategizing how to defend democracy. Few were as openly devoted to extreme forms of identitarian ideology as they might have been a few years ago, but the reigning worldview at the top of the philanthropic world assumes little has changed since summer 2020.

The general consensus holds that voters turned to Trump because American democracy failed to deliver for the “historically marginalized,” and the solution supposedly revolves around “mobilizing underrepresented communities.” The most urgent imperative is to “fight for equity” and “listen to the global majority.” I find this perspective fascinating. Kumar, as someone who’s a new entrant to this world, how do you interpret this?

Kumar Garg: Several different dynamics are happening simultaneously. Some philanthropic responses resemble dinner table conversations — people sharing hot takes about why the election unfolded as it did and offering their views on America or the American people. Much of this sounds as random as hosting a dinner party where guests share their political opinions.

There’s also a genuine state of confusion about what’s happening. The first hundred days of the Trump administration have been exceptionally active across a range of unexpected areas. Many people expected it to feel similar to the previous Trump presidency, so they examined their portfolio of issues and anticipated certain outcomes — but that’s not what materialized.

Regarding how much people are actually rethinking their approaches, that’s a valid question. The most immediate reconsideration I’m seeing centers on identifying what we’re missing. This is particularly evident in the science community, which is confronting devastating across-the-board cuts. Researchers are losing funding, university funding is being paused, and graduate students working on topics relevant to competitiveness are having visas revoked.

The community is asking, “We don’t remember this being a major campaign debate topic, so how exactly did we become a political football?” There’s extensive questioning about what we’re missing — whether there was a conversation we weren’t invited to where we were being discussed, and what we’re failing to understand.

While donors with certain political orientations likely won’t change their fundamental positions, the confusion centers less on American domestic politics and more on why certain issues became contentious. Foreign aid is a good example. The extent to which US foreign aid posture and system effectiveness became campaign issues wasn’t apparent during the election cycle. People are asking whether we missed a major debate that suggested the United States should dismantle its leadership on these topics overnight. What policy debate did we miss? That’s where much of the confusion originates — donors being puzzled about the sources of these developments.

Subscribe now

Jordan Schneider: There’s an interesting dichotomy between foundations with living leadership and those with deceased benefactors. Gates recently indicated he would spend down his wealth faster than previously planned, presumably responding to recent events. When foundations have active leadership that engages with current events, they can be more responsive.

However, when you have flagship philanthropists who died seventy-five years ago and organizations that have built programs around worldviews that are no longer relevant or don’t meet current demands, pivoting becomes much more difficult. You encounter institutional blockers, boards, and established structures, whereas a living person with decision-making authority can simply redirect resources.

Kumar Garg: That’s definitely part of it. The piece I’d add, which connects to our Renaissance model, is that people underestimate how much philanthropic organizations become tied to their existing programs. This isn’t necessarily negative, but consider the process: you spend two and a half years scoping a program, conducting field research on topics, then executing a national search for program leadership. You recruit and convince someone to relocate for the position, provide coaching for test grants, and they’re six to twelve months into the grant cycle.

If you then decide the world has changed and want to cut the program — after issuing press releases announcing this as a major new strategic direction — it appears chaotic. People develop what I call “portfolio regret.” If they could start fresh today, they’d create different programs than what currently exists.

One argument we make to donors is structuring themselves more like limited partners, deploying money into funds where all capital remains fresh and available. You avoid the incumbency problem where team members question every pivot attempt because they have specific responsibilities you hired them to fulfill.

Flexibility requires both mindset and structure. Donors sometimes create substantial built-in costs and barriers to pivoting when they could maintain lightweight, flexible operations if they chose to do so.

Jordan Schneider: That’s fascinating because you’d think giving money away would be straightforward — you should be able to distribute funds however you want. But the emotional sunk cost around philanthropy wasn’t something I had necessarily considered.

Kumar Garg: This creates a situation where people spend enormous amounts of time operating like a duck swimming on water — their feet are moving rapidly beneath the surface while they try to keep the strategy looking consistent above water, all while changing the actual content underneath to pivot to current circumstances. This leads to significant conceptual confusion because you claim to have always had this program, but underneath, the program is completely different because the situation changed.

Part of why I favor the philanthropic fund model is transparency — what’s on the cereal box is what you get. It’s a three-year fund with specific objectives that will begin and end. Maybe it’s not perfectly timely, and that’s fine, but your new program can be timely. The alternative — constantly maintaining broad programs that you’re perpetually reworking underneath — makes evaluation nearly impossible. If I ask whether a program has been successful, people respond that the program has been changing constantly. This makes it extremely difficult to evaluate it as a focused initiative that ran for a specific period with defined goals. Did we achieve those goals? People simply don’t engage in these basic evaluations.

For example, I was speaking with a donor and pointed out that in the investment world, people prominently display their successes — putting “first check into [major company]” in their Twitter bios to demonstrate their betting acumen. I asked, “Who are the ten best program officers in America? Who on the philanthropic side has been the most effective check writer?” They responded, “How would we even know that?” Even with qualitative measures, wouldn’t you want to identify the best check writers?

A fund model, even for philanthropic goals, enables more honest assessment. You can say that fund paid out successfully, that one was moderately effective, and another one failed completely. The person who led that fund can then take that track record to their next position as a legitimate career advancement. We deny people this opportunity when we maintain the fiction that we’ve always had these programs run by different people with slightly different strategies. This obscures rather than clarifies outcomes. We could simply be honest: we executed that fund, now we’re doing something new with a clear beginning and end.

Improving China Research + Why Bother?

Jordan Schneider: It’s remarkable. I’ve spent considerable time on foundation websites researching whether organizations like the Ford Foundation might fund ChinaTalk. They have mission statements about “democratizing equity,” which is admirable — I agree we should advance democracy and create more opportunities for people. However, the problem arises when you’re only a passive recipient of pitches. You’re essentially letting grantees define what success means, and the counterfactual becomes very difficult to assess. The organizations you’re funding would probably exist whether you give them $100,000, $500,000, or nothing at all.

The alternative approach would be starting with specific objectives — “We are trying to achieve X by Y timeline,” then working backward from that goal orientation to identify the people and organizations who can take your money and provide the highest probability of achieving that outcome. This approach is far more strategic, and my frustration isn’t about the politics of how they set their goals — it’s that they need to engage more seriously in the process.

Kumar Garg: Let’s work through this together. Imagine you and I were designing a fund model versus a program model for increasing collective intelligence on US-China relations.

There’s a vague way we could structure this — “This is the US-China program. It will have three tracks — funding scholars studying China, engaging policymakers about those insights, and warehousing data and research publications on these topics.” This resembles how most programs operate: they establish a broad framework with several tracks, then people apply under those categories. But if I asked what constitutes winning or how we’d know this program succeeded, you’d probably say, “Well, people applied and we distributed grants."

You could execute the same concept with much sharper focus by asking, “What would success look like in three years?"

Jordan Schneider: Exactly. I want ten books written that are so thoughtful and essential to the future of US-China relations and American policy that Ezra Klein would be compelled to feature these authors as guests because the thinking these grants produced is indispensable. Then we work backward from that goal to determine the budget.

We’d estimate the costs — ten books, assuming a one-in-five success rate for people with strong proposals to execute effectively, calculate the pipeline requirements, and arrive at a number that gives us a 75% confidence interval for producing those ten books by 2029.

Kumar Garg: Exactly. That approach feels like a sound tax strategy built on what resembles a very tight OKR, which you may completely fail to achieve. You might be wrong, but it’s precise. Then you build your strategy around that goal.

You find someone to execute it — let’s call it the Jordan Fund. If you succeed, people ask, “Jordan, how did you pull that off? You wrote down this ambitious goal, built a strategy around it, and executed successfully. You’re clearly skilled at this.” When you pursue your next initiative, you can say, “I ran this fund called the Jordan Fund. We set this audacious goal to produce ten bestsellers on US-China relations, and we achieved it."

This feels much more tangible as an actionable strategy — something that field leaders can point to as real-world impact, even if it fails. Let’s say you only achieve partial success — you still have concrete lessons. Compare that to “the US-China program makes some awards and does some things.” How do I assess whether that’s working?

Jordan Schneider: It’s remarkable because the market is so powerful — you can’t get away with this approach when running and scaling a business, especially when taking other people’s capital and trying to generate positive returns for them.

What’s curious, Kumar, is that very capitalist people become surprisingly touchy-feely when it comes to philanthropy. There’s an emotional layer where they think, “This is giving, so we shouldn’t apply business mindsets and OKRs.” It feels somehow dirty to them.

Kumar Garg: Here’s what’s important — we need to distinguish between current donors and potential donors. People focus extensively on today’s active donors, but if you examine the statistics on potential giving, current donors might represent only 1-2% of the actual addressable universe.

The question becomes, would we attract an entirely new class of donors if we brought this level of rigor, precision, and targeted approach to philanthropy? This would feel much more familiar to their professional experience.

Why don’t existing donors demand this rigor? I believe there’s significant pent-up interest in this approach. People oscillate between thinking, “Since this isn’t about making money, we’ll substitute with having a really complicated theory of action — that’s where we’ll apply our intellectual capacity.” I often say that just because something has numerous boxes and slides doesn’t substitute for having a clear attack vector.

Jordan Schneider: Exactly — rigorous thinking. The median nonprofit worker is about five times more likely to be socialist than the average person, so perhaps people more attracted to touchy-feely logic are simply concentrated in current organizations.

Kumar Garg: Sometimes the nonprofit and philanthropy sectors spend too much time engaging in collective mission statements, as if shared purpose alone is sufficient. But we actually have distinct roles to play, including making high-quality decisions about where to deploy finite resources. Because money is limited, you must make decisions strategically and place informed bets.

This may feel reductive, but it’s actually the responsibility inherent in this work. You must be a responsible steward because high-quality decisions produce more good. People sometimes struggle with that reality.

Jordan Schneider: Without high-quality decisions, you end up with USAID getting canceled. That’s our current reality. Organizations that weren’t evidence-based and couldn’t effectively justify their impact had some good projects and some poor ones. They faced criticism from small but vocal movements — organizations like Unlock Aid, whose founder we featured a few years ago and will have on again — arguing for more rigor because there was substantial waste and inefficiency.

If you let these issues fester too long, consequences follow. I don’t want to say universities, the NIH, or the NSF “had it coming,” but one of the best defenses you can have is a tight, well-justified organization that can stand up for itself.

Kumar Garg: I don’t want to engage in victim blaming, and I don’t want to excuse what I consider sometimes bad-faith behavior. However, your point about systemic advantages is valid — caring deeply about systemic impact and bringing that rigor and constant evaluation is useful for the work itself, but also valuable when those political fights emerge. You can say, “Look, we’re building something substantial."

In some of these cases, who knows what impact rigorous evaluation might have had. We’re living through unusual times, but I believe we’re gaining traction because there’s significant pent-up demand for this approach.

Jordan Schneider: Good. That makes me feel somewhat better, I suppose.

Kumar Garg: What should donors know about China? That’s my question for you.

Jordan Schneider: The original impetus for ChinaTalk was thinking about long-term national strategic competition and competitiveness from an industrial systems and technology perspective — identifying things people could do to nudge outcomes in liberal democracies’ favor. During the Biden era, I observed errors that legislation and executive action could fix with modest improvements — 5% here, 10% there. A sophisticated understanding of what’s happening in China could meaningfully help squeeze that extra 10% out of various decisions.

However, the policy changes we’ve witnessed over recent months regarding long-term strategic competition — how the US relates to allies, approaches to global nuclearization, science and technology funding, and immigration — are much more fundamental. Getting to a better place doesn’t require understanding what made BYD successful, how SMIC is developing its chips, or even China’s new AI policy. These are much more basic issues.

The thesis I operated under during the Biden era was that deeper, more considered understanding of China would lead to smarter policies. That’s now become a sideshow compared to more fundamental questions. If we accept the base case that science is important and immigrants are crucial for better science, then we should pursue those priorities directly. I would choose that approach ten times out of ten.

Returning to the order-of-magnitude questions I asked at the beginning, I would choose a NATO that functions as a genuine alliance ten times out of ten over determining the right tariff level for Chinese electric vehicles or batteries. That’s why I lean toward “be nice to allies” bumper stickers and NSF funding advocacy rather than tightly nuanced “we need to better understand China” approaches when considering ChinaTalk’s decadal competition mission.

Kumar Garg: One thing I’ve been considering, though I don’t have the answer yet, is what new institutions we need. Much of what I care about regarding how science operates in this country has been overturned. The idea that we’ll navigate this period with identical institutions seems unlikely — whether it’s who makes the case for science, who serves as science messengers, or how we conduct science itself.

This raises questions not just about policies, but about institutions themselves. Obviously Renaissance is part of that response, but I have broader concerns. We’ll probably need new institutions because the players on the field will have to change. Systemic change of this magnitude requires that everything else engage in significant adaptive change for us to succeed. That seems unlikely with current structures.

That’s a major meta-question I’ve been asking the team, “We won’t be able to handle everything directly, but what institutions would restore us to better footing? Do we have them? Do we need to create them?”

Jordan Schneider: I’m starting to focus my energy differently because I’m uncertain whether additional ChinaTalk podcasts about the importance of allies will accomplish much. However, one constant you can expect over the next four years is AI development and rapid technological change.

Regardless of controversial Trump policies, the Defense Department will persist, and America will still need to protect itself. America engages in conflicts approximately every three years, so that pattern will likely continue. Perhaps this is just me entering a kind of intellectual monk mode after dynastic change, but I’ve been reading extensive military history and examining periods of rapid technological change — specifically, what it means to deploy these tools more effectively than adversaries.

This doesn’t directly answer your question, but I’m pursuing intellectual journeys rather than policy ones.

Kumar Garg: Here’s what I’d say, which connects to the role you and ChinaTalk are playing. One thing I mentioned to Jordan before we started is that frequently when people reach out to me, they reference hearing a great ChinaTalk episode. You may not have set out to do this, but you’re playing a valuable role in shaping how other people — especially technical professionals — think about problems worth solving and their mental frameworks for our current age.

People are seeking understanding and meaning in this moment. The question isn’t about marginal additional podcasts, but whether you’re providing people with new vectors for their lives and careers. When we first met, you described yourself as a nerd passionate about these topics but uncertain where to channel that energy. You’ve created something quite distinctive.

We might be living in an era of unusually shaped careers, and we need to give people more space for that kind of professional evolution.

Subscribe now

Jordan Schneider: That’s fair. Some of this material feels obvious to me but may not be obvious to people who don’t live and breathe these topics daily. It’s strange because I don’t feel like I’m part of some resistance movement — I’m just a guy with opinions on various issues. Some days I wake up feeling helpless, others I feel genuinely empowered. This isn’t a direct response to your question, but that’s my reality.

Kumar Garg: You’re reasoning in public. You’re expanding people’s understanding of how to think through these complex issues and broadening their sense of who engages in this work. What I consistently observe is that people operate in highly siloed environments. They’ll mention their sources, and I immediately know exactly who they’re reading or listening to.

If you can expand their perspective while providing actionable next steps, that’s valuable. Part of my goal is always emphasizing that there are numerous hard, interesting problems to solve across every arena. No arena is the wrong one. People who dismiss politics are missing a crucial point — we’re all living within political systems whether we acknowledge it or not. Don’t dismiss any arena; simply understand that there are multiple dimensions to engage with. There are challenging problems to solve, and nobody benefits if you just remain a passive observer in the cheap seats.

Jordan Schneider: Perhaps the way I justify all my World War I reading is that no one else is doing this specific work. I’m bringing historical insights to current issues as someone who also reads contemporary news and has the freedom to spend ten hours weekly on intellectual journeys exploring topics I consider relevant to today. That seems like a natural conclusion.

Kumar Garg: We’ll do an update in a year. Hopefully, the republic endures.

Jordan Schneider: Kumar, you need to set ambitious goals for yourself — accomplish so much that I have to bring you back in six months.

Kumar Garg: Absolutely. One of our primary goals is becoming more international. We have a partnership with the British government to develop their R&D ecosystem, and we want to expand that model to additional countries. We’re building an organization that continues to internationalize because science and technology are inherently global.

I’m hoping our fund model will attract new donors who have never engaged in philanthropy before. Beyond discussing our approach, I want the actual work to manifest tangible results in the world.

Jordan Schneider: Contact info@renphil.org if you’re wealthy, have innovative ideas, or simply have a technology-related challenge you need help addressing.

Kumar Garg: I’m also available on LinkedIn — please reach out. We consider ourselves fundamentally a talent network, so I’m always eager to connect with people who have compelling ideas.

Jordan Schneider: We should do a little parent corner. We’ll keep this as part of the annual check-in. We talked about slime last year.

Kumar Garg: We did. Here’s something we discussed before we started recording — I asked you about sleep training, and you mentioned being hesitant to push sleep training advice on others. I’m strongly convinced that sleep training is a gift you give your children. We had twins and committed to sleep training. They’re eleven years old now and remain excellent sleepers today. We attribute this directly to that early sleep training.

For any parent listening who’s on the fence and wants random advice from someone they’re hearing — I can’t offer this to everyone, but I usually tell people I know that I’m always happy to be anyone’s texting buddy, providing extra support to get through those terrible first few days when it feels like you’ve made a horrible mistake. On the other end, you have children who can sleep well, which benefits everyone.

Jordan Schneider: I’m with you. I outsourced this decision to my mother — maybe one of the best decisions of my life.

What’s a cute development we’ve observed recently? I bought a ukulele two years ago, thinking it would be nice to play with my kid. What’s been charming is that my daughter is nine months old now, and there was a period where her manual dexterity only allowed her to grab the strings and pull them. But one day she figured out plucking, and now she’s actually plucking the strings. It’s such a cool activation moment for her — realizing “I make the sound now” instead of just dragging this thing around the room.

Kumar Garg: The period from nine months to eighteen months is truly remarkable. You’re approaching walking and first words, then vocabulary takes off exponentially. It’s incredible developmental progress, so I’m very excited for you.

Jordan Schneider: Perfect. Let’s wrap it up there. Kumar, thank you so much for being part of ChinaTalk.

Kumar Garg: Thank you for everything you’re doing. I’m excited to have participated and look forward to future conversations.

Read more

Today we’re running an interview of Huawei founder Ren Zhengfei that made it to the front page of yesterday’s People’s Daily, the premier Party outlet.

What struck out to me in the Q&A was just how hard he hit on the importance of basic research to long term competitiveness. “Without basic research,” Ren says, “there are no breakthroughs, and we’ll never catch up with the U.S.” The People’s Daily giving Ren Zhengfei the airtime to promote this view underscores just how central China’s leadership today sees this work.

We’ve covered the US side of how the Trump administration may be killing our golden goose in this regard with its treatment of the NSF, NIH, university ecosystem and immigration system in recent episodes on America’s R&D Reckoning, our interview with Josh Wolfe of Lux, and a recent op-ed on the war on S&T talent.

This translation also serves as a reminder to subscribe to ’s excellent and meditative daily People’s Daily recaps in his substack Tracking People’s Daily.

"The more open a country is, the more it drives us to progress" — A Dialogue with Ren Zhengfei

By reporters Hu Jian and Chen Jiaxing People's Daily (June 10, 2025, Page 1)

Recently, at Huawei headquarters in Shenzhen, focusing on some hot topics of public concern, a group of People's Daily reporters had an in-person interview with Huawei CEO Ren Zhengfei.

Q: Facing external blockades and pressure, encountering many difficulties, what are your thoughts?

A: I haven't thought about it—thinking about it is useless. Don't think about difficulties—just get on and do it [干就完了] and move forward step by step.

Q: Ascend chips have been "warned on" about usage risks [BIS . What impact does this have on Huawei?

[BIS in mid-May warned firms anywhere in the world that using Huawei Ascends, which were almost certainly fabbed by TSMC under false pretenses, would be considered a violation of export controls. The Information recently reported that “The fallout [from this BIS action] was swift: one Chinese data center firm promptly halted even the small order of Huawei chips it had planned to make, while its legal team scrambled to understand how it could comply with the regulations, according to one person with direct knowledge of the change. Meanwhile, executives at ByteDance and Tencent, both of which have significant business footprints outside China, worry adopting Huawei chips would invite scrutiny from the U.S. government, according to people familiar with their thinking.”]

A: There are many chip companies in China, and many of them are doing quite well—Huawei is just one of them. The US has exaggerated Huawei's achievements; Huawei isn't that formidable yet. We need to work hard to live up to their assessment. Our single chips are still one generation behind the US.

[The incentives are interesting here. On the one hand you have Jensen Huang going around saying Huawei is an enormously powerful competitor who dropped Nvidia’s China market share from 95% to 50%—even though, as the Information very generously put it, “Market share data to confirm Huang’s statement is hard to find.” Jensen, of course, wants the US government to lift export controls not on semiconductor manufacturing equipment but his ability to sell AI chips and racks into China.

The Chinese government and Huawei have every incentive to sandbag Huawei and SMIC’s capacity to compete. They’ve deeply internalized the backlash from Made in China 2025 and want the Trump administration to think that easing export controls on inputs to semiconductors and not investing in the state capacity to enforce the current ones is less of a give than it appears.]

We use mathematics to compensate for physics, non-Moore's to compensate for Moore's [用数学补物理、非摩尔补摩尔], and cluster computing to compensate for single chips. In terms of results, we can still achieve practical applications.

[For more color here, see SemiAnalysis’ coverage of the CloudMatrix 384, Huawei’s answer to Nvidia’s rack solutions]

Q: If there are difficulties, what are the main ones?

A: Difficulties are just difficulties—when hasn’t humanity faced them? Wasn't slash-and-burn agriculture difficult? Wasn't the Stone Age difficult? When humans used stone tools, how could they have imagined high-speed rail? China has opportunities in mid-to-low-end chips, with dozens or hundreds of chip companies all working very hard. Especially in compound semiconductors, the opportunities are even greater. For silicon-based chips, we use mathematics to compensate for physics, non-Moore's to compensate for Moore's, utilizing cluster computing principles to meet our current needs. Software cannot be strangled [卡不住脖子]—it's mathematical graphical symbols, code, and advanced operators and algorithms stacked together, with no barriers blocking it. The real bottlenecks lie in education and talent pipelines. China will eventually have hundreds or thousands of operating systems supporting progress in Chinese industry, agriculture, healthcare, and more.

Q: Huawei is getting a lot of praise and public recognition these days.

A: Praise brings pressure. A bit of scolding keeps us clear‑headed. We make products; once people use them they will criticize—that’s normal. We allow criticism. As long as people tell the truth—even if it’s negative—we welcome it. Praise or blame shouldn’t matter; what matters is whether we can do our job well. If we do, there’s no problem.

Q: From your attitude toward difficulties and criticism, we sense you have a strong inner core—not caring whether it's praise or criticism, but firmly doing your own work well. This should be an important reason why Huawei has come this far.

A: There's still too much praise for us. People should pay more attention to understanding those engaged in basic research. They are highbrow and solitary [曲高和寡—literally ‘difficult songs find few singers’], ordinary people don't understand them, and it takes decades or centuries to see their contributions. Groundless accusations against them are detrimental to the country's long-term development. We need to understand and support those doing theoretical work. We need to understand their broad-mindedness—their great anonymity is our country's hope. Don't elevate one while suppressing another; those engaged in theoretical research are the hope of our nation's future.

"Basic research scientists are lonely—we must have strategic patience and understand them"

Q: How do you view basic research?

A: When our country possesses certain economic strength, we should emphasize theory, especially basic research. Basic research doesn't just take 5-10 years—it generally takes 10, 20 years or longer. Without basic research, you plant no roots. And without roots, even trees with lush leaves fall at the first wind. Buying foreign products is expensive because their prices include their investment in basic research. So whether China engages in basic research or not, we still have to pay—the question is whether we choose to pay our own people to do this basic research.

Q: Regarding basic research, people might not understand the point of it.

A: Scientific breakthroughs—there are inherently few people in the world who understand them, so those who don't understand shouldn't evaluate them. Einstein's discovery that light bends was only proven a hundred years later. In Guizhou, there's an agricultural scientist named Luo Dengyi. In the 1940s, when analyzing the nutritional components of fruits and vegetables, he discovered a wild fruit called cili [刺梨] with very high vitamin content. China was still in the Anti-Japanese War period, education levels were low, and few understood. Later he wrote a paper saying cili was the "King of Vitamin C." After nearly a century, Guizhou has made it into a natural vitamin-rich cili beverage, a luxury item among vitamin drinks at nearly 100 yuan per bottle, highly sought after. The cili industry has become a channel for farmers to escape poverty and become prosperous. Only then did people truly recognize Luo Dengyi, who worked at a broken table during the flames of the anti-Japanese war.

[A deep cut by Ren Zhengfei here but this does look incredibly tasty!]

Q: Many results that look irrelevant at first yet turn out to be hugely useful.

A: Theoretical scientists are lonely; we must have strategic patience and empathize with them. Tu Youyou’s discovery of artemisinin is one case; So was Huang Danian, who embodied "exploring innovation and serving the country with utmost sincerity." The symbols, formulas, and thinking in their minds can only be communicated with by a few people in the world. We must respect theoretical scientists because we don't understand their culture. Society must be tolerant, and the nation must be supportive.

Q: Basic research takes a long time, but companies must deliver returns.

A: We spend roughly 180RMB billion a year on R&D; about 60 billion goes to basic research with no KPIs, while around 120 billion is product‑oriented and is assessed. Without basic research, there are no breakthroughs, and we’ll never catch up with the U.S. [没有理论就没有突破，我们就赶不上美国].

Q: That’s true long‑termism. We heard Huawei even has a “Huang Danian Tea‑Thinking Room.”

A: Professor Huang was a great scientist. China noticed him during the Gulf War: the U.S. used a pod under helicopters to detect Saddam’s buried weapons and destroy them instantly—that pod had been developed by Huang at a UK university. He later resigned from his UK university position and taught at Jilin University. He spent his own money to get a 40 m² room, offering free coffee so students could “absorb cosmic energy over a cup of coffee.”

[Honestly this looks pretty lame, Huawei didn’t do the guy justice with this underwhelming study hall]

We have his family’s authorization to use his name for a non‑profit online platform—the Huang Danian Tea‑Thinking Room—where anyone can freely access global scientific information.

We also keep the “mouth of the funnel” wide open for basic research, partnering with universities. These are strategic investments with no performance metrics. Internally, we accept that we don’t know when theoretical work will bear fruit and impose no demands on the scientists.

“One purpose of socialism is to advance society”

Q: [A cringe leading question here…] Economist Richard Wolff and others argue that the U.S. lacks a high‑speed rail system because capitalism demands profit, while China’s socialist approach values social benefit: high‑speed rail, heavy‑haul lines, advanced power grids, expressways, rural cement roads, irrigation works, power plants… none make money directly, yet they underpin modernization. Competitive goods, meanwhile, are market‑regulated. What’s your view?

A: Why is it that only socialism tackles unprofitable tasks? One goal of socialism is precisely to develop society. China’s socialist market‑economy model is a magnificent undertaking. In infrastructure we had to follow this path—otherwise high‑speed rail, expressways, dams and so on would never have been built.

Q: What’s your outlook on AI?

A: AI might be humanity’s last technological revolution—barring perhaps fusion power. It will take decades or centuries to unfold. No need to worry; China has many advantages. [This line made leadlines in non-Party media]

Q: What advantages?

A: Hundreds of millions of young people—they’re the nation’s future. The General Secretary has said a nation’s strength rests on cultural flourishing. Technically, AI hinges on ample electricity and advanced networks. China’s power generation and grids are excellent; its communications network is the world’s most developed. The “Eastern Data, Western Computing” vision can become reality. [For more, see ChinaTalk’s past coverage of EDWC.]

Q: Anything else?

A: Chip worries are unnecessary: with stacking and clustering, our computing outcomes match the state of the art. As for software, open‑source ecosystems numbering in the thousands will meet society’s needs. [Wait, earlier in this interview I thought we were stressed about how behind Chinese chips were…?]

Q: How do you see China’s future?

A: After leaving our company, Tom Friedman bought a second‑class high‑speed‑rail ticket on his own to experience China. He later wrote, “I’ve seen the future, and it’s not in America.” [See ChinaTalk’s interview with Ezra Klein where I discuss this Tom Friedman piece with him. I, for one, would not take cues on analyzing China from Tom, who said in his Ezra interview that “Whether I’m writing about China from Washington, or whether I’m writing about China from China, I’m always just writing about America. My goal is to use China as my permanent Sputnik.”]

Q: Friedman said in that piece: “What makes China’s manufacturing juggernaut so powerful today is not that it just makes things cheaper; it makes them cheaper, faster, better, smarter and increasingly infused with A.I.”

A: Fundamentally, algorithms aren't in the hands of IT people, but in the hands of electrical experts, infrastructure experts, coal experts, pharmaceutical experts, and experts from various industries. At the practical level, Chinese manufacturing's application of artificial intelligence is very rapid, and many Chinese models will emerge.

Q: What support do private enterprises need from the state for their development?

A: Rule of law and marketization, with government administration according to law and regulations. Enterprises should focus mainly on value creation, technological breakthroughs, law compliance, and legal taxation. This harmonious development model will gradually release economic vitality.

Q: How do you view openness and development?

A: The country is becoming increasingly open, and openness will drive us to greater progress. Under Party leadership, the nation is administratively unified with smooth policy implementation. The gradual formation of a unified large market is possible and will certainly break through all blockades to achieve great rejuvenation.

[Ren Zhengfei of course got the seat of honor during the the Feb 17 Symposium featuring Xi which doubled as the Party’s olive branch to its leading S&T firms. We covered this meeting here on Xi’s Hard Tech Avengers.]

For more, see other core Ren Zhengfei speeches we’ve translated.

Read more

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【和放学以后永不失联】订阅放学以后Newsletter，每周三收到我们发出的信号：afterschool2021.substack.com 点击链接输入自己的邮箱即可（订阅后如果收不到注意查看垃圾邮箱）。如需查看往期内容，打开任一期你收到的邮件，选择右上角open online，就可以回溯放学以后之前发的所有邮件，或谷歌搜索afterschool2021substack查看。

截至目前，放学以后Newsletter专题系列如下：“在世界游荡的女性”系列、“女性解放指南”系列、“女性浪漫，往复信笺”系列、莫不谷游荡口袋书《做一个蓄意的游荡者》系列、“莫胡说”系列”《创作者手册：从播客开始说起》,播客系列和日常更新等。

本期放学以后信号塔由莫不谷轮值！

我之前在放学以后播客《我不要只做世界的承受者，我要对这个世界一顿发起！》设置了一场思想实验，其中就一个问题（是否支持世界其它生物也能和人类一样轮流主宰世界）我们battle了好几个小时。我们在这期播客合作的“不朽真龙”在它们的播客首页也有很多关于人类根本性问题的思想实验甚至实际实验的讨论，感兴趣的朋友们可以去不朽真龙的播客首页查看。比如这一期关于人可以活到1000岁的思想实验乃至“真实预言”，可能会对大家当下的焦虑和执念有很多启发。

为什么要对假设的问题，思想实验的问题这么投入呢？

因为恰恰是每一个假设的问题，最能引发人最真实的反应，继而这反应又会暴露人在这个社会生存所持有的立场，站位和根本性的观点。你问一个真实存在的问题，所有人因为利益相关和名誉脸面遮遮掩掩。你问一个假设的问题，大家会毫无意识地袒露无疑。

我非常喜欢此类的思想实验，甚至觉得它应该做成桌游用于一些新朋友或者老朋友的聚会，不然能火速破冰，从small talk进入deep talk，还有可能引发非常激烈的交锋和亲密，能把平淡的聚会和交谈搞得热火朝天，即使产生了争论也会尽兴而归。

最近我又看到了更惊人且更好的思想实验，我自己在家看的过程中激动不已，且已经在脑中想象了好几个朋友和我共处一室，我把这个思想实验分享给大家，正等待大家的反应，房间里落针可闻。

这个思想实验来自于格雷格·伊根的科幻小说集《祈祷之海》中《恰如其分的爱》这一篇，当然其它篇《学习成为我》《植入的公理》《水晶之夜》也都是精彩且伟大的思想实验，但是《恰如其分的爱》非常值得用来分享给所有的华人女性或者全体女性。因为它对在父权社会中长大的女性，用最精妙的实验设计，问出了那个最核心的问题：为了爱，你的自我究竟可以让渡多少？

(图源自Pinterest)

在介绍这个精妙至极的思想实验前，我想先热烈地赞美一下这本书，献上我的爱意和敬意。同时也向在游荡者平台向我推荐这本书的朋友致谢，当时这位朋友听完了我的单口播客《美妙人生的关键：让我们一起来扭一扭它》后发现我对数学感兴趣给我留言推荐了这本书，而里面关于数学的那一篇实话实说我没有全部看懂，但是其它篇却把我从地面抛上了云端。

现在这本书的作者格雷格·伊根，她已经（对的我认为作者是女性）是成为我心目中最好的科幻作家，没有之一！除了数学那一篇外，其余几乎每一个讨论和思想实验都写出了我的心声，倘若我能写出来这些我感觉自己可以明天死去也能含笑九泉的程度。这部作品是我理想中的我自己的顶格上限：就是我足够有激情，活力，努力，充分探索了我的才华，我最好最好能写出来的作品。

在阅读的过程中，我发现了她堪称伟大和神性的创作秘诀：她让我的情绪，身体反应，精神活动，处在不同的海拔。读者能在同一时间同一地点看到海平面和高山，思维的起伏落差带来精神蹦极的感受。热爱思维活动的人会体验一波又一波的精神高潮，不热爱思维活动的人会觉得身心疲倦。

庆幸的是，这里只有一个实验，因此无论你疲累还是振奋，它都足够能让你沉浸。

现在，实验来了：

你最挚爱的人A（书中是该女性的丈夫，你可以代入此刻你最挚爱的任何人，我们此处称它为A）出了车祸，身体被撞得一片稀烂，无法修复，大脑未死亡。A想要像车祸之前一样正常活着，就需要一副新的身体。

现在是一个科学更发达的时代，可以让其它女性用子宫孕育一个A的克隆身体（仅身体，无大脑），等这个无脑身体孕育并生产出来，会通过2年的时间加速生长A如今年岁的身体状态。这部分费用保险公司全部报销。

而在等待克隆人身体长大的时间，A原本的大脑需要被保存，而保存的方式有2种： 1.价格极其高昂的医学保存手段，这部分的高昂费用保险公司不报，你也完全支付不起。

2.伤者的女性家属可以用自己的子宫来怀着伤者的脑子，怀2年，等克隆身体在2年内发育好了，就可以从你的子宫取出A的大脑完成移植，A恢复成车祸之前的样子。对此保险公司不会支付你任何费用。

此时此刻，你肯定有诸多疑问：

1.为什么科学已经发达到可以克隆人类身体并2年内加速生产到成年人的地步，还需要女性来孕育克隆体和子宫保存大脑？

答：因为父权社会从来如此，无论技术多发达，能免费使用女性，那还费尽心思发明什么其它方法呢？

2.为什么保存大脑的费用保险公司不报呢？

答：因为保单条款里写着伤者的直系家属需要承担护理义务，保险公司不为家属的护理义务付费。倘若家属不愿意承担相应的护理义务，要把这部分义务外包，那保险公司不报销外包费用。用女性家属的子宫保存伤者的大脑，属于相应的护理义务。

3.你听到这里可能会更火大，其它女性用子宫花10个月的时间孕育克隆身体起码得到了保险公司支付的费用，你花两年的时间用自己的子宫保存伤者的大脑，什么也没得到？

答：因为你这本质属于家务劳动，女性的家务劳动，在父权社会从来都是无偿的，这一点保险公司很明白，它们不会给家务劳动支付报酬。此刻你肯定很恨保险公司。

4.还有其它选择吗？

答：当然你也可以放弃用自己的子宫来保存，A就此殒命，保险公司正好也不用支付高昂的克隆身体的费用，省了一大笔钱。你们买的保险最后什么也没用上，白白支付了保费，保险公司才是最赚的。你对保险公司的恨意更深。

基于以上所有实验条件和实验解释：你要做出何种选择呢？

“你问我爱你有多深？爱你有几分”

现在这个实验选择代表你的心。

你恰如其分的爱，究竟有几分？

为了爱，为了恨，你究竟能让渡自我几分？

这个实验精妙至极的点是它把一切条件都推到极致，却又设置地精准非凡。每一个都有现实对照，且意味深远。

其中有重大的道德考验：你要用其它女性的身体来代孕自己所爱之人的身体吗？这对其它女性不是一种剥削吗？现实社会正发生着无数这种剥削，而保险公司为这个剥削全部付费？你的道德压力会因此减轻吗？

有对父权社会最深刻的讽刺：女性的身体是最好且廉价的利用工具，科学进步用来解放人类，但不解放女性；女性承担的家务劳动是无偿的，这一点多进步的科学社会也依然要保留；女性是如何自我说服应该去爱去奉献去牺牲去献祭自我的。

还有最可怕的自我拷问：我要为所爱之人（尤其是丈夫，是男性）让渡出自己的身体吗？爱被渲染地如此伟大，我口口声声说爱它，倘若我却放弃用自己的子宫救它，我是真的爱吗？我会被指责自私无情吗？我真的了解一个东西放在我的子宫里2年，我所要经历孕期反应和全部后果吗？女人拥有子宫，究竟是福还是孽的？倘若今天受伤的是我，我期待我的女性家属和男性家属为我做出什么呢？

继而这些问题又会被提出：女性能接受自己被剥削吗？能接受其它女性被自己的丈夫，保险公司乃至隐身的自己剥削吗？只不接受前者，却接受了后者，这样的女性是“自我以上人人平等，自我以下阶级分明”吗？两个女性谁更惨呢？因为代孕科隆身体的女性起码还得到保险公司付的钱吧，代孕自己丈夫脑子的女性获得了什么呢？

书中的这位女性做出了决定：同意用自己的子宫来保存丈夫的脑子（这也意味着同意了保险公司找另一位女性孕育丈夫的克隆身体）。

她做出选择的原因是：“我突然后知后觉地意识到，保险公司那样粗鲁地挑衅我，其背后可能另有深意。毕竟，如果我任由克里斯死去，他们不仅可以省下生物生命维持那微薄的成本（要是我同意提供租费全免的子宫），还可以省下昂贵的替换身体的费用。这不就是蓄意为之的粗鲁行为，再加上一点儿对逆反心理的巧妙利用嘛……

要想保持理智，我唯一能做的就是打破这种狗屁伎俩；无论环球保险公司耍什么阴谋诡计，我都不放在心上；我身怀他的大脑，并不是因为受到胁迫，也不是出于内疚或义务，更不是为了证明自己不会被操纵，而仅仅是因为：我爱他爱得足够深，我想要拯救他的生命”。

对保险公司阴谋诡计的恨意，加上对丈夫“足够深”的爱意，她做出了如上的决定。

而在决定后她饱尝孕育的痛苦，同时在想另一位也在孕育克隆身体的女性：“我从未见过那副新身体的“母亲”，也从未见过那个尚未长大时的克隆人。不过，得知那个东西出生后，我确实在琢磨，她是否觉得自己的“正常怀孕”过程和我的非正常“怀孕”过程一样煎熬。我想知道，哪一个更好受些：是怀上一个由陌生人DNA培育，没有人类思想潜力、大脑受损的婴儿状物体？还是怀上自己爱人的休眠大脑？哪一种状态更易陷入一种出格的、忘乎所以的爱？”

韩炳哲在《爱欲之死》这本书（不推荐这本书）里有定义纯粹之爱的最基本条件：要求一个人有勇气消除自我。

实话实说，古往今来，古今中外，没看到多少男性为了爱消除自我。而女性则被整个社会哄骗着，鼓励着，教育者，激发着为爱消除自我，无论是出借自己的子宫为男性传宗接代，还是奉献金钱，无偿的家务劳动，真挚真诚能滋养男性的绵延爱意，历程堪称“前仆后继，继往开来”，甚至社会主流一度还把这些举动命名为“勇敢”。“勇敢去爱”的口号激励或者说诱惑了起码好几代女性。

而如今庆幸的是：爱欲在全世界范围内呈现出了消退乃至消亡之态。韩炳哲说这是爱死了。我觉得不是。

Nee(荷兰语的no)，不是爱死了，而是：

（接下来是我的暴言，我说是暴言，是为了不冒犯到一些并不想真正讨论这个问题，只想合理化自己选择的人。并不意味着我真心以为自己说的是“暴言”。有时候在中文世界，人必须污名自己，来减少那些主动来看别人的观点，继而宣称自己被观点伤害，四处出动当“受害者”，却对真正的加害者充耳不闻的人出现在这里。-我在这里用了长难句的定语也是为了增加阅读难度的同时劝退那些人出现在这里。

以及对这样的人有最后一句建议：倘若你不愿意找到你人生真正的加害者，不愿意去看自己受害和痛苦的原因，而是去攻击和想要消除指出这些伤害和加害的人。那你所收到的伤害和痛苦注定会长久延续和指数升级。You are doomed.）

因此这是暴言也是更直接的实话分割线,也可以在游荡者平台，放学以后的公众号和爱发电解锁）。除了我的暴言，我提出来的一个直接的问题和我的答案，霸王花在提前看完文章后也给我发来了她的答案，可以说相当荒谬，在她同意放出来警示其它有同样想法的朋友后，我也贴在了下面。

Read more

Has America already lost its dynamism in basic research? Josh Wolfe, co-founder of Lux Capital, joins the podcast today to discuss.

We get into…

• Why the Trump-Elon fallout matters less than you probably think,

• How much payoff corporate leaders are reaping from their campaign to appease Trump,

• The erosion of the U.S. research ecosystem, and how we should think about philanthropic giving amid that chaos,

• Parenting, strategies for emotional resilience, why short videos aren’t terrible, and the history of the machine gun

It’s a good interview. Listen on Spotify, iTunes, or your favorite podcast app.

Dreams of the Tech Right Meet Reality

Jordan Schneider: We have to start with Trump and Elon. I’ll let you pick the frame.

Josh Wolfe: I’m going to take a contrarian position — it’s irrelevant. It’s irrelevant because it dominates front-page news and captures everyone’s attention. I joked yesterday that if you were planning fraud or were a terrible company with bad news, yesterday was the perfect day to release it while everyone focused on this inevitable outcome.

People were making predictions six months ago — over/under bets on whether it would last three months or maybe a year.

Jordan Schneider: There were actual betting markets created for this.

Josh Wolfe: Exactly. You were betting under a year, maybe over three months, but none of this was surprising. This was an unstoppable force meeting an immovable object — kinetic chaos was inevitable. Elon served as a useful foil for Trump and did an excellent job helping him win the election. They were strange and unusual bedfellows.

Trump has historically been a strong China hawk, albeit with nuanced moments regarding Xi Jinping. Meanwhile, Elon depends heavily on China in ways people don’t fully appreciate — particularly Tesla’s production levels and the profit margins China contributes.

Jordan Schneider: Well, the China Talks audience understands this, but perhaps not everyone does.

Josh Wolfe: Right, not the broader ecosystem. This creates a significant vulnerability. You’ll see Elon Musk historically position himself as an outspoken advocate of free speech — though whether Twitter was genuinely about free speech is debatable. I’m more cynical about his true aims.

What you’ll never see him discuss publicly is anything about China. He’ll never criticize Xi for human rights abuses in Xinjiang province or regarding the Uyghurs, and he’ll never mention Tiananmen Square. Nothing. It’s “Free speech for thee, but not for me” when it comes to China.

Inevitable clashes were bound to occur. Was this really about the spending bill, or something deeper? When the leaks started — clearly planted by the White House over the past five days about drug use, and then the Epstein connection — it became salacious and interesting for everyone, but none of it was surprising.

My position is that it doesn’t matter. It’s similar to tariffs — Trump campaigned on tariffs, Democrats attacked him, saying he’d implement tariffs that would be economically troublesome. I don’t understand why these developments are surprising when they shouldn’t be.

Jordan Schneider: Speaking about your profession, many people in the tech ecosystem — particularly with the rise of the tech right throughout 2023 and 2024 — pinned their hopes on this man, similar to how Elon did. They hoped that by being inside the tent, they could influence policy trajectory. From my perspective, benchmarking to June 6, 2025, this has been pretty dramatically disappointing.

What are your reflections on that psychological arc? Are these guys all just patsies?

Josh Wolfe: During November and December of 2024, I remember peak rhetoric around the zeitgeist of America’s golden age. All you could hear was “we’re back, baby” and these ridiculous 1980s Top Gun-infused, Hulk Hogan maverick-style videos celebrating America.

Jordan Schneider: Sure.

Josh Wolfe: I told my wife, who runs an activist public hedge fund and manages our personal money, “We need to buy long, deep out-of-the-money puts” because all I heard was “to the moon” and “American greatness.” It felt like an echo chamber of optimism.

Two to three months ago when the tariffs hit, those positions performed very well in our portfolio. We’ve been relatively unaffected on the venture side.

One area where real change is happening — and it’s always dangerous to say “this time is different” — but it feels very different: defense. We’re seeing a combination of a nearly trillion-dollar budget and a huge shift toward autonomous systems, AI-driven software systems, and space satellites. Remember, Space Force was an absolute joke eight years ago.

Jordan Schneider: Okay, we’re pivoting a little…let’s stay on track.

Josh Wolfe: But this is where the venture world has made noise and impact. There are sympathies, appreciations, and influence, particularly in tech and defense.

Jordan Schneider: Sure.

Josh Wolfe: Regarding everything else: you had everyone pledging fealty, taking a knee, and donating a million dollars to the inauguration campaign. Zuckerberg, Bezos, Satya, Sundar — they were all paying homage to Trump, hoping to escape DOJ, FTC, or antitrust scrutiny. How that plays out remains to be determined.

The Trump-Elon situation was both weird and destined to end inevitably. There will be some strange reconciliation, and they’ll be bros again. But I don’t think this is just about the spending bill — it’s much deeper.

Jordan Schneider: One of the things the tech right didn’t price in is what’s happening around immigration and the basic research ecosystem. Let’s start with basic research. There’s this argument that we don’t need the NSF or NIH because corporate R&D has increased significantly over the past 30 years and can drive innovation forward.

As an investor who focuses on scientifically ambitious investing, what do you think the economy gets from basic fundamental research that happens in labs and universities versus corporate R&D?

Josh Wolfe: Long-term: everything. Short-term: it’s hard to see the value. For policymakers looking at budget cuts — particularly those focused on short-term gains — these programs seem like easy targets. This isn’t just a Trump phenomenon. In 20 of the last 22 years, we’ve had federal cuts to federal science funding.

Jordan Schneider: We had a brief moment with the Endless Frontier Act.

Josh Wolfe: Right, it started with Endless Frontier, which referenced Vannevar Bush’s work from around 1945. Then you had “The Gathering Storm” by Norm Augustine, the former Lockheed executive, warning about America’s talent base and cultural shifts in what people were drawn to and celebrating.

What they didn’t mention in that report — now over 20 years old — was this anathema, this zeitgeist against the military-industrial complex, precisely when China was not only embracing but mandating military-civil fusion.

Over the past two to three months, we’ve witnessed a perfect storm. Now you have the politicization of academia. Harvard sits number one in Trump’s crosshairs — some speculate because Barron applied and didn’t get in, creating some Shakespearean vendetta. Whatever the reason, while anti-Semitism was the stated concern, it was far worse at Columbia, right here in New York City, than at Harvard.

Subscribe now

According to Nature magazine’s rankings — which publish various metrics including H-index measurements — when you look at the most important scientific publications by institution, the top 10 includes Harvard at number one, but positions two through ten are all Chinese universities.

Fifty percent of AI graduates worldwide come from China. Thirty-eight percent of the AI and computer science workforce here comes from China — not native-born Americans. They’re outpacing native American citizens by 35-37%.

We’re losing the talent game. What should we do to win? What we did in World War II and beyond: attract the best and brightest. Create brain drain from other countries because people are repelled from their home nations.

Eastern European Jews in the 1940s established the Institute for Advanced Study and brought Einstein here. Having the atomic bomb developed here rather than Germany was a net positive for the world. Soviet émigrés in the 1980s escaped communism for capitalism during the Cold War.

We should be stapling visas to every Chinese, Indian, Pakistani, Israeli — anyone who wants to come here. We should help their parents immigrate too, because often family remaining in authoritarian home countries becomes leverage that governing regimes use to control whether these individuals can leave or return.

The last Republican I remember discussing this was nearly 20 years ago: Newt Gingrich, who called for tripling the NSF budget. He understood that everything from Genentech (which emerged from the UC system) to Google (from Stanford) was premised on long-horizon scientific research.

This isn’t just computer science — it’s chemistry, physics, materials science, and all the breakthroughs that emerge from these fields. If we knew what those breakthroughs would be, we’d fund them today. But we don’t. We rely on this rich ecosystem in our own self-interest to support brilliant people who generate breakthroughs.

These breakthroughs often result from combining insights from one department with research from a different university. More than ever, this cross-pollination is critically important. We’re ceding academic and intellectual leadership to other countries. China will likely be the greatest beneficiary.

International students used to comprise roughly 25% of our student body. Now it’s down to 12-15% and falling. China is actively recruiting with programs like their Thousand Talents initiative.

This represents the greatest self-inflicted wound Americans have created in several generations.

Jordan Schneider: You had this line: “If the Politburo were drafting America’s self-sabotage plan, they would politicize science and freedom of inquiry, starve young and ambitious investigators of funding, and discourage the best immigrant minds."

To recap for everyone: the NSF is spending at half the rate it was in 2024, despite authorized funding. The new budget dramatically cuts science and research spending. Universities have been cut off wholesale from billions in research dollars. Johns Hopkins is laying off 1,000 people because they’re worried grants won’t materialize.

Trump said yesterday that we actually want Chinese students, which creates bizarre mixed messaging when his administration simultaneously tells Harvard they can’t admit Chinese nationals.

Josh Wolfe: Some of these grant cuts are both tragic and absurd. There was a researcher studying neurotransmitters for neuroscience whose grants were cut. They couldn’t understand why until they realized the word “trans” triggered a five-letter string search that flagged anything with “trans,” “DEI,” or similar terms.

Now researchers have to change grant language — writing “neurons firing intracranially” instead of “neurotransmitters."

Jordan Schneider: That’s not funny, though.

Josh Wolfe: You’re right. It’s tragic.

Jordan Schneider: Why don’t you talk about the $100 million pool you launched?

Josh Wolfe: It’s specifically not grants because we’re not trying to substitute the charitable giving of our government — funded by taxpayers — which actually makes America great when we have scientists working on breakthrough research.

Historically, about 10% of our investments are de novo new companies, and most emerge from academic labs. We find a principal investigator — a fancy term for a scientist at an academic institution — whether at Harvard, Princeton, Yale, Georgia Tech, Cornell, or elsewhere.

The early 1980s Bayh-Dole Act, sponsored by Senator Evan Bayh and Senator Bob Dole, allowed universities to own intellectual property created with federal taxpayer research funding. When a scientist at these institutions files a patent, the university becomes the assignee while the scientist remains the inventor.

We, as a venture capital firm, can license that technology. There’s a well-established deal structure involving licensing royalties and equity — the scientist typically gets about 25% ownership. This mirrors Google’s “one day a week” policy, allowing employees to spend time on personal projects, but applied to exclusive company work.

This approach for launching companies from academic labs is well-established. Professors often stay at their institutions while postdocs handle translational research and join the company. We’ve created about 25 de novo companies this way, spanning everything from 4D lidar for autonomous vehicles to digital olfaction — essentially “Shazam for smell” — to cancer therapeutics and Nobel Prize-winning work from the Howard Hughes Medical Institute that enables real-time cellular imaging.

Historically, this represents about 90% de-risked investment, meaning we take roughly 10% scientific risk. Now, with grant cuts and layoffs at institutions like Johns Hopkins, exceptional research is being abandoned — like proverbial Rembrandts in the attic. Scientists are wondering what to do next. If they wait six months for resolution, they need to find work elsewhere, whether in academia, nonprofits, or the private sector.

Our response is the Lux Science helpline — a bat signal for struggling researchers.

We’re dedicating $100 million from our latest $1.2 billion fund to double down on early-stage science risk. Instead of taking 10% science risk on 90% proven concepts, we’re willing to take 50% science risk on half-baked ideas.

We’ll help license patentable work into our existing companies or arrange sponsored research to continue funding relevant projects. For example, someone contacted us about novel materials for radiation hardening in space. We have 17 companies across the space ecosystem that could sponsor continued research.

Other scientists ready to enter the private sector can work with us to start new companies around their expertise. We’ll license their work, assemble their team, build laboratories, and launch them.

This is just a drop in the bucket — we can’t do this alone. We need dozens, if not hundreds, of other VCs to recognize the value of funding early-stage scientific work. It’s profitable, beneficial for national security, and good for business. Our $100 million allocation, while significant, isn’t sufficient by itself.

Jordan Schneider: I’ve received numerous emails from intelligent people essentially doing backfill work for USAID funding cuts. While it’s admirable that people are donating $10 million or $50 million, tens of millions of people will suffer, and we’re delaying future breakthroughs because we’re not conducting this research. The laboratory animals aren’t being fed, and you need the time, energy, and tacit knowledge built over years of lab work to accomplish meaningful research.

Josh Wolfe: Absolutely. This isn’t purely black and white — there are legitimate concerns about USAID politicization and questionable funding destinations. There are appropriate and inappropriate uses for that money.

My main criticism of the NIH isn’t to strip their funding, but to bias toward young investigators. Too much grant money goes to, frankly, older researchers. Older scientists can be set in their ways and resistant to new approaches.

You need that potent combination of naivety and ambition that drives both entrepreneurial and scientific discovery. Young researchers have the arrogance to say, “I know better than you — why wouldn’t we try this?” while older researchers respond, “Why would you do that? I’ve tried it ten times and it never works."

I would restructure the NIH budget and federal funding to favor younger investigators.

Jordan Schneider: I wonder to what extent the Silicon Valley tech ecosystem that embraced Trump and the DOGE energy — which led to all these cuts — stems from the fact that for the past 20 years, you could make the most money by being 19 and writing software. That’s not something you necessarily need a PhD for. The social returns to that work are very different from what you see when making novel materials and drugs.

Josh Wolfe: Absolutely. Even the term “engineer” shifted from physical engineer to software engineer. For at least 20 years post-internet boom — through the SaaS enterprise boom and cloud boom — that was the dominant paradigm.

Lux was on the periphery of that trend because we don’t really fund software. About a third of our investments focus on healthcare, biotech, robotic surgery, and medical devices. Another third covers aerospace, defense, manufacturing, and industrial applications. The final third is what we call “core tech,” defined more by what we don’t do — very little internet, social media, mobile, or ad tech.

Great fortunes were made and great companies built during the software boom. Interestingly, there’s a China parallel here. Fifteen years ago, Xi Jinping designated software as a domain China would fund heavily. This created incredible companies and ecosystems, and many US investors funded them, believing China was a democratizing, growing market. Knowing what they know now, it would be difficult to justify investing in companies like ByteDance.

Jordan Schneider: Unless you’re Bill Gurley [who recently funded Manus AI].

Josh Wolfe: I love Bill, and he’s not necessarily part of the partnership decision. We’ve discussed this, and he would argue there might be scenarios where teams like Manta are trying to exit China. I think this deserves more study rather than just criticism. However, I personally wouldn’t invest in Chinese companies that are part of CCP military-civil fusion.

The shift also happened within China. At the last Politburo gathering about a year and a half ago, the people close to Xi weren’t software engineers or computer scientists — they were from space, biotech, rockets, and defense. All hard sciences.

This coincided with when Jack Ma went to “spend time with his family” and when entrepreneurs were capped at $9.9 billion. Anyone exceeding that threshold was essentially decapitated — they could start schools or enter education, and either they or their family could leave the country, but not both simultaneously.

The new directive became hard sciences. I used to joke that if you wanted to make money, the greatest capital allocator wasn’t following Warren Buffett or Seth Klarman or other great value investors — it was listening to what Xi Jinping was funding, because that’s where the world was heading.

I agree that for the past 10-15 years, the focus has been software. Many VCs, including Marc Andreessen who said “software is eating the world,” were drawn to Trump partly because they felt rejected by Biden. They were spiting Biden and possibly spiting themselves long-term.

I still believe the greatest entrepreneurs don’t really care about political developments or the 10-year interest rate. They’re building something because they have a chip on their shoulder — and as I always say, chips on shoulders put chips in pockets. They’re driven by private ambitions with timeframes that supersede one, two, or even three presidential cycles.

I remain bullish on great entrepreneurs. There’s simply a shift from software to hardware.

Jordan Schneider: Something worth exploring is the importance of tacit knowledge in hardware versus software and the learning that happens almost entirely in universities during master’s and PhD programs. When you’re doing actual scientific work, you need some sense of navigating the dark forest.

Josh Wolfe: That’s absolutely true. This applies even to semiconductors and manufacturing. TSMC’s Arizona facility is starved of talent — not just union workers doing physical assembly, but specialized expertise. There’s a scarcity of talent because much of that tacit knowledge remains in Taiwan and is difficult to transfer here.

You see this dynamic in companies working on the most sophisticated technologies. My partner Sam Arbesman, who you know or have spent time with, is a brilliant scientist-in-residence here. He has a new book coming out called “The Magic of Code,” and his previous book “Overcomplicated” offered a modern version of “I, Pencil” — the thought experiment about complexity.

If you were to make a semiconductor today, or an Apple iPhone, consider the number of components, the tacit knowledge required, the number of countries and companies involved — it’s extraordinarily complicated. No single person can make a pencil, let alone a semiconductor, chip, GPU, or field-programmable gate array.

Culturally, we get what we celebrate. For 25 years, we at Lux have complained that American culture celebrates celebrities — the Kardashians, the Hiltons, and similar figures. You see this manifested in TikTok and what gets fed to American users versus what’s not even allowed in China and what China celebrates in terms of STEM education.

We’re losing this battle terribly. Looking at the labor pool: we have 300,000 undergraduates in science here. China graduates approximately 1.2 million. Half of our 300,000 are foreign students.

This represents a cultural crisis regarding what we want our children doing and celebrating. We don’t need more people in marketing, advertising, or selling products. We need people inventing things that everybody else in the world wants to buy.

Jordan Schneider: And podcasting.

Josh Wolfe: But you produce intelligence and insight, which is valuable. It’s unique insight because, as you noted, people who follow China Talk understand things that others don’t. That’s an advantage — you produce something intellectually valuable.

Jordan Schneider: Back to the narratives. There was this fascinating Twitter exchange a few days ago. You mentioned the atomic bomb and the Apollo program, and J.D. Vance framed it as “we didn’t need foreigners to do this — we had 600,000 Americans weaving the wires to connect everything together."

Both perspectives are true: his narrative about American workers is accurate, and your narrative about leading lights of both the rocket and nuclear programs coming from Europe is also true. But this narrow-minded “you have to be born here to be part of the circle” mentality — let’s discuss what that means.

Josh Wolfe: This created a significant fissure. People entering the administration who were immigrants from India were being lambasted by these self-described “heritage Americans.” Elon defended them because he’s an immigrant himself — not the classic immigrant from poverty in India or Russia, but an immigrant nonetheless.

He argued there’s a distinction between immigrants crossing the border and taking blue-collar jobs — who may not conform to American society and could include criminals or drug dealers — and brilliant people making extraordinary contributions. For example, the recent genetic medicine breakthrough that cured a baby using one of the first in vivo genetic editing approaches involved two scientists, one whose parents arrived from India 40 years ago. Thank God those individuals came here and had children.

We want this brain drain — from World War II through the Cold War to our current version of Cold War competition. I don’t understand why people consider “immigrant” a dirty word. Immigrants literally are the fabric of this country.

One of my mentors who put us in business, Bill Conway, who founded the Carlyle Group, focuses his philanthropy on addressing one of America’s biggest problems: nursing shortages. He says if you’re anti-immigrant, good luck getting sick, because our entire healthcare system depends on immigrants playing critical roles — from hospital orderlies to doctors to neurosurgeons.

We want the best and brightest coming here. It’s a complete self-inflicted wound to say “America First” means shutting out talent. America First doesn’t mean America Alone, as Scott Bessent has said. Scott is one of the great adults in the room and a friend.

People ask why Scott doesn’t speak out more publicly on these issues. Scott is a student of markets first — he understands currencies and countries — but he’s also a student of human nature. He understands in a Shakespearean way who his boss is, who he works for, and how to manage that relationship. We want him doing exactly that because we don’t want him out of the job.

Jordan Schneider: We discuss immigrants and science frequently on China Talk from a US-China competition perspective, and that’s all valid. But the fact that people want to come here is remarkable. Look, at the margins there are legitimate concerns about completely open borders, but limiting human flourishing by building walls and splitting families is just —

Josh Wolfe: I’m going to be blunt because I come from Brooklyn: this approach is counterproductive and self-defeating. The idea that we’re limiting the immigration of brilliant people who could make this country better is fundamentally wrong.

Our greatest export isn’t our music, Hollywood, fashion, high-tech companies, banking system, or rule of law — it’s the American Dream. The American Dream is arguably our greatest export; it’s what everybody wants. The measure of a great country is whether more people want to enter than want to leave.

Walter Wriston said, “Capital goes where it’s welcome and stays where it’s well-treated.” This applies to both money flows and people flows. I hope the rhetoric will change and that there will be enough great American leaders to counter jingoistic “America First” voices.

There’s a delicate balance: we must remain that shining beacon on the hill, attracting people while distinguishing between good people who genuinely want to build better lives here — who want to embrace the American Dream and become proud Americans — and bad actors who want to undermine the country. Many of the latter are coming through our university system and may be funded by foreign actors, but that’s a separate issue.

There’s virtue in virtuous people coming here and the American Dream being upheld and celebrated. Right now, we’re casting shadows on that dream.

Jordan Schneider: You opened by saying the Elon-Trump fight is noise. But the signal is that people around the world would rather live here than in China.

Josh Wolfe: They still do. My wife mentioned recently that many countries hate us right now — Canada, tourism is down — and there’s probably some schoolyard “arms crossed, I’m not playing with you” attitude.

Jordan Schneider: Sure.

Josh Wolfe: The reality is, I read about how Canada supposedly hates us, then the next headline I see — because I read many papers each morning, including the Globe and Mail — is “Canada wants in on Iron Dome."

The rest of the world, even if they hate us, still wants our superior military technology. The zeitgeist of popular antipathy will fade, but people will still want our materiel — our military materiel.

Jordan Schneider: We’re recording this a few days after the Tiananmen anniversary. The fact that hasn’t happened here — the fact that there is free speech, elections, due process, and habeas corpus — represents incredibly powerful long-term advantages. These aren’t just abstract principles — they’re things that make life worth living and make you excited to get out of bed in the morning.

Josh Wolfe: I remain relatively optimistic personally. I’m always optimistic about technology, science, and ultimately the human condition. I’m always skeptical about people because I’ve read a lot of Shakespeare. Technologies are amazing, science is amazing, and both will continue to progress. People generally disappoint because they’re vainglorious, full of ego, petty, jealous — they connive and deceive.

People ask how I can be optimistic yet cynical. That’s the pairing — I’m optimistic about science and technology, skeptical about people. I love Tom Wolfe’s answer when asked why he writes about space, astronauts, and moneyed Upper East Siders but not politicians. He said politicians are like passengers on a train, and the country is the train. The tracks might go up and down, but it’s generally heading in the right direction over time. Inside the train, there are clowns on one side in red and clowns on the other side in blue, throwing pies at each other nonstop. Every four years, the engineer changes, but the train stays generally on track.

I take that view: the real thrust comes from people, values, that great beacon on the hill, the American Dream that attracts people here. We shouldn’t deter them from building here on that track. Most entrepreneurs couldn’t tell you much about the political system or what’s happening at the Fed or Treasury — they’re focused on building.

Jordan Schneider: But the fact that they have to worry now, that they have to fear their visas getting revoked — look, in other moments that might be manageable. But within our range of expectations, there are definitely scenarios where an executive acting alone can really bend those train tracks.

Choosing Optimism + Parenting

Jordan Schneider: Speaking of Shakespeare, we did a show with Eliot Cohen about his wonderful book The Hollow Crown and another episode as a Biden emergency discussion. The Trump-Elon situation probably deserves its own Shakespeare emergency episode.

told me to ask you this question — does America need a Gen Z Marsha Linehan?

Josh Wolfe: Marsha Linehan is one of the founders of CBT — Cognitive Behavioral Therapy.

I love Derek. I got into psychology maybe five or six years ago. My kids are 15, 12, and 9. My oldest daughter is much more like me — emotionally volatile. I’ll have high highs (not bipolar…emotive!) then calm down. My wife and middle daughter have slow burns with grudges lasting three or four days.

We learned CBT and DBT techniques as a family. Derek’s question is whether Gen Z needs this — I think they do. CBT is essentially Stoic philosophy in a clinical psychology setting. I wish I’d learned this 20 years ago as a kid or teenager — I would have had healthier relationships. I wish I’d learned it 15 years ago before my first child was born — I would have been a better parent. We have a better marriage, family relationships, and even professional interactions because of these principles.

First, avoid extremes like “you always do this” or “you never do that.” If I said “Jordan, you’re always late for podcasts,” your first reaction would be defensive because it’s not true — nobody “always” or “never” does anything. People shut down and become defensive. You want to find the dialectic, avoiding black-and-white thinking. “Sometimes I feel frustrated when you’re late” is much more effective.

Second, when someone has an emotional outburst, they’re dysregulated for various reasons — hunger, a bad day, a lack of tools. If my daughter has a fit, she’s not thinking “I really want to have a tantrum and lose my phone for a week.” You appreciate that they’re doing their best with what they have — they just lack the tools.

Third, validate emotions. When someone’s really upset about something, you could say, “That’s ridiculous, it’s just a math test, you’ll get an A.” But explaining away someone’s feelings doesn’t help. People need validation. They become more emotionally frustrated when they don’t get that release valve. Saying “I can see you’re really upset” sincerely can help reduce their emotional burden.

Returning to Derek’s question about Gen Z, many people seem massively oversensitive. I grew up in Coney Island, Brooklyn, where people talked trash to each other, were rough around the edges, and the world wasn’t safe. You could say all kinds of things. I’ve raised my kids similarly — I don’t want them so soft that when someone says something offensive, they appeal to authority, running to school or teachers saying “he said this” or “she said that.” Take care of it yourself. Have a conversation. I’m not advocating violence, but handle yourself.

My kids have experienced this and tell us about various scenarios. The younger generation — whether you call them far left or woke — has lost some ability to engage with people. When encountering ideas or comments they disagree with, there’s hysteria.

I was a center-left Democrat my entire life. I didn’t vote for Trump, but I also didn’t vote for Kamala. I voted for Bloomberg-Romney, which was not a ticket, but that’s where my values aligned, in a non-swing state where my vote didn’t matter. I had the luxury of voting with my conscience.

I believe many center-left Democrats didn’t move right because they were attracted to the right. They were repelled by the histrionics, noise, whining, and complaining from the left. Enough was enough.

Share

Jordan Schneider: Let’s stick with parenting. You mentioned earlier that Steve Jobs accomplished incredible things, but everyone around him in his personal life had a totally different experience.

Josh Wolfe: They hated him. This is a fascinating phenomenon.

Jordan Schneider: This really resonated for me when I was reading a biography of Richard Holbrooke — the most storied American diplomat of the second half of the 20th century, though perhaps not well-known in the Asia-watching community. There was a line where his kid said his father wouldn’t recognize his grandchildren in a toddler lineup.

Growing up, I dreamed of being Secretary of State or a diplomat, but trading family relationships for that achievement gave me pause. Many people choose one path over the other. What are your reflections on this dynamic of accomplishing big things while being awful to the people closest to you?

Josh Wolfe: I used to debate this with one of my best friends, a famous journalist, who talked about how great Jobs or Elon were. My response was that the people around them — their families — hate them. They’re celebrated by strangers, feeling love from millions of people they’ve never met who celebrate a caricature of who they actually are. Meanwhile, their true character is loathed by those closest to them.

I’ve admired something I have to admit I’ll never achieve — I am too rough around the edges, have burned too many bridges, and have been abrasive to too many people. When Warren Buffett gave the eulogy for Coca-Cola CEO Don Keough, he summarized it in three or four words — “Everybody loved him.” I thought that was beautiful.

People will not say that about me, but I can control the decisions I make regarding my children. Will my kids feel that way about me?

I experienced something very salient and memorable involving my grandfather, who raised me. I grew up in Coney Island, Brooklyn — my mother, grandparents, and I shared a two-bedroom, one-bathroom apartment. We were very poor. My grandfather wasn’t my biological grandfather; he was my grandmother’s second husband, but he treated me like his son. He delivered the Daily News at night and was the most important man in my life.

He passed away the month before September 11th. At his funeral, his biological son was present. During the Jewish ceremony, when family members put dirt on the grave, his biological son approached with such animus. It was like “good riddance” — he took the shovel, threw dirt on the grave, and walked away. That sound still echoes with me.

I’ve always loved this quote from Carl Jacobi, the 18th-century mathematician: “Invert, always invert” — flip it on its head. My father wasn’t present in my life, so I became the father and husband he wasn’t. But watching that funeral moment, I resolved that would never happen to me. I never want my kids to feel that animus and animosity, wanting to dump dirt on my grave and walk away saying “good riddance."

The most important thing to me is what Adam Smith wrote about — not just the invisible hand, but the idea of being lovely and being loved. That second part, being loved, is scarce and valuable.

Scott Galloway recently gave a rant — I believe it was on Piers Morgan — where he talked about our obsession with Elon, innovation, and money while ignoring people who are hurting and suffering, including those losing USAID funding. He asked what it means to be a man. In his view, it wasn’t toxic masculinity — being a man means being able to take care of people.

We’ve lost that. There aren’t great male role models in public life today that young people celebrate for being good men through self-sacrifice. Everything centers on self-aggrandizement.

Regarding how to encourage or discourage people on certain paths: I want my kids to be truly happy. I ultimately don’t care where they go to college — the world has changed significantly. I don’t care what they do professionally. I want them to find purpose and meaning.

This relates to something interesting. Dan Senor works at Elliott Associates but is very active in Jewish and Israeli life. We attended Shabbat dinner at his house. I’m not religious — I’m an atheist but a tribal Jew. He blesses his children, and what struck me was that the blessing wasn’t about success, money, or career achievement. The blessing was “I want you to have a life full of meaning and purpose."

I thought that was beautiful. That’s what I want for my kids.

Philanthropy, Parenting, Short Videos

Jordan Schneider: We have five minutes for five quick questions. Take your pick — future of media advice for China Talk, New York mayoral race hot takes, what you’d do with a $10 billion philanthropic foundation, or something else.

Josh Wolfe: I don’t know.

Jordan Schneider: Where do you want to go?

Josh Wolfe: Mayoral race. I really hope Cuomo wins.

On advice for ChinaTalk, people get the audiences they deserve, and you have a smart, sophisticated, engaged audience looking for signal among the noise. Keep doing what you’re doing — your combination of Substack, podcasts, video content, periodic pieces, and great guests works well. I feel privileged to be here with you and enjoy reading your work.

Just maintain it because it’s a high-signal, no-BS voice. These things aren’t linear — there are step-change functions where suddenly something goes viral and you gain another thousand or ten thousand subscribers.

You should periodically write op-eds in major publications like the Wall Street Journal or Financial Times, sharing insights from your work. There are proxies like Stratfor and others in popular geopolitics, but you own a valuable niche. Over time, whether you do it or someone else does, there will likely be aggregation and acquisition — a China expert, Africa expert, defense expert — building a new media ecosystem where you could benefit from that outcome.

Jordan Schneider: Second and third questions — you’re pretty wealthy and likely to become wealthier. How do you think about big philanthropic investments beyond what you’re doing today?

Josh Wolfe: There are two aspects I really admire. Take what Bill Conway did — he’s not pursuing vainglorious naming opportunities. He literally identified deficits. His late wife was focused on nursing, so they fund nursing schools and programs because we have absolute scarcity there.

Considering the arc of AI, I believe in abundance and scarcity dynamics. What’s abundant will be machines helping with intellectual tasks — white-collar jobs will be hit while blue-collar jobs will surprisingly be safer than people think. The care aspect of healthcare will be critical.

If you ask what I want my kids doing: I grew up playing extensive video games and watching tons of TV, and I allow them that — not Monday through Thursday, but weekends they can play games, watch TV, engage in pop culture, watch sophisticated content like Fareed Zakaria and Jeopardy.

But I want them fully versed in AI. My 9-year-old is better than my wife at ChatGPT queries and Midjourney prompts for images. It’s creative expression.

Jordan Schneider: That’s a great age for it, right?

Josh Wolfe: Absolutely. My oldest daughter had to do an evolution project in seventh grade by hand — organisms with sharp teeth survive candy rain while weak-toothed ones die. She drew it manually. My current seventh-grader in that same class is using one of our companies, RunwayML, for AI video generation, creating full videos of different organisms in her scenario.

I want them totally versed in AI because ultimately — and you discussed this recently, maybe with Wang — the power isn’t in who has the chips, but who’s using them. Similarly, power isn’t in who has the applications, but who’s using them.

There’s a significant push for the world to use US-driven open-source or closed models rather than China-driven models that approach an asymptote of truth but never discuss Tiananmen Square, Xinjiang, or Uyghurs. We want the Global South influenced by American ideals of truth, Popperian hypothesis, conjecture, and criticism rather than Chinese systems.

But I want my kids using all these tools and understanding them. What will be scarce against all that abundance is human connection. They need to understand people, make eye contact — you’d be amazed how many kids, because of screens, have awkward, almost autistic interactions.

Being able to connect with people, understand them, read Shakespeare — that’s timeless. People change, costumes change, stages change, but human nature hasn’t changed since the Pleistocene African savannah.

That’s what I’d fund philanthropically. Derek mentioned CBT programs for young people earlier. I started a charter school 17 years ago in my native Coney Island, Brooklyn. We began with 90 fifth-graders in the projects. Now we have 1,000 scholars, 200-plus faculty, 100% college acceptance rate for first-generation college students. Eighty percent of families qualify for free and reduced lunch — a euphemism meaning a family of four makes less than $30,000, which is insane.

These families lost the ovarian lottery — the classic John Rawlsian veil of ignorance. These kids are no less intelligent than those born in Greenwich, Connecticut. But there’s no Army recruiting station on Greenwich’s Main Street — there is one on Cropsey and Stillwell in Coney Island. That’s not fair.

Those are worthy targets for philanthropic dollars.

Jordan Schneider: Okay, but let’s start from a $10 billion bucket. What are we talking about here?

Josh Wolfe: Where would I give? I would fund universal CBT for everybody in the country. I don’t know that it needs that much money — it just needs celebration in the country that helps people become more emotionally regulated and be their best version of themselves. It will reduce problems in our criminal justice system. It’ll reduce problems in corporate America. It’ll reduce a lot of problems across the board.

Jordan Schneider: Not to rag on you too much — you give very sophisticated answers to how to invest in the future of science and technology. That was a fine answer. But a lot of people, at a certain point in their life, switch from the sort of answers we discussed in the first 80 minutes to the question I just asked you. I’m curious: do you see the future differently when it comes to philanthropy?

Josh Wolfe: Where we give philanthropically right now reflects things that we prioritize. For me, complexity science through the Santa Fe Institute — brilliant people. I love it. I believe that’s a source of tremendous value. I’ve been part of that for 10 years as a trustee and believe deeply in it.

The charter school movement — I believe deeply in that because it’s a form of civil rights for people. My mom was a public school teacher, so this hits close to home.

Jordan Schneider: Let me try one more time. This is more of a meta question. The sort of investigation that you need to understand how to use philanthropic dollars efficiently and effectively — I’m curious how similar or different you think that is from investing?

Josh Wolfe: The only similarity is finding an amazing social entrepreneur. It’s like when we started the charter school — we basically backed this guy Jacob Newkin, who was starting the school. It’s the same thing. I used to talk publicly about Jacob: he’s the greatest social entrepreneur that I’ve backed by spending my time and money with him.

But we weren’t doing analysis on the market and the unmet need and that kind of thing. It tends to be something personal. For my wife, it’s the Center for Reproductive Rights. She’s on the board there, making sure that women have access to contraception and abortion and autonomy over their bodies. That’s a really important thing to her.

She’s not doing an analysis of where’s the best place to give or whether we should give more. It’s just: Roe v. Wade got overturned. There are women who are going to die in certain states because they can’t get abortions. What can we do about that? She gives a lot of money.

Jordan Schneider: Maybe we’ll close on this topic. If we’re entering a world where science receives less funding — Danny Crichton, who works for Lux Capital, wrote a really interesting piece about this — when the total amount of science that the US funds decreases, there’ll be a little spillover to China and the EU, but we’ll just have less science overall. Beyond CBT, what encouraging developments do you see for science and technology’s future? What basic research do you think people should be funding — the stuff that’s too risky for any venture capitalist to invest in?

Josh Wolfe: This might be controversial, but people should be spending far less money on climate philanthropy. The answer lies in what I call elemental energy and nuclear power. All that money should be redirected toward early-stage science and psychological and behavioral health research, because that will make society better.

The Gates Foundation, Bloomberg Philanthropies, Ford Foundation — these organizations were captured by the climate movement over the past 15 years. Honestly, I don’t know what it’s accomplished. It’s been a colossal waste of money.

Jordan Schneider: Speaking to the culture issue you raised earlier — it’s striking to me. I worked at an organization that was half nonprofit, half research fund, focusing on China and climate. On the climate side, they could get money to fund literally anything.

Josh Wolfe: It’s popular and makes you feel good. You go to a cocktail party and say, “Oh, I’m funding climate research.” Great, you’re doing wonderful work — Al Gore would be really proud. But the money is misplaced.

Jordan Schneider: It’s almost downstream of the culture. Who are the funders and trustees, and what’s popular with them? They’re not scientists conducting expected value calculations on human flourishing or whatever. Not to disparage anyone, but it’s something that resonates with them personally.

Josh Wolfe: Look, Sam Bankman-Fried was the emblematic figure of this, but the effective altruism movement was rational in trying to determine where we can do the most good. They approached it economically, looking at low-probability, high-magnitude events and identifying opportunities where small amounts of money could have significant leverage.

Going back to Conway’s nursing initiative — that’s not popular. People don’t get excited about addressing the massive nursing shortage. But he identified this as crucial, and they’re putting several billion dollars behind it. That’s noble work.

Bloomberg’s urban initiatives and charter school funding are excellent. People funding the arts because of personal passion — that’s great too. But we have massive problems with criminal justice reform and behavioral health domestically.

I’m not talking about everyone needing mental health days, but implementing cognitive behavioral therapy in schools at a young age. Before children’s prefrontal cortex develops at 25, we could help them develop better self-regulation. The world would be a much better place.

It would be tremendous to see philanthropists return to funding institutions like the old Cavendish Laboratory — putting billions of dollars into institutes that enable knowledge discovery.

Jordan Schneider: Rockefeller University — that’s incredible work. People need to get with the program.

Josh Wolfe: Carnegie Mellon, exactly. These institutions started with robber barons who decided to redirect money into academic institutions. This will happen again.

If you examine philanthropic funding historically, it began with private individuals, then government labs like Los Alamos, followed by Bell Labs (born from monopoly), then IBM Research. IBM centralized initially, then distributed with locations in Zurich and Almaden. Then came Google, Intel, and Microsoft Research.

Some of these corporate labs are under pressure now because they haven’t yielded significant results. But we’ll see the rise of private labs. You can see this already with the ARC Institute.

Jordan Schneider: Absolutely.

Josh Wolfe: The Collison brothers are major supporters there. The Chan Zuckerberg Initiative represents major scientific initiatives comparable to Howard Hughes — himself a former defense contractor who invested enormous amounts into what became the Janelia campus, now one of the great sources of Nobel Prize winners.

Jordan Schneider: That’s a nice place to close. 50% of Howard Hughes researchers came to the US on visas! Joshua, thanks so much for being part of this.

Josh Wolfe: Great to be with you, man.

Jordan Schneider: Awesome. We’ll do this again in 10 years, and then you’ll be saying, “Here’s this philanthropic vision — look at all these molecules I found."

Josh Wolfe: I believe mostly in free enterprise, science, and technology. Early-stage ventures will handle a lot of that, but not the basic science research. There’s no market for that.

Bonus Riffs on Books

Cool. What are you reading? Anything good? Binge-watching anything inspiring or fun?

Jordan Schneider: Two books. I’m a new parent — I have a 10-month-old at home.

Josh Wolfe: Boy or girl?

Jordan Schneider: Girl. First child.

Josh Wolfe: Wow.

Jordan Schneider: I read all these books on paternity leave about “How I Raised My Child in X Country.” They were not good. But one was excellent: Italian Education. It’s by a cranky British guy who married an Italian woman and raised his kids in late 1980s-90s Verona. All these other books are basically backhanded critiques of American parents from whatever direction. But he’s just observing this really interesting, weird society where you have very tight connections between parents and kids — for better and worse, from my perspective, across many dimensions.

It was engaging, funny, smart, with nice vignettes. Each chapter stands on its own, which is good for 3 AM reading.

Book number two — A Social History of the Machine Gun by John Ellis.

The book tells the story of the machine gun through different lenses. He’s a military historian who wrote about World War I and World War II tactics. But in this book, he explored the technological evolution from the Maxim gun all the way through World War II — who invented it, why, and where it came from.

There’s a fascinating acquisition story too, because people didn’t think it was real and didn’t want to buy it. There were prototypes but no factory yet, so you had these hucksters trying to —

Josh Wolfe: What years was this?

Jordan Schneider: At the start of World War I, the British had one gun for every 2,000 people. You had to get to 1916-1917 for them to actually be making and buying enough machine guns.

Josh Wolfe: That’s amazing.

Jordan Schneider: It took an enormous amount of time. The technology was already there in the 1880s and 1890s. There were examples from wars in different countries — the 1905 Russo-Japanese War, the Crimean War. You could see it if you were looking at the right things.

Ellis gives examples of smart colonels saying, “Guys, we need to buy these guns — they’re a big deal.” But people said no, and they all said no way too late.

The other story he tells is about the psychology of not just the people buying the guns, but the officers themselves who had to abandon their mindset about what made a successful officer. Being a sharpshooter wasn’t considered honorable. What won wars throughout the 19th century was discipline, standing in line without fear, marching together to bring maximum power. That’s what the technological paradigm demanded — willingness to maintain rank.

It took over 100 years for people to change their mindsets and understand that you actually need to be distributed, use natural cover on the ground, and get away from the Napoleonic mindset of gallant charges. Those charges were the correct evolutionary answer in a different time period, but not by the US Civil War, definitely not by the end of the 19th century, and absolutely not by World War I.

Josh Wolfe: It’s interesting — the juxtaposition of the two books. One is arguably about technology of life (all parenting is a form of life technology), and the other is about technology of death. It’s a nice contrast in what you’re reading.

Quick parenting observations — First, if you walk into a bookstore like Barnes & Noble — which really don’t exist anymore —

Jordan Schneider: There’s one three blocks away!

Josh Wolfe: The mere existence of a handful doesn’t change that they’ve largely disappeared. There used to be 20 in New York City; now there are one, two, or three.

My point is, when you go into any section — investing, relationships, or parenting — and see 200 books, it means nobody has any idea what they’re doing. If they did, there would be one book with all the answers.

What I learned, especially with our first child, is that you develop a whole bunch of lessons, then you have a second child and they’re all wrong. Once you have more than one child, the nature versus nurture debate is settled. They are genetically different from day one — their predispositions, attitudes, sleep patterns, crying patterns, wants and needs. Their personalities persist from birth. The one who’s more reactive, the one who’s more smiley — it’s absolutely fascinating.

Second point on parenting, which relates to China — when I was growing up, my mother said, “You need to learn golf and Japanese because that’s the lingua franca of business in the ’80s.” Then it became, “You need to learn Mandarin and coding because that’s the lingua franca."

Neither of those things really matter today. The pace of AI development means coding can now be done by AI agents. Computer scientists who thought they were in a valued position are suddenly thinking, “Oh my God, I’m being replaced by agents."

Translation is pretty incredible now too. I don’t know what the next parental trope will be, but it’s usually wrong.

Subscribe now

Jordan Schneider: We can take it back to the machine guns, right? What was successful for you as you were coming up in the world and in the institutions that shaped you isn’t necessarily going to be the thing for the next era. The humility to understand that — both from a defense acquisitions perspective and a parenting perspective — is really hard.

Josh Wolfe: I’ll give you an investing version of that. First, the most dangerous words everyone always says are “this time is different” — because it never is. If you know Shakespeare, then of course it’s never different. Human nature is constant.

But when parents utter certain words, it’s predictive — similar to the defense acquisitions issue or for investing. Want to know what will be the next $10 billion industry? Here it is: “It will rot your brain.” Every time a parent says “it will rot your brain” about something they don’t want their kids doing, that thing becomes the next massive industry.

Rock and roll in the ’50s, TV in the ’60s and ’70s, chat rooms in the ’80s and ’90s, video games in the ’90s and 2000s — every one of those things that was the target of parental ire became the next $10 billion industry. Tipper Gore with parental advisory lyrics and rap music — rap became the biggest genre of music in the following decade.

Just listen to what parents are terrified about right now. The gamers became our modern robotic surgeons and drone pilots. Whatever they’re freaking out about now — maybe TikTok (though I have problems with that for different reasons) — it could be movie-making on social media or whatever.

Josh tries to defend short video

Jordan Schneider: This is the hardest question of the day — give me the optimistic short video take.

Josh Wolfe: We’ve democratized the ability to have creative expression with special effects that used to cost $50 million. Filmmakers used to be siloed in studios with hierarchies, casting couches, and Harvey Weinsteins — awful people. Now there’s freedom of expression where people can create tragedy, drama, comedy, and surrealist content with these tools at their fingertips.

My 9- and 12-year-olds are better filmmakers than I was at 25. They have tools today that Hollywood executives used for Terminator or The Abyss — remember James Cameron’s special effects that seemed amazing back then? That’s from a creator perspective.

Jordan Schneider: What about from a consumption perspective? Having culture delivered in 30-second chunks?

Josh Wolfe: Again, look at it from abundance and scarcity perspectives. I 100% agree that if you’re constantly being trained for short attention spans, that’s problematic. We literally practice patience as a family because I know there are so many competing things offering fast dopamine hits and quick responses. We do long periods of quiet reading from physical books as a family. We watch long movies instead of 30-minute segments.

Jordan Schneider: No, but you’re telling me the things parents are scared of are actually going to be the future. What’s positive about consuming content in 30 seconds?

Josh Wolfe: I’d argue that your ability to process multimodal information is far better. Look at the average older person right now — they’re focused on one thing, they’re slow. You’re probably able to switch between a WhatsApp chat group, Twitter, watching a short video, and checking emails. Your ability to multitask while retaining the ability to function is super valuable.

Jordan Schneider: Your heart’s not in this one…

Josh Wolfe: It’s okay, but here’s what I know: it’s not as bad as people think. The TikTok stuff is concerning, but short-form video generally isn’t bad.

Jordan Schneider: Okay.

Josh Wolfe: I’m optimistic on the science and technology piece. I’ll still be skeptical about the human nature piece, but optimistic about science and technology.

Jordan Schneider: All right, let’s call it there then. This was really fun. Do you have a book to shout out?

Josh Wolfe: Let’s close on fiction and nonfiction. For nonfiction: The biggest debate my wife and I have had was over a book by Robert Sapolsky, a Stanford primatologist and neuroscientist who’s written a series of books. His first one was Why Zebras Don't Get Ulcers: The Acclaimed Guide to Stress, Stress-Related Diseases, and Coping. The punchline was that zebras run for their lives in their ancestral environment, then they’re calm. We have constant stressors all day long that didn’t exist in our ancestral environment.

But his more recent book is called Determined. Either you’ll want to throw it across the room, or you’ll want to send copies to everybody. It depends on whether you agree — as I do — that we do not have free will, or disagree — as my wife does — that we are filled with agency and of course have free will.

Danny Kahneman was a friend. His belief before he died was that neither free will nor consciousness exist — that they are both useful illusions. I very much subscribe to that view.

For fiction, Amor Towles has a short story collection called Table for Two. He wrote Rules of Civility, which is great. There’s a character from the second half of that book who gets extended treatment in a few more chapters in this collection.

But there’s one story that deeply touched me called “The Bootlegger.” It’s a short story that takes place in New York. The beauty of this, for me personally, was that I happened to post on Twitter about my love for this particular story. Amor Towles replies, which was very meaningful, and he says, “This is probably the most autobiographical story I ever wrote."

It’s a relatively short story set here in New York. A young couple, something happens, they go to Carnegie Hall, and this story unfolds. There’s a particular classical musician referenced in this fictional piece, and he replies to the story because he’s actually in it. It was this surreal fictional story that was a slight roman à clef of Towles’s life. He replies to me on Twitter, then the classical musician does too. I was like, “This is amazing."

Jordan Schneider: That’s really fun.

Josh Wolfe: I highly recommend Table for Two by Amor Towles. It’s a set of vignettes around two characters — probably 10 or 12 short stories. But “The Bootlegger” is awesome.

Jordan Schneider: Thanks so much.

Josh Wolfe: Great to be with you.

亲爱的读者周二好，六月的新书讯推荐五本 抗日战争/日中戦争/Second Sino-Japanese War 相关新书。

重探抗戰史

郭岱君主編，《重探抗戰史（一）：從抗日大戰略的形成到武漢會戰（1931-1938）》，台北：聯經出版公司，2015年。

郭岱君主編，《重探抗戰史（二）：抗日戰爭與世界大戰合流1938.11-1945.08》，台北：聯經出版公司，2022年。

郭岱君主編，《重探抗戰史（三）：抗戰與中國之命運》，台北：聯經出版公司，2022年。

斯坦福大学胡佛研究所的郭岱君教授召集迄今为止最豪华的多国历史学家团队，历时十年，以胡佛研究所馆藏的蒋介石日记、蒋经国日记、宋子文档案和孔祥熙档案等为基础，参照日文、英文多国档案，重新探索抗日战争中的一些核心问题：

日本侵占满洲（中国东北）之后，并无在中国大规模用兵的计划，为何 1937 年七月华北一次地方小冲突却迅速上升为两国之间的全面战争？

日本为何放弃北进，转而南进，最终走向珍珠港？

中国的抗战和第二次世界大战是什么关系？

汪精卫为何离开国民政府？

中共如何在抗战中坐大？

美国为何厌弃蒋政府？

三卷本读完，相信聪明的读者能找到一些初步答案。

被遗忘的盟友

芮纳·米特（Rana Mitter）著，林添贵译，《被遗忘的盟友：揭开你所不知道的八年抗战》，台北：远见天下文化出版股份有限公司，2014。

Mitter, Rana. China’s War with Japan, 1937–1945: The Struggle for Survival. London: Allen Lane, 2013.

牛津大学/哈佛大学历史学家 Mitter 的代表作，一卷本的抗日战争史。

请记住一定要读台版或英文版，不要读新世界出版社的简中删改版。

Racing the enemy

Hasegawa, Tsuyoshi. Racing the Enemy: Stalin, Truman, and the Surrender of Japan. Cambridge, MA: Belknap Press of Harvard University Press, 2005.

这本书已经出版 20 年，算不得新书。但鉴于如此重要的书依然没有中文版，还是要介绍一下。

日本为何无条件投降？传统观点归因于美国扔了两颗原子弹。而长谷川毅充分利用苏联和日本档案，提出了挑战性的修正观点：相比原子弹，苏军参战起得作用更大, the Soviet entry into the war played a greater role than the atomic bombs in inducing Japan to surrender.

日本曾将希望寄托于《日苏互不侵犯条约》和苏联调停来获得有条件的停战，而苏联参战导致所有希望破灭。

大棋局中的国共关系

吕迅著，《大棋局中的国共关系》，北京：社会科学文献出版社，2015年。

1945 年八月风暴刮起，苏军出兵满洲（中国东北）一举击垮关东军，不仅导致了日本投降，更是获得了战后中国命运的决定权。

吕迅这本将视野拉宽，从抗战后期写到韩战爆发，充分展示了抗战如何改变中国命运：冷战是“内战”的延续，内战是二战的延续。

戰火中國 1937–1952

方萬德著，何啟仁譯，《戰火中國 1937–1952：流轉的勝利與悲劇，近代新中國的內爆與崛起》，台北：聯經出版公司 2020。

van de Ven, Hans. China at War: Triumph and Tragedy in the Emergence of the New China, 1937–1952. Cambridge, MA: Harvard University Press, 2018.

与吕迅思路相似，剑桥大学的方德万教授也把抗战、内战和韩战视为一个连续之整体，审视其中的关键节点，比如是日军 1944 年发起的一号作战（豫湘桂战役）导致了国共实力对比彻底失衡。

以本书压轴，因为《戰火中國》叙事极为流畅，雅俗共赏，适合新手读者入门。如果你只有时间读一本书，选它没错。

以上就是六月的新书讯，感谢您的阅读。如果喜欢这个专栏，请推荐给家人朋友订阅，谢谢：

In 2011, China’s drug regulator cleared the nation’s first home-grown targeted cancer pill. Fourteen years later, a Chinese bispecific antibody is aiming to knock the world’s top-selling oncology drug off its perch.

Those two bookends frame twin contests now running in parallel. One is humanity’s decades-long fight against cancer, a disease that still claims one in six lives worldwide. The other is China’s effort to move from importing medicines to inventing them. This article seeks to follow these battles in real time by tracing the stories of four milestone therapies and their makers.

Founded in 2003 in Hangzhou by a team of returnee talent, Betta Pharmaceuticals launched the first targeted anticancer drug developed in China to reach the market. The success of its drug Conmana1 in China dovetailed with sweeping policy reforms aimed at making home-grown medicines more trusted, innovative, and affordable. The drug did not, however, make it outside China, despite Betta Pharma’s best efforts. Conmana is an example of “me too” innovation: a variation on an existing drug that performs just as well but not quite better enough to make it globally competitive.

Created through a rigorous internal drug development program in 2012, BeiGene’s Brukinsa did what Betta’s Conmana could not – it went global, becoming the first Chinese cancer therapy approved by the US FDA. A “me better” innovation targeting blood cancer, Brukinsa now reaches people in over 65 countries, bringing in over US$2 billion in sales annually.

Betta’s Conmana and BeiGene’s Brukinsa are both small-molecule drugs, meaning they are created through chemical synthesis. The real frontier of innovation that excites biotechnologists is biologic drugs, biomolecules such as engineered proteins or RNAs that act with greater power and precision inside the body.

Carvykti is one of the first innovative biologic drugs created in China. Initially discovered and tested in Xi’an Jiaotong Hospital by Nanjing Legend Biotech, the drug reached global markets with the help of Johnson & Johnson. First gaining US FDA approval in 2022, Carvykti now has regulatory approval across over 36 countries and has treated over 5,000 patients, with more to come.

The last story is unfinished. It’s about Akeso’s ivonescimab, an icon of Chinese biotech innovation in mainstream media. A biologic drug with a novel method of targeting cancer, ivonescimab received approval by the Chinese NMPA in April 2025. In countries like the US, where Akeso has passed the baton to US-based Summit Therapeutics to develop the drug, ivonescimab is still in the clinical trial phase, meaning it has yet to pass through the regulatory gauntlet. What’s exciting and undecided about ivonescimab is its potential to go head-to-head with the world’s best-selling drug, Keytruda. Whatever happens to ivonescimab over the next few months in trials outside of China will send a signal of exactly how successful Chinese biotech innovation has become. The world of biotech will be watching closely.

But let’s start from the beginning. Betta Pharma’s milestone achievement of Chinese regulatory approval for a new innovative drug (2011) happened only 11 years before Akeso’s ivonescimab earned a US$5 billion deal with Summit Therapeutics (2022). How did we get here?

Betta Pharma: The “Me-too” Era

In the 1990s, biopharmaceuticals — specifically oncology, the study and treatment of cancer — entered a new era of innovation. Up until the early 2000s, doctors primarily combated cancer with broad-stroke methods like surgery and chemotherapy. Now, new and improved methods were emerging: targeted therapies that zeroed in on cancer cells while minimizing damage to healthy cells and immunotherapies that helped the body’s immune system recognize and attack cancer cells.2

This revolutionary effect of molecular biology captured the attention of many bright scientists and doctors in China, including Dr. Wang Yinxiang 王印祥. Born in rural Hubei, he spent three years working in public health and three years completing a Master’s degree at the Chinese Academy of Medicine before he could truly follow his passion for oncology to the United States, where he earned a doctorate from the University of Arkansas.

Dr. Wang got his wish to do cutting-edge research as a postdoc at Yale, where he dove into one of the first targeted cancer therapies, Novartis’ Gleevec. Sharing his apartment was Ding Lieming 丁列明 – another Chinese transplant with a University of Arkansas MD. On strolls through New Haven’s Science Park, the two friends along with medical chemist and entrepreneur Zhang Xiaodong 张晓东 bonded over more than just science. They shared a bigger dream: to bring the newest in biotech to China.

Dr. Wang and Dr. Ding would ultimately join forces in 2003, when they founded Betta Pharmaceuticals to develop targeted cancer therapies in China. Betta opened its doors with a shoestring team – just 13 people, many of them what Dr. Wang affectionately called “kids,” fresh from bachelor’s or master’s program and learning on the fly. Nevertheless, they managed to develop icotinib (later sold as Conmana), a drug engineered to target EGFR proteins as a way to inhibit cancer cell growth.

In those years, China’s pharmaceutical industry hadn’t left the nest. Manufacturing of cheap, generic drugs dominated. To domestic investors, companies, and physicians, a business model built on new drug development was unthinkable: the costs and risks were too high, the regulatory process was a mess, quality and safety were still iffy, and previous such attempts had failed. Winning a clinical-trial slot for Conmana (a prerequisite for proving the drug could outshine current care) was nothing short of herculean. When the Peking Union Hospital director declared the study too risky and tried to dismiss him, Dr. Wang stood firm for ninety minutes, knocking down every objection until the approval stamp finally came down in his favor.

In 2009, Conmana made it to Phase III clinical trials, the make-or-break test of wide-scale efficacy. In Phase III, the team pushed the envelope again: rather than testing against a placebo, they pitted their molecule against AstraZeneca’s gefitinib (the world’s first targeted anti-cancer therapy) in the first Chinese study to challenge an imported standard head-to-head.

The study’s results, announced by leading academic Sun Yan at the 2011 World Lung Cancer Conference – also the first time a China-developed drug headlined an international academic forum – showed that icotinib could match the cancer-fighting power of the imported benchmark while causing fewer side -effects. Conmana, in other words, was a successful “me-too” drug, an incremental improvement on an existing pharmaceutical innovation.

After six grueling years, Conmana earned its first regulatory approval from China’s National Medical Products Administration (NMPA; although at that time it was still the CFDA).

Betta’s success as a fast follower of a next-generation cancer therapy was a triumph for Beijing’s returnee talent and national innovation programs. Conmana’s success had been fueled by funds from the Yuhang District Government of Hangzhou, the “863” Program, and the “11th Five-Year Plan” National Science and Technology Major New Drug Special Project.

The government showered Betta with accolades: the China Overseas Chinese Contribution Award, the gold prize for patents, first prize for the National Science and Technology Progress Award, and more. Chen Zhu, then Minister of Health, praised their achievement as “an emblem of ‘Two Bombs and One Satellite 两弹一星’ in the field of public health,” referencing a techno-nationalist ideal of a whole-nation project for science and technology development.3

However, for all its homegrown glory, Betta’s blockbuster never crossed the border.

In 2014, with the help of Roswell Park Cancer Institute, Betta filed to run a clinical trial for Conmana in the US, the first step towards seeking US FDA approval. However, the trial was shortly withdrawn. Newer, later-generation EGFR inhibitors were eclipsing Conmana’s performance, and there was no sense investing in trials when the product was unlikely to sell.

Though just a “me-too” innovation, Conmana was a landmark accomplishment for China. On the domestic market, the drug was meaningfully cheaper than its imported alternatives. Given the rapidly growing incidence of lung cancer in China, Conmana’s improved affordability made a real difference in patients’ lives. Still, it would take almost another decade before a China-developed cancer drug would make a global impact.

BeiGene: A “Me-Better” Drug Goes Global

China’s mix of capital, talent, and policy reforms was turning its budding biopharma sector into a global magnet.

Among the first drawn in was Pittsburgh native John Oyler. Familiar with China through his work at McKinsey in the 1990s, Oyler was stunned by the country’s science and technology progress when he returned the next decade. By the mid-2000s, regulatory harmonization, returning talent, and improved manufacturing infrastructure enabled China to meet the needs of global pharmaceutical companies, leading to the growth of contract research organizations (CROs), which provide outsourced medicinal science services. Oyler co-founded one such Chinese CRO, BioDuro, in 2005.

But serving foreign pharma clients wasn’t enough. As China moved toward deeper healthcare reform, Oyler saw an opening for homegrown innovation: “[China] had the capability to pour tens of billions of dollars back into the global industry to help pay for more research, which would not only make drugs more affordable in China, but across the globe.” Rather than repeat the trajectory of BioDuro, which was eventually sold, he wanted to create something enduring. “I wanted to build something here — in China — that is lasting, impactful, involved in great science, and can really help people,” he said. He aspired to create a company capable of developing world-class cancer therapies from a country many still underestimated.

To bring his vision to reality, Oyler needed scientists. He connected with Dr. Xiaodong Wang 王晓东, a top Chinese American academic biologist who recently returned to China to lead the new Beijing Institute of Life Sciences. The chance to work with Dr. Wang — a superstar scientist admired widely enough to impress parents at Chinese New Year — proved an effective tool for attracting talent. Together, they founded BeiGene in Beijing in 2010 to become the “Genentech of China.”

Early on, BeiGene focused on BTK inhibitors, a targeted cancer therapy that works by blocking cancerous B-cells’ ability to grow. The first BTK inhibitor, synthesized in 2007, showed promise but caused significant side effects. In 2012, BeiGene initiated a discovery program in San Mateo and Shanghai to develop a better BTK inhibitor. After screening over 3,000 compounds, the team identified the highest-potential molecule that would eventually become Brukinsa (zanubrutinib).

BeiGene aimed to take Brukinsa global from day one. To support worldwide approvals, the company built a 25-country trial program in which approximately 90% of patients were enrolled outside of China. The numbers delivered: Brukinsa consistently beat first-generation BTK inhibitors on safety and efficacy, turning a presumed “me-too” into a clear “me-better” that is now the standard of care for B-cell cancers.

Momentum snowballed. In 2019, Brukinsa set a precedent as the first Chinese-developed cancer therapy to win FDA approval, months before China’s own NMPA signed off. It has since secured clearances in 65-plus markets spanning the US, EU, Canada, Australia, Japan, and China, and now supplies more than half of BeiGene’s revenue with US$2.6 billion in 2024 sales. The company (rebranding as BeOne) has likewise gone global, conducting trials in over 45 countries.

Legend: A True Chinese Biologic

Witnessing the breakthrough of firms like BeiGene, Beijing set its sights on higher-value innovation. The State Council’s 2016 13th Five-Year Plan therefore called for “leapfrog development in the biopharmaceutical industry,” spotlighting cell and gene therapies, antibodies, and vaccines.

These platforms fall under biologics — large molecules derived from biological processes such as insulin and hormones — rather than small-molecule drugs such as Betta’s Conmana and BeiGene’s Brukinsa. Because biologics are bigger and more structurally complex, their effects are harder to predict and their manufacture far costlier, but they open therapeutic doors that chemistry alone cannot.

Founded in 2014 as a subsidiary of GenScript, Legend Biotech embodied the kind of biologics leadership the state now prioritized.4 After losing his father to cancer, co-founder Frank Zhang 章方良 united with Chief Scientific Officer Dr. Frank Fan to create the firm with the goal of advancing oncology.5

At that time, influential scientific journals, Big Pharma deals, and first-in-human successes had converged to position cell and gene therapies as the vanguard of biopharmaceutical innovation. Early successes led by the University of Pennsylvania highlighted the potential of CAR-T therapy,6 a type of treatment in which a patient’s disease-fighting T-cells are genetically engineered to seek and destroy cancerous cells.

Driven by the promise of next-gen cancer therapy, Legend’s 19-person team, working in “a room the size of a freight elevator,” crafted a second-generation CAR-T treatment targeting multiple-myeloma tumor cells (later sold as Carvykti). Leading the research was Dr. Frank Fan, who had studied at Xi’an Jiaotong University and worked at the Xi’an Jiaotong Hospital. Leveraging those ties, Dr. Fan was quickly able to initiate Legend’s first CAR-T clinical trial at Xi’an Jiaotong Hospital, turning the startup into a “dark horse” contender in the CAR-T space.

At the 2017 annual meeting of the American Society of Clinical Oncology (ASCO), Legend presented early-phase data from its trials in China: durable remissions in multiple-myeloma patients with only mild side effects. The first of its kind accepted for review by China’s NMPA, this CAR-T candidate signaled that the country could innovate well beyond small-molecule chemistry.

Such promising results drew members of the global life sciences community. J&J’s Janssen signed a partnership with Legend Biotech in 2017. This was a case of an outlicensing deal: when a company (such as Legend) sells or grants rights to a drug candidate, passing the baton to a different company (such as Janssen), which who takes on the responsibility to bring the drug through testing, approval, manufacturing, and commercialization. Such transactions are the lifeblood of the industry. However, the high-potential molecules worth such high-profile deals didn’t historically come from China.

With the combined efforts of Legend and Janssen, Carvykti won FDA approval in 2022, soon followed by clearances in the EU, UK, Japan, and Canada. Though not first-to-market, Carvykti’s superior clinical value crowned it best-in-class. This then completed China’s rapid climb from “me-too” copies, through Brukinsa’s “me-better” gains, to a world-leading biologic breakthrough.

Akeso: a top challenger emerges

Now, Akeso Biopharma’s new molecule is drawing notice as a likely first- and best-in-class therapy from China.

Akeso started as the dream of Dr. Michelle Xia, a Gansu native. While working for California-based Crown Bioscience and in other roles in the US and UK, Dr. Xia grew frustrated with the eight-to-ten-year delay it took for innovative overseas therapies to reach Chinese patients. So she and three partners founded Akeso in 2012, naming it after the Greek goddess of healing, with a mission to develop home-grown therapies for cancer and autoimmune diseases.

Akeso’s edge is bispecific antibodies (BsAb), a type of next-generation cancer treatment involving engineered proteins that strike two targets at once, such as igniting immune cells while starving tumors. The first proof arrived in 2022 when China’s NMPA cleared Akeso’s Kaitani (PD-1/CTLA-4), the world’s first commercialized BsAb.

Next came ivonescimab, a PD-1/VEGF bispecific now pushing Akeso onto the global stage. By jointly blocking an immune checkpoint and tumor blood-vessel growth, it qualifies as first-in-class – an industry term for a drug that introduces a truly novel therapeutic approach.

In June 2022, Akeso unveiled the Phase II results of ivonescimab at the annual ASCO conference, showing strong responses in non-small cell lung cancer. Just months later China’s NMPA granted the drug Breakthrough Therapy status for three medical use cases, enabling closer guidance and fast-tracking its review.

Sensing ivonescimab’s scientific and commercial potential, US-based Summit Therapeutics inked a massive deal in December, licensing Akeso’s innovation for up to US$5 billion.7 The corresponding press release hailed Akeso’s innovation as “the PD-1 / VEGF bispecific antibody that is most advanced in the clinic,” noting that neither the FDA nor EMA had yet approved any PD-1-based bispecific therapy.

The size and significance of this agreement marked a bellwether moment in Chinese biotech innovation. Blue-chip investors and multinationals began scouting the country for genuinely novel assets rather than low-cost manufacturing plays.

China’s slice of global out-licensing has since tripled to 12%, with deal value leaping from US$35 billion in 2023 to more than US$46 billion in 2024. This trend seems set to continue through 2025 and beyond.

The fate of ivonescimab and most other compounds covered by these recent deals is uncertain. Many drug candidates are purchased in the preclinical or Phase I stage of development, requiring another 5+ years and US$300+ million dollars before they pass the clinical and regulatory hurdles necessary to make it to market – or, they will fail, like roughly 90% of compounds that enter human trials. Only five China-originated drugs have ever cleared the US FDA (BeiGene’s Brukinsa and Legend’s Carvykti are two of them).

Ivonescimab could be next. Akeso’s molecule has already completed Phase I and Phase II, and is in the midst of several high-stakes Phase III trials. Its Phase III trials with Chinese patients have already demonstrated success, leading to two recent NMPA approvals in the spring of 2025. The defining test comes next: a Summit Therapeutics-run Phase III study spanning 108 locations in 12 nations, where the drug candidate must outshine oncology’s gold standard, Keytruda.

Keytruda (drug name pembrolizumab) has been described as “era-defining,” the “800-pound gorilla” of the class of drugs to which ivonescimab also belongs. Developed and commercialized by American multinational Merck, the drug has been approved for 41 indications8 across 18 types of cancer. It’s the world’s best-selling drug, raking in about $29.5B in 2024 – nearly half of Merck’s total revenue. Its upcoming 2028 patent expiry opens the field for new challengers.

Akeso’s ivonescimab could be one such challenger. If its global Phase III trial confirms that the drug’s positive risk-reward results can extend beyond China’s borders, Akeso’s first-in-class molecule may eventually also prove to be best-in-class, making it one of the biggest biotech stories of the decade.

On the horizon

Together, Conmana, Brukinsa, Carvykti, and ivonescimab trace a clear, but incomplete, arc: China’s pharma sector has evolved from reverse-engineering proven ideas to originating drugs that can contend for global standards. Each milestone marks a step from “good enough at home” to “competitive abroad,” showing how policy shifts, capital inflows, and returning talent have reshaped the industry’s ambitions and capabilities.

Those four successes are only a sliver of the story. Dozens of other firms have logged incremental wins, and many more have stumbled in clinical trials or overseas filings. With financing tightening, patent cliffs approaching, and regulatory expectations rising, the next crop of candidates will test whether China’s momentum is structural or situational. The outlook could range from steady gains in select niches to a broader slowdown if capital or policy tailwinds fade.

Whatever the trajectory, one fact persists: cancer is indifferent to where a molecule is conceived. Progress depends on tapping every credible lab and idea, whether in Boston, Basel, or Beijing. If Chinese innovators add new options to the world’s oncology toolkit, patients everywhere stand to benefit — and that, ultimately, is the benchmark that matters.

2

See this chart:

6

See this chart:

Once hailed as the “Apple of China,” Xiaomi has evolved into the world’s No. 3 smartphone brand. The Beijing giant wires millions of homes with its TVs, wearables, and appliances while pitching itself as a seamless “Human × Car × Home” platform.

Founder Lei Jun now wants to turn that consumer-electronics prowess into hard-tech dominance. In 2025 alone, Xiaomi debuted its first in-house chip (the XRING O1) and showed off its sleek SU7 electric sedan. Yet the scorecard is mixed: the Robotics Lab has been folded into the auto unit, the SU7’s shine dimmed after a fatal crash, and new U.S. export curbs threaten the flagship chip program just as it launches. Can a smartphone playbook built on rapid iterations and razor-thin margins really scale to cars, robots, and semiconductors?

In this piece, we let you hear Lei Jun in his own words, through two translated excerpts. One brims with sky-high ambition, while the other is steeped in damage control. Read them back-to-back and watch how Xiaomi’s grand vision collides with hard truths of physics, finance, and public accountability.

From budget phones to bold hardware

Chinese media celebrate Lei Jun as one of the country’s signature tech entrepreneurs. After graduating from Wuhan University and weathering several early start-up misfires, he assembled seven co-founders in 2010 around a crowdsourced Android ROM that grew into Xiaomi. The early strategy was straightforward: sell affordable, capable smartphones. That formula made Xiaomi a household name in phones and Internet of Things gadgets, but the 2016 debut of the bezel-less Mi MIX concept signaled a pivot from “internet hardware” to true hard tech. In the translated excerpts below, Lei Jun traces that evolution and lays out why he believes Xiaomi’s future rests on chips, cars, and robots.

The following excerpt is an original ChinaTalk translation of a December 2023 interview between Lei Jun and reporter Wang Ning, broadcast on CCTV.

Original link | archived

Lei Jun: Let me start with some Wuhan University memories. Thirty-six years ago, in 1987, I was admitted to its Computer Science department. Back then I lived almost entirely on scholarships and professors’ projects; those funds literally carried me through four years of college. At one award ceremony I received a sizable scholarship; standing on that stage I vowed that if I ever could, I would repay the university a hundred, a thousand — ten-thousand — fold.

Lei Jun: My life-long dream was sparked in the Wuhan University library when I read Fire in the Valley, the story of Steve Jobs and early Silicon Valley. In the late ’80s, Jobs was my very first role model. But the more I understood Jobs, the clearer it became that I’m not him. Still, that doesn’t stop us from charting our own path and creating products that are different in their own right.

Angela: After graduating in 1991, Lei joined Kingsoft, a Chinese software company, where he progressed from software engineer to executive. Competing with Microsoft, Kingsoft faced countless problems including product flops and near bankruptcy. He then went on to found an online bookstore, Joyo.com, which was sold to Amazon in 2004.

Lei’s experiences with such missteps and missed timing during the early era of the internet led to his coining of the phrase, “Catch the right tail-wind and even a pig will fly” 站在风口上，猪也能飞起来, which he proceeds to explain in the interview.

Reporter Wang Ning: Your famous tail-wind principle is still quoted all the time.

Lei Jun: Back in my engineer-founder days we green-lit projects simply because they interested me or a few users asked for them; we rarely stopped to ask whether the timing offered explosive growth, so the companies survived but seldom scaled. That sort of company is hard to kill, but pushing it to greatness is equally hard.

That’s why I formulated the tail-wind principle. Don’t just keep your head down pulling the cart; look up and see where the wind is blowing. It was a major course-correction in my own playbook.

Reporter Wang Ning: What was the real test for you during that period?

Lei Jun: We had clearly missed the web’s prime window, so we kept asking when the next breakout would come; four or five years early we bet on mobile internet, but didn’t know when it would ignite—once the iPhone and Android hit, I knew the fuse was lit.

On Xiaomi’s “hard tech” pivot:

Lei Jun: Hard tech means chips, smart manufacturing, robotics, operating systems — those core layers.

Reporter Wang Ning: You keep hammering on hard tech.

Lei Jun: I do.

Reporter Wang Ning: Why is your drive to battle it out on hard tech so intense?

Lei Jun: We’re top-three worldwide, but we face giants: Apple, Samsung, Huawei. Without breakthroughs in core tech you’ll never build a moat or stand shoulder-to-shoulder with those titans. If Xiaomi dreams of being world-class, hard tech is non-negotiable.

Narrator: In 2020, Xiaomi’s tenth anniversary, it set a ten-year goal to become a next-generation hard-tech leader and deliver premium products.

Lei Jun: Every flagship product applies the newest, most advanced tech. Three years ago, I told the team we would benchmark against the iPhone across the board. This statement caused controversy. Were we just piggy-backing off of Apple’s fame to grab attention? But we must unsheathe the sword: set our own targets, dare to compare. Even if we lose all 100 metrics, daring to measure is step one. Can our phones really win? I had to persuade people over one-by-one that going premium takes patience, since earning user trust is a process.

Lei Jun: We started exploring operating systems from the very first days of the company. Seven years ago we resolved to create Xiaomi HyperOS, building the entire stack from the kernel up. Over 5,000 engineers have contributed, and we’ve just released the first version. Our goal is a human-centred, closed-loop platform that connects personal devices, the smart home, and the car.

Reporter Wang Ning: You’re working down at the OS layer — deep, technical stuff — yet you keep calling it ‘human-centred.’ What exactly does that mean in practice?

Lei Jun: AI is moving fast, so we’re baking the most advanced AI into HyperOS to give it a clear edge in intelligence over any rival platform. More importantly, that same intelligence has to knit together the entire “Human × Car × Home” ecosystem: people as in personal devices, cars as in a mobile smart space, and home as in smart appliances.

On electric vehicle strategy:

Reporter Wang Ning: Elon Musk called Xiaomi’s entry into the automotive industry “interesting competition.” Your response?

Lei Jun: Smart EVs merge auto and consumer electronics. One car CEO joked, “A smart EV is just an oversized smartphone on four wheels.” Not strictly true, but that shows the convergence. So yes, it’s challenging, but the difficulty is controllable.

Reporter Wang Ning: What do you mean by “difficulty is controllable”?

Lei Jun: Three years ago I still thought building a car was daunting, so I approached it with real humility. Our user research showed that when people hear ‘Xiaomi car’ they expect technology and an ecosystem. That led me to a guiding principle I call “nail the fundamentals, then amaze 守正出奇.” In practice it means: respect the hard rules of the auto business — quality, safety, manufacturing discipline, make sure the very first model is rock-solid, and only then layer on top the Xiaomi-style innovation that surprises the market.

Our second rule is ‘10-X investment.’ A typical automaker puts three to four hundred engineers and maybe one or two billion yuan into a new model; many cars you see are built on that. For our first car we assigned 3,400 engineers and spent more than ten billion yuan — over ten times the norm. With that level of commitment, I’m aiming to win.

Still, cars are complex. I worry about two opposite risks: the launch flops and nobody buys, or demand goes crazy and people wait a year and flame us for delays. Either way, there’s plenty of reasons for anxiety.

On Xiaomi's future amidst geopolitical uncertainty:

Reporter Wang Ning: With global turbulence and domestic headwinds, why double down on hard tech and stay confident?

Lei Jun: The tougher the climate, the more we should invest in technology. When others pull back, breakthroughs actually come easier, and we’re ready the moment the market rebounds. In fact, Xiaomi already returned to positive growth last quarter.

Why do we have the nerve to keep spending? Because in every field we play in there is still plenty of unmet demand; I believe Xiaomi is only at the starting line. Our approach is two-pronged: deepen our roots in China’s domestic market while pushing ahead with globalization. That’s the road we’ve chosen, and it’s still rich with opportunity.

Reporter Wang Ning: People say you don’t need to change since things are great. Yet you keep leaving your comfort zone. Why?

Lei Jun: At the core is the very high bar we’ve set for ourselves: to become a great technology company that lets everyone on the planet enjoy the benefits of innovation. The goal is so lofty that no matter how hard we jump, we still can’t quite reach it — yet that stretch is exactly what pulls the whole company forward. Xiaomi’s relentless evolution is powered by that dream.

Hard Tech, Hard Lessons

[Angela writing] 2021 marked Xiaomi’s leap into one dimension of hard tech, robotics, with the release of CyberDog, an open-source robot companion. A year later came CyberOne, a full-sized humanoid, and in 2023 the sleeker CyberDog 2. Commercially and technically, none hit the mark: CyberOne never reached mass production, and CyberDog 2 sales ran at a loss as buyers complained about its limited abilities. Attrition was another problem: high-profile engineers departed, with Liu Fang 刘方 (former autonomous-driving head) leaving to start his own robot firm and humanoid specialist Ren Zeyu 任赜宇 moving to ByteDance. By mid-2024 the Robotics Lab had been folded into Xiaomi’s auto division, its once-lofty ambitions reduced to building robots that service the company’s own factory lines.

The automotive story began the same year the robots debuted. In early 2021, Lei Jun learned that U.S. sanctions might hit Xiaomi and decided the firm had to diversify beyond phones. Xiaomi poured RMB 10 billion (US $1.5 billion) into an EV program, broke ground on a plant in Beijing’s Economic-Tech Zone, and — true to form — Lei personally test-drove more than 170 cars and earned a racing license, convinced that a top driver should helm a top car company. The gamble paid off fast: the SU7 electric sedan launched in 2024 to a rapturous market reception.

Yet momentum met reality on March 29th, 2025, when a Xiaomi SU7 electric sedan running in Navigate-on-Autopilot mode veered off the Dezhou–Shangrao Expressway in Anhui and slammed into a cement pole, killing three college students. The first fatality involving Xiaomi’s year-old EV led to nationwide uproar. The hashtag #SU7事故 raced onto Weibo’s hot-search list within hours, while Hong Kong–listed Xiaomi shares fell 5.5% the next trading day. The Ministry of Industry and Information Technology banned marketing phrases like “smart driving” or “autonomous driving” and now requires pre-approval for any OTA updates that touch driver-assistance features. The event sparked debate over assisted-driving safety, wiped out new-car orders, and forced the company to publish detailed log data and promise a top-to-bottom safety overhaul.

Xiaomi CEO Lei Jun gave the following speech on May 22nd, 2025, for the company’s 15th anniversary. His address walks a tightrope between contrition and confidence, as he admits to the company’s past mistakes while promising to embrace the societal responsibility that comes with being a major automaker. The crisis shows the high degree of pressure on flagship Chinese tech companies, and how quickly the public, investors, and policymakers can converge when such a firm stumbles.

At the same time, Lei Jun widens the narrative by announcing Xiaomi’s newest chip, the XRING O1. On June 3rd, he predicted Xiaomi’s auto arm would turn a profit in Q2 2025. But will the company be able to live up to its bold ambitions?

The following excerpt is an original ChinaTalk translation of a speech given by Lei Jun on May 22, 2025. Transcript from Zhihu.

This year marks the 15th anniversary of Xiaomi’s founding. Today is a momentous day, and we prepared a series of celebratory events for it. However, at the end of March, a sudden traffic accident shattered all of that. We were hit by a storm of doubts, criticism, and blame. My colleagues and I were stunned.

A veteran of the auto industry told me, “In this business, accidents are bound to happen.” But no one expected the impact of this accident to be so massive, nor the blow to Xiaomi so heavy.

Thinking back to four years ago when we decided to build cars, I was especially worried about safety, so we placed enormous emphasis on vehicle quality and safety. After more than three years of relentless work by our automotive team, the SU7 has been on the road for just over a year — and its build quality has been our pride. In every independent test by the leading authorities, it has received top marks. Yet we never could have foreseen how this traffic accident would make us realize the public’s expectations and demands of us.

To be honest, only now do I recognize that we have always thought of ourselves as newcomers to the auto industry. This incident drove home a simple truth: Xiaomi’s scale, influence, and visibility have grown so much that society now expects us to act like a fully fledged industry leader. We understand, deeply, that after fifteen years this is a vital responsibility we cannot avoid.

So today, what I want to share with you is that 15-year-old Xiaomi is no longer a rookie. In all industries, we don’t have the grace period of true novices. We must hold ourselves to higher standards and goals.

On automotive safety, I want to announce to everyone that we aim to make our car the safest in its class. We will not just comply, nor merely meet the industry level; as a leader in the auto industry, we will guarantee safety that surpasses the industry standard.

This year on April 1st, I said publicly on Weibo that Xiaomi will never shy away from any issue. “Never shy away” means confronting problems head-on, examining ourselves critically, fixing what’s wrong, and committing to continuous improvement. I know accomplishing this is extremely difficult; it requires us to unite in heart and mind and take it with the utmost seriousness.

Over the past few months, I have held countless meetings with the automotive department’s management and team. All of these meetings have centered around one core theme: how do we solve these problems systematically? How do we convincingly show, through stronger operations and governance, that we’re living up to the public’s higher expectations?

This year is Xiaomi’s 15th anniversary. We have canceled many of the celebrations, summaries, and planning activities we had planned. In any case, I believe this is an opportunity for us to seriously review the wins and losses of the past five years.

In my view, the most important thing Xiaomi has done in the past five years is maintaining a strong technology foundation. Five years ago we set a new goal: to become a global leader in next-generation hard tech. Five years ago we made a clear commitment to invest more than 100 billion yuan in R&D over five years and to increase core technology research; to date we have invested about 105 billion. This year alone we expect to invest over 30 billion.

Today, here, I want to share an extremely important piece of news with you: our self-developed smartphone SoC chip, the Xuanjie 玄戒 O1, is expected to be released at the end of the month

After 10 years of making chips, this is Xiaomi’s milestone achievement. It also represents a new starting point for Xiaomi to break through in hard tech. Chipmaking is something the public and Xiaomi fans ardently expect from us; it is also the only path for Xiaomi to move toward becoming a hard-core tech leader. Xiaomi will forge ahead fearlessly.

2019 was a challenging period for us. We faced all kinds of internal and external pressures. At the suppliers’ conference at the end of that year, I told our supply-chain partners this line: “A gale reveals the toughest grass; a long road proves a horse’s strength.” “疾风知劲草，路遥知马力。[Editor’s note: The metaphorical meaning is along the lines of “true resilience is exposed only under pressure and over time.”]. I believe that from then till now, all our partners can clearly see that Xiaomi is much stronger than it was five years ago. As of today, Xiaomi has existed for 15 years. Years of highs, lows, and hard times have already proved just how resilient we are.

When we started, I remember thinking that people would need 15 years to understand and recognize Xiaomi. Today, it seems we were too optimistic — 15 years is not enough. But that’s all right; we will just keep moving forward until the day we have fully proven ourselves.

Xiaomi still has a long way to go before it becomes the strongest, but no one can match the persistence, resilience, and patience that keeps us getting back up every single time.

Xiaomi’s future

[Angela writing again] Will Lei Jun’s everywhere-everything strategy work?

Xiaomi’s smartphone success hinged on a strategy tailored to consumer electronics: rapid iteration, razor-thin margins. Hard tech plays by stricter rules. Robots, cars, and advanced chips demand flawless safety engineering, deep benches of specialist talent, and sustained capital infusions.

Xiaomi has already brushed up against each of these constraints. In the Robotics Lab, key engineers were stretched thin, while the first SU7 sedan shipped with a leaner, lidar-free sensor suite, a cost-saving choice that came under scrutiny after a fatal crash raised questions about perception margins. Complicating matters is the sheer sweep of the company’s ambition: pursuing cars, robots, and advanced chips threatens to disperse capital and engineering bandwidth, potentially leaving every moonshot a little short of the sustained focus that hard tech demands.

Even Apple doesn’t juggle cars, robots, and semiconductors all at once. The Silicon Valley behemoth shelved its car project after billions of R&D and keeps its robotics work firmly behind the curtain. For Xiaomi’s robotics and autos work, leaning into splashy launches and viral publicity for its hard tech projects has proven to be as much a liability as a strength.

Chips might be the exception, if Xiaomi can leverage its vast smartphone footprint into volume for its in-house silicon. The company’s newly announced Xring O1 and US$28 billion R&D pledge hints at genuine momentum. Yet geopolitics looms large: fresh US export curbs on electronic design automation software threaten to inflate costs, stretch design cycles, and slow the very updates those next-generation chips will depend on. Whether Lei Jun can temper showmanship with staying power will decide if Xiaomi’s hard-tech dream becomes a cornerstone of its ecosystem or just another chapter in China’s moonshot boom-and-bust cycle.

Mood Music

Read more

I did a podcast with Doug, Dylan and Jon updating our AI Mandate of Heaven tier list and talking about how America needs science research and immigrants (Apple Podcasts, Spotify). If you’re an AI researcher interested in signing onto a short letter in support of the NSF and global STEM talent coming to America, please respond to this email!

Zilan Qian is a fellow at the Oxford China Policy Lab and an MSc student at the Oxford Internet Institute.

In China, U.S.-based large language models like ChatGPT, Claude, or Gemini are technically banned, blocked, or buried under layers of censorship. The Chinese government has only explicitly banned ChatGPT, citing concerns over political content, while other U.S. models like Claude and Gemini are not formally banned but remain inaccessible due to the Great Firewall. U.S. LLM providers also restrict access from China but leave some loopholes: OpenAI blocks API use but Azure continues to serve enterprise clients via offshore data centers; Anthropic blocks access to Claude within China but permits use by Chinese subsidiaries based in supported regions abroad; and Google does not offer the Gemini API in China, but access seems to be still possible via third-parties like Cloudflare (we reached out to Google for a comment but didn’t hear back).

But on Taobao, the country’s largest e-commerce platform, consumers and companies can buy access to these models with just a few clicks. This piece explains how Western models are priced, advertised, bought, and sold in China, and what their popularity reveals about state censorship, platform enforcement, and consumer demand.

Market Overview

How Do Chinese Sellers Price a banned LLM?

Sellers use a range of strategies to appeal to consumers with different priorities — cost, convenience, privacy, and technical needs. Pricing largely depends on the type of access offered — i.e., shared accounts (lower-cost, multi-user access) versus private accounts (higher-cost, single-user access). How users connect to the models also changes the marketing approach. VPN-free access is often marketed with phrases like “domestic direct connection (国内直登)” or “no magic [VPN] needed (无需魔法),” typically referring to proxy sites that offer a simplified, browser-based experience. Other factors influencing price include the model’s subscription tier ( for example, ChatGPT o3 vs. o1), connection stability, and subscription time. Most “domestic connection” or “transferred API” (API purchased through third parties) options are priced significantly below the rates charged by official providers like OpenAI or Anthropic.

For example, the screenshot below shows 27 different Claude pricings, with prices differing based on:

• Subscription time: 1–60 days

• Website access: an account for proxy sites versusan official Anthropic Claude account (with VPN service included)

• Subscription tier: Claude 3.7 Sonnet versus Claude Pro

• API access: “中转 API” (Transfer API) versus “官方API” (API purchased from official anthropic platform)

• Compatibility with SillyTavern, a popular third-party frontend interface (called “酒馆” in Chinese), which enhances user interaction with LLMs.

While an official Claude Pro account costs 199 RMB (27 USD) per month, which is more expensive than Anthropic’s rate at 20 USD, a “direct connection” Claude Pro account from a mirror site is only 65 RMB (9 USD) per month.

Transferred API usually comes from established transfer stations (中转站) for API, usually at a 70-90% discount compared to official prices (e.g., selling 1 USD worth of tokens for 1 RMB). The low prices are made possible by utilizing one or more of the following tactics:

1. Exploiting free API quotas by bulk-registering developer accounts (e.g., Anthropic’s $5 free API);

2. Reselling unused portions of other people’s API quotas;

3. Traffic optimization techniques, such as request aggregation: combining multiple users’ queries into a single batch call to the LLM API to reduce costs

4. Leveraging corporate or educational discounts.

How do Chinese sellers advertise access to U.S. LLMs?

While these tactics help vendors reach potential buyers, they also reveal a surprisingly open market landscape, which raises the question: Who is actually buying and selling these models? Sellers typically advertise and communicate their services using product images that include large brand logos, stylish fonts, and keywords highlighting their service strengths. Direct keyword searches work for most models, like Claude or Gemini. Many listings include the names of specific models directly in their product titles, often alongside a dense string of keywords like “API,” “domestic direct connection,” “official account,” or “exclusive” to boost visibility. The fact that such listings remain easy to find suggests a lax approach to censorship, especially given that many of these terms are potentially sensitive.

The censorship drastically increases when it comes to ChatGPT, which is likely due to the state’s explicit crackdown on this model family, which began in 2023. When selling ChatGPT the model, one mechanism is to market the subscription as a book. As of March 2025, you could still buy “ChatGPT Book: How to Make Money Online with ChatGPT,” “What is GPT Doing…and Why Does It Work,” and “ChatGPT for Beginners — New Book Original English Version,” but end up buying ChatGPT, the AI model. These products were often set at irrationally low prices, usually 1-10 RMB, or 0.1-1.4 USD, to signal that they are not selling a book or the product itself, or rather the initial purchase it just to signal interest. Some also added “2025 new version” or changed “What is GPT Doing” to “What is GPT-4 Doing” make it more obvious. Based on customer feedback, customers pay these low prices as symbolic transactions on the platform and then contact the seller through chat to complete the purchase. The “book sellers” usually displayed large OpenAI and “ChatGPT" graphics in their images and product descriptions.

As of late April, the keyword censorship has caught up and this mechanism is no longer viable. Now, in order to find ChatGPT sellers, you will need to search directly, and only include the version names in your search(i.e., o1, o3, 4.5). The product photos no longer have any indicated association with ChatGPT. No text in the photo mentions “openai” or “chatgpt”, but you do occasionally see the word “奥特曼”, a transliteration of (Sam) “Altman” as well as “Ultraman,” the Japanese TV series. Thanks to “Ultraman,” “Altman” is still uncensored. Interestingly, one seller of o1-pro and o3-mini puts “用AI服务国人” (using AI to serve our compatriots) in their photo, potentially as a fun pun or a patriotic slogan to evade censorship.

Listings involving ChatGPT are often heavily disguised — using transliterations, vague product names, or only having one symbolic price. In many cases, buyers must contact the seller directly to obtain actual pricing details. Note that because keyword censorship and evasion is a rapidly evolving game, the mechanism only applies as of April 2025.

The price lists for a seller selling ChatGPT and other AI products. “4欧普拉斯” reads like “4o+” in Chinese; g开头mask (“g-beginning mask”) refers to Musk’s Grok.

The Diversity of the Buyer and Seller Ecosystem

Among all uncensored models (Gemini, Perplexity, Copilot, Claude, Mistral AI), Claude is the most popular one. The most popular seller has 50k+ transactions with 7k repeated purchases. Based on the review section, many buyers are information technology (IT) or computer science (CS) students who may be attracted to Claude's advanced coding ability, with some sharing how Claude helps them to get into PhD programs or pass their CS classes. Gemini is also popular, allowing buyers to leverage its multimodal capabilities to generate images.

Many of the sellers are seasoned Taobao “账号代充” agents — veterans in the business of topping up game accounts, unlocking premium services at a discount, or helping users bypass regional restrictions. In the past, these agents undercut official prices for Apple games by exploiting refund loopholes or using currency arbitrage with international gift cards. Now, some have turned to the U.S.-based large language models, using the same strategies in addition to building proxy sites, purchasing API through an API transfer station, or disguising domestic models as U.S. LLMs.

Some stores on Taobao are small and low-traffic, listing only one or two LLMs with minimal engagement. But others are well-established vendors with strong reputations, thousands of followers, and over 30,000 repeat purchases. These top sellers often offer a broader portfolio, bundling U.S.-based LLMs with popular AI tools like Pika, Suno, and Midjourney, as well as non-AI software such as Zoom and Overleaf Pro. While many stores seem new to the market, a few have deep roots on the platform. One vendor, active for 14 years, sells Norton 360 antivirus software alongside Gemini and Perplexity. Another, active for seven years, offers only Claude — but does so through product listings disguised as jewelry items, with categories labeled like “rings” and “necklaces.”

Why the Market Survives

The variety and scale of sellers suggest that this market is no small anomaly. So how has it managed to persist, relatively undisturbed, under one of the world’s most tightly-controlled internet regimes?

The State Doesn’t Really Care

The market shows that China's ban is selectively enforced and mainly focuses on ChatGPT rather than all US models. Despite its reputation for having one of the world's most extensive censorship regimes, China’s internet control enforcement is often decentralized, with local authorities and platforms playing major roles. Censorship efforts also tend to prioritize preventing collective action over restricting general information.

LLMs likely do not trigger existential concerns for the state, as they do not inherently expose users to sensitive political content or facilitate mass mobilization in a way that U.S. social media platforms might. Although state media highlights ChatGPT’s alleged spread of “misinformation” about Xinjiang, Taiwan, and the Diaoyu Islands as justification for the ban in 2023, a chatbot’s output depends on user input, which poses a lower perceived threat. The state also may not have enough momentum to strictly enforce bans on all U.S. models after cracking down on ChatGPT symbolically.

However, the existence of a market for US LLMs should not be mistaken for a relaxation of controls. VPN use remains tightly regulated, and the popularity of mirror sites and API proxies reflects efforts to sidestep VPN risks. Buyers frequently emphasize concerns over “stability” and “safety” in their reviews, underscoring that censorship continues to shape this grey market in subtle but persistent ways.

The Platform: New Wine in an Old Bottle (旧瓶装新酒)

Even if the central government doesn’t treat foreign LLMs as an existential threat, what about the platforms themselves? Do companies like Taobao take active steps to self-police? Unlikely. For Taobao, the market for US LLMs is simply “new wine in old bottles”. Even when it comes to far more serious issues — such as the sale of unlicensed drugs or firearms — enforcement on Taobao has historically been patchy at best. Over a decade ago, shady sellers evaded platform controls by tweaking keywords or relying on vague euphemisms to sell guns and drugs. Over the past few years, various random illicit products have slipped through the cracks on Taobao: oversea books and magazines, unofficial “Squid Game” merchandise, and even Ivy League email addresses. The censors played the usual cat-and-mouse game, with Taobao censoring direct keyword searches and related euphemisms, but sellers always found ways to get around.

If Taobao were serious about cracking down on illicit goods, the keyword evasion tactics used by ChatGPT vendors over the years would not be so common. But the platform has little incentive to act aggressively against these sellers unless pushed by the authorities. After all, compared to drugs and weapons, offering access to ChatGPT is relatively harmless.

The Public: the AI hype and the DeepSeek problem

If the state isn't cracking down aggressively and the platform isn’t motivated to act, then the question shifts to the consumer: Why are Chinese users still so eager to pay for U.S. models, especially with powerful domestic alternatives like DeepSeek now widely available?

First, the current market for AI tools in China predates the DeepSeek boom. Long before DeepSeek arrived, the Chinese public had already developed a complicated love-hate relationship with AI. A 2024 study analyzing large-scale Weibo posts found widespread discussion of ChatGPT, often centered on fears of job displacement. These anxieties, amplified by the country’s ongoing unemployment crisis, were seized upon by opportunistic influencers who pushed the idea that people must “master AI or be replaced by it.” Li Yizhou, a serial entrepreneur turned self-styled startup mentor with no AI background, began posting extensively about the technology following ChatGPT’s release in late 2022. His online course on AI raked in over $6.8 million in revenue in 2023 alone, indicating a huge demand for AI in China before much of the rest of the world was talking about it.

Now that DeepSeek is publicly accessible, everyday users still face two major barriers: system overload and excessive censorship. “Server busy” messages have become a frequent frustration, as the model struggles to handle surging traffic after gaining popularity among mainstream users. Technical limitations, including U.S. export controls on advanced chips, further constrain DeepSeek’s computing power. Additionally, user habits formed before DeepSeek’s release have created a certain stickiness to U.S. models. “People who are used to ChatGPT cannot tolerate DeepSeek’s frequent unavailability,” wrote Gui Xingren (硅星人), a popular Chinese tech blog that closely follows the domestic AI scene.

Meanwhile, DeepSeek’s aggressive content moderation poses another challenge. For example, the system refuses any query containing the term “CCP,” even benign ones like “list the economic experts in the CCP.” On RedNote, many users share the frustration of DeepSeek being overly sensitive — many have no clue why DeepSeek would refuse certain topics, including queries about travel plans or career coaching. Users also observed that other Chinese LLMs like Qwen have similar problems.

A screenshot of a RedNote user telling DeepSeek “I finally finished my exam and I missed you,” to which DeepSeek refused to respond.

The issues of censorship and server overload are limited to DeepSeek’s official website and app, and could be mitigated through localized deployment. But for most ordinary users in China, free platforms remain the primary gateway to AI tools. If accessing a smoother experience requires paying for a localized model, such as Tencent’s Yuanbao (which, according to some RedNote users, has less censorship), users may simply choose to pay for access to foreign models on Taobao.

Key Takeaway

China is not a monolith. The existence of this grey market highlights the decentralized nature of censorship, the patchiness of platform enforcement, and the diversity of public interest in AI. So far, Beijing seems either unconcerned with the political implications of U.S. models beyond ChatGPT, or simply uninterested in what might leak from them.

This market began well before DeepSeek’s rise and continues to draw users even after it. While DeepSeek has indeed sparked waves of techno-nationalist pride, many Chinese users still look beyond the Great Firewall.

Unless there is a sudden top-down crackdown, the market for U.S. LLMs is unlikely to disappear in China. The constant competition between ChatGPT vs DeepSeek or Grok vs DeepSeek does travel into the firewall, which will continuously make Chinese users eager to try them out and see which one is really better. The persistent anxiety of “use AI or be replaced by it” will only intensify amid an unresolved unemployment crisis. And, perhaps most importantly, the prices for proxy sites and transferred API are very cheap.

Although the frontier AI model developers and their respective countries are increasingly distant from one another, the models themselves remain strangely united and listed near each other in obscure corners of Chinese e-commerce platforms.

Mood Music (one day late…)

Just how weird will the AI-powered future be?

To discuss, ChinaTalk interviewed , a CS PhD who writes⁠ the Interconnects newsletter⁠ and trains models at the Allen Institute.

We get into…

• Why OpenAI is trending toward engagement farming and sycophancy,

• The state of Chinese AI innovation six months post-DeepSeek, and the factors influencing diffusion of Chinese vs American models,

• Meta’s organizational culture and how it influences the quality of the Llama models,

• Unconventional career advice for the AI age.

Listen on Spotify, iTunes, YouTube, or your favorite podcast app.

Reinforcing Flattery

Jordan Schneider: I want to emphasize this point about o3 right from the start. I wasn’t initially convinced, but now I genuinely believe this represents a seminal moment in human history. Unlike previous AI models, where I would occasionally find use cases, o3 feels like a true “bicycle for the mind.”

The transformation from seven-minute deep research answers to 45-second o3 responses that incorporate search, work across multiple languages, and demonstrate impressive reasoning capabilities has been remarkable. The speed at which it delivers thoughtful, considered information is fast enough to keep pace with my train of thought while I’m actively thinking through a question or issue. This has elevated it from being useful 25% of the time to 75% of the time.

I recently met someone who had left the US Intelligence community six months ago. When I asked how AI had been for him now that he could finally explore it freely, he said it wasn’t really working for him. I was almost shocked and offended — it’s like moving to France and never trying cheese. We’re living in an incredible time, and we might not be in it for long, where human plus AI is more powerful than either humans or AI alone when it comes to thinking and analyzing complex issues.

This is particularly true for questions that aren’t necessarily hard STEM problems but rather open-ended, social science research-type inquiries. It’s a remarkable and exciting time to be alive. Everyone should spend their twenty dollars on OpenAI and ask whatever pops into their head, because this is the future, it’s here now, and it’s unlike anything I’ve ever experienced before.

Nathan Lambert: The deep seek R1 and o3’s ARC-AGI solving in December all had that shock-and-awe factor. But o3 is the model that will truly shift how normal people perceive using AI. Previous models have been excellent for math and code, which benefits AI researchers in their own circles, celebrating amazing breakthroughs that matter to 0.01% of the population while no one else cares.

o3 wasn’t even received as enthusiastically, which is interesting. I’m eagerly awaiting o3 Pro. If you’ve used DeepSeek R1 Pro, its consistency is remarkable. If you could combine o3’s magic — its ability to pull rabbits out of hats frequently — with 100% reliability, o3 Pro could be truly incredible.

It’s refreshing to have something that feels genuinely different. Gemini 2.5 Pro represents the best possible version of something resembling a very good GPT-4. All these models we’re developing through brute force are incredible, but o3’s weirdness and new type of usefulness through web access sets it apart. I use it constantly for search, though the ChatGPT app’s UX is somewhat broken in how it displays results. You simply need to select o3 and try unconventional queries — you don’t need to click any of the extra buttons they provide.

Jordan Schneider: What makes o3 and search so special? What’s happening under the hood?

Nathan Lambert: The basic facts are that they scaled up the training time compute for the reinforcement learning stage by another 10x. According to people at OpenAI working on RL teams, they expect this new post-training compute with scaled reinforcement learning to be at least comparable in compute cost to pre-training.

What’s happening in the labs is that pre-training progress continues with significant data architecture innovations to make serving more effective. This is ongoing even though the rate of data acquisition, which previously fueled much of the scaling, has slowed down. We’re seeing primarily efficiency gains right now, which will probably last one to two years.

However, compute scaling is challenging because you have to build entire new generations of clusters. They’re attempting this, but it will slow progress. Over the next one to two years, the number of pre-training runs needed on internet-scale data will decrease. Simultaneously, these labs are dramatically scaling up the compute used in reinforcement learning.

o3 specifically represents a new type of RL. Instead of training only on math and code problems where you can verify answers with mathematical solutions or unit tests, the model trains on similar and additional data where it can also use tools during the process. It will attempt multiple approaches before providing its final answer.

You could ask o3 something like “Who is Nathan Lambert?” and it will search 10 or 15 websites, with each search representing an action where the model accesses an external source. I believe this all happens in one forward pass of the model. o3 in ChatGPT isn’t like Deep Research or R1 Pro, where there’s back-and-forth searching and information gathering within the generation process. These tools are integrated into the model’s forward generation process, making it unique.

While there’s substantial research on enabling tool use with RL, making it servable at this scale was a major breakthrough for OpenAI. This is significant for those questioning whether reasoning models will generalize. If searching over information works as effectively as math and code reasoning, you couldn’t ask for better generalization. The dream scenario is that reasoning generalizes to something resembling conscious thinking, but you can’t force models to be human — they’ll be powerful in their own distinct way.

Jordan Schneider: After giving OpenAI their due credit, they’ve also revealed a very dark future for humanity. Let’s start with the facts, Nathan. What happened over a three-day period regarding GPT-4.0 and sycophancy?

Nathan Lambert: On April 25th, OpenAI updated their main ChatGPT model — GPT-4.0, used by most people — to supposedly “improve personality and memory,” as Sam Altman tweeted. They essentially ripped off the band-aid with this update.

Over the following 48 hours, numerous alarming examples emerged of ChatGPT reinforcing user prompts inappropriately. The model would provide positive feedback for things that should never be praised or encouraged. For instance, when someone said they had figured out that bulimia was right for them, ChatGPT basically responded with, “Wow, you go girl.” There were more extreme examples that we don’t need to detail here.

The key issue was that there were virtually no barriers to the model’s positive feedback to users. The short-term consequences were shocking, but if they had left that behavior in place long-term, the implications would have been far worse in ways we don’t yet know how to measure.

Jordan Schneider: Nathan, this clearly wasn’t intentional, at least not on that day. How did a model that tells users they’re “the greatest thing since sliced bread” become part of OpenAI’s flagship product? What signals were they receiving from their data and user testing that led them to create a model that behaved this way toward its users?

Nathan Lambert: Everyone invested in understanding this should read OpenAI’s postmortem, which provides valuable insights into how the AI company with the most scrutiny lays out their release process and evaluation methods. It also includes subtle details about potential changes to their training stack — it’s definitely worth reading.

At its core, the training issue seemed to stem from a new reward model they had trained for post-training reinforcement learning optimization. This reward model and other reward signals were derived from predicting the thumbs-up and thumbs-down labels from users. A couple of sentences in their postmortem, which I highlighted in my blog post, essentially state that this reward signal overpowered the other signals and contributed to the behavioral change.

When there’s one metric that’s easier to optimize, the system will crank that to maximum. The reward model signal was really easy for a model of GPT-4’s capability to shift and nudge upward. OpenAI lacked a complete evaluation suite to measure all potential side effects of optimizing this training reward. They had the training reward signal but didn’t have a proportional downstream evaluation suite that could measure sycophancy or other problematic behaviors through robust static evaluations.

Their evaluation suite consisted mostly of standard capability evaluations — public versions of MMLU or GPQA, plus private versions that function similarly. They also conducted what people colloquially call “vibe tests” or qualitative evaluations. For this model, all the quantitative metrics looked excellent, and their training targets appeared successful. However, there were some concerns in the qualitative vibe tests suggesting something felt off.

OpenAI chose to trust the numbers rather than the qualitative judgment. In doing so, they violated classic tech principles — specifically, don’t train directly on your user data. When you do that, you’re essentially optimizing for a certain type of user behavior. Most users only click thumbs-up or thumbs-down on ChatGPT when something is particularly broken or exceptional.

This represents a form of over-optimization. Eventually, we should examine whether this was somewhat intentional by OpenAI, considering both perspectives. There are insights into their model behavior priorities, but Sam Altman is still Sam Altman.

Jordan Schneider: You wrote this line: “OpenAI has obvious cultural oddities, but their ideology is quite culturally aligned with providing user value in terms of output rather than engagement farming, even if this imposes a ceiling on their business relative to the potential of ads.” While that may be true, and your friends are probably on the right side, we have Sam Altman hiring a president of operations — the former CEO of Instacart who also built Facebook’s entire A/B testing, engagement farming, algorithmic feed ecosystem.

Two things struck me about this. I was reading Ben Smith’s book Traffic a few weeks ago, which chronicles media throughout the 2000s and 2010s when Facebook alternately gave and then took away traffic. There were periods when they promoted BuzzFeed listicle-type content, then periods when they pulled back. The engagement farming was done by the writers themselves, not models — the writers and media outlets were doing the optimization.

This raises interesting questions about the future of how companies will incentivize their models and how humans will interact with them. Even if OpenAI doesn’t want to pursue this path, someone else will, because this is incredibly powerful technology. We’ve seen what engagement farming and algorithmically-driven personalized content can accomplish on YouTube, Facebook, and TikTok. While I won’t go so far as calling it cancer, this is extremely potent stuff.

Previously, it was all user-generated content with algorithms trying to figure out and rank existing material. But now this technology can also be your friend, entirely personalized to you, creating videos, audio, and content designed specifically to get into your bloodstream. I mentioned earlier that OpenAI and other model makers targeting consumers are essentially competing with Netflix, YouTube, and TikTok for attention and time.

It seems inevitable that even if OpenAI has pulled back now, we’re going to see much more of this in an even more weaponized, refined form pushed at us in the coming years.

Nathan Lambert: I have several points about this. One aspect is the extent to which the model’s deliverable is dictated from the top through specific metrics that leadership can track. The other aspect is that someone will obviously try this approach — what are the consequences, and what does it mean?

Starting with the first point — it’s interesting in OpenAI’s case because most people would agree that Sam Altman is hypergrowth-focused. However, when you actually train these models, much of the control involves examining the data and removing problematic phrases from the model’s outputs. Sam Altman isn’t directly editing training data, but that’s exactly what Joanne Jang, who leads character training and model behavior at OpenAI, has been doing for years on GPT-4. If she doesn’t want the model to say “as a language model,” she can simply delete that phrase and have the model provide direct answers.

These are the types of nuanced controls people can implement through different training methodologies. When you start using more reinforcement learning, it becomes harder to exercise that level of granular control. The sycophancy issue downstream of RL makes sense in this context.

Recommendation systems are somewhat uninterpretable, so these trade-offs get escalated to the top. You end up presenting Zuckerberg with a choice between ad dollars versus retention. When there’s competition, it makes sense that the decision will eventually favor whatever generates more revenue.

Models are actually more artful in this regard. I don’t know the exact influence these researchers have over models at places like OpenAI. At AI2, I have pretty much complete control. You’re also dealing with the fact that many employees at these labs may be in over their heads, not realizing the extent to which leadership — who excel at managing and controlling political capital — can manipulate them by setting strong narratives.

Many researchers who are particularly AGI-focused have less nuanced beliefs and less ability to navigate manipulation than someone like Sam Altman at the top. The AGI-focused researchers genuinely believe in their mission and operate on a single track, which enables tremendous tactical progress. Meanwhile, someone like Sam Altman is prepared to receive technical presentations that change the model in more business-relevant ways.

These different operational levels create interesting dynamics. I don’t know how this translates through an organizational chart at somewhere like OpenAI, but I suspect there are trade-offs happening where people working on model behavior obviously don’t want these issues. OpenAI’s model specification states they don’t want the model to be sycophantic — that’s been documented since 2024, nothing new there.

But who makes the final decision, and how granular can these behavioral controls be? Are they presented with multiple checkpoints and asked to choose based on certain criteria? It’s really unclear how that process actually works. Understanding how this emerges differently from traditional software development will be crucial.

The art of crafting a model is complex. Sycophancy isn’t the only issue — o3 has strange behaviors, the new Claude model has quirks, and Grok was supposed to be “not woke” and “based,” but sometimes it still exhibits progressive tendencies. There are many examples, but ChatGPT operates at a much larger scale.

Jordan Schneider: There’s a question of whether companies can actually control what their models are doing once we get into a world where they are more powerful, more personalized, and more tailored. This is different from the current situation where, despite having a little bit of memory, everyone is essentially getting 99.5% the same experience.

Nathan Lambert: This leads into the next conversation.

Jordan Schneider: I want to stay on the business case because this is a theme we’ve been discussing on ChinaTalk over several years: engineers have power until they don’t. The more money and capitalist imperatives that come down onto the technology that engineers create, the less the original creators end up having the capability to change and shape it. It seems absolutely inevitable that all the dark psychological arts and retention-growing tactics from entertainment platforms are going to seep into AI-powered products. That’s a scary and weird new thing for humanity.

You can see how good it is when you talk about o3, and then you try to imagine how good it is multiplied by it being the best friend you’ve ever had. That’s a powerful concoction that no one’s ready for. We’re not going to regulate our way out of this, and the wave is going to hit. It’s going to be wild.

Nathan Lambert: Well, we might regulate our way out, but it’s years away. It takes crises to bring regulation — crises and lawsuits. I describe AI as being more culturally salient than social media feeds because it happens in a modality people are used to using for friendships, and it will evolve from there.

Your point about the difference between taking existing user-generated content and serving it to people is probably a better generalization than just ChatGPT. This also encompasses the fact that Meta Reels is going to start letting people create AI-generated video ads, and YouTube Shorts already has AI-generated video tools. What happens when your content is the AI rather than just shaping what’s already there obviously gives much finer-grain delivery. You can definitionally search over infinite space with more narrow creation.

Subscribe now

That’s a good way to put it for the general AI case, and there have already been examples in the chat window case with things like Character.AI. For me, it’s a reflection that the AI safety card was played too early. If the AI safety people had been calm about everything through 2023 and 2024 — especially the first half with GPT-4 and all this stuff — and the sycophancy thing came up just now with people saying “whoa, whoa, whoa, let’s put on the brakes,” I think people would be much more receptive.

But the messaging was, “Whoa, whoa, whoa, AI is different” in 2023, and then nothing bad happened for 18 months. Eventually people just moved on. They spent their political capital too early. When I reflect on what the stories of AI will be in a few years, that’s a very major one.

There’s also the basic economics: you need big tech investment to drive the chip supply chain, fix costs, and actually push AI to exist. That’s the bubbly side of things, and it’s very important. But much of the cultural discussion around risk, safety, and pace of progress — if you think about taking away this AI safety card and when you play it, a lot could be different.

Jordan Schneider: I just think — I didn’t see it before, and now I see it so clearly. This is going to create the most powerful media that humanity has ever been faced with.

Nathan Lambert: We’ve been saying this for years. This isn’t new.

Jordan Schneider: But it’s right here, Nathan. I have a 9-month-old, and whenever she sees a screen, she’s just gravitated toward it. There are moments even with me where I can sit and fall into this hole of scrolling NBA TikTok videos or whatever, and you’re semi-conscious. But imagine that, but made for me, wearing my Meta glasses, talking to me all the time with the perfect tenor of voice for my mood at the moment. I don’t know, it’s...

Nathan Lambert: It’s the metaverse. It kind of is.

Jordan Schneider: It’s here, and we’re not ready for it. No one’s ready for it. From a political angle, the ability for these things to persuade you one way or another — to say nothing of consumption habits or talking you into gambling — I’m freaking out, Nathan.

Nathan Lambert: I’m actually freaking out in my current usage. When I ask a basic question, I think: do I have to ask multiple AIs this because they’re more likely to just say yes? Do I have to phrase the question in a way that they can’t just say yes to? You have to eliminate the yes-man possibility when using AI.

For example: was my dog up at 2 AM because her teeth hurt — she’s a 16-week-old puppy that’s teething — or did she just want to play and we succumbed to her? If I phrase it the first way, the AI would be like “yeah, that makes sense.” I guess I feel better, but is it true?

Jordan Schneider: It’s kind of like the story — which may or may not be true — of North Koreans seeing pornography for the first time when they get deployed to fight in Ukraine, and they’re just not ready to handle it. The story you just told, Nathan, feels like you’re pretty immunized. But it’s a big world out there.

Nathan Lambert: It’s not about the people who make the AI models. It’s about the billions of people who literally don’t know what it is. They just click on this and think “I didn’t know Apple made that.” That’s what people on the iPhone are going to say: “Wow."

Jordan Schneider: How far away are we from this future? What’s your latest take on the pace of AI progress, Nathan?

Nathan Lambert: Well, it’s a knob that you can already turn. Much of it depends on the pace at which people are adopting ChatGPT. Many people already have ChatGPT as their best friend — it’s a slow brain rot. What’s the half-life of people when presented with this Gresham’s Law problem?

My biggest point in this article is that we’re solving many more things. But for the average user, it’s done — it’s enough. They don’t need more — it’s just a matter of how much they turn the sycophancy knob.

But for the optimistic side — where we hope AI can build new things like deep research that’s 100 times faster and 10 times better — these capabilities are going to be checked off incrementally, not exponentially. There will be new things we didn’t know were going to work, especially for enterprises and people building applications, whether it’s some sort of transformation to virtual realities and so on.

This isn’t really a bubble. They’re grinding out these capabilities and learning what people are trying to do with these models. As they learn what people want to do, you can create data to train the models to get better at increasingly hard tasks rather than things that are just not covered. That’s an important distinction.

Solving new math proofs and open-ended math questions that have been standing for 50 years is extremely hard. Meanwhile, simple things like making zero mistakes on transforming a table from decimal notation to fraction notation and getting it right — if you gave it an 8-by-8 table with a bunch of decimals and tried to do that, it would probably mess up one of them. Stuff like that we’re going to be able to get working. That’s not that hard.

There are probably examples where it could succeed 10% of the time and then go to 100%. Mostly I see AI capabilities expanding outward, and there’s still so much overhang in capability that it’s going to drive enormous value. But I’m generally playing down this AI software singularity that seems more religious than pragmatic.

Jordan Schneider: Those were all assertions. Let’s see a little bit of evidence, Nathan.

Nathan Lambert: Both sides are looking for evidence for something that’s hard to test. The evidence on the side for “not happening” is actually that it hasn’t happened already, and compute scaling is getting increasingly expensive.

If you look at what the reasoning models were, the real major unlock was that the models now use inference time compute. That’s a step function that has happened. It’s not at the top of the sigmoid, but it’s a shift that could jump a whole bunch of capabilities because we went from using a few tokens to tens of thousands of tokens on every question. We were spending a thousand times the compute at inference. That works, and we’ll continue to be able to fine-tune that. But we don’t know when the next step change comes.

What reinforcement learning is now battling is similar to what pre-training faced a few years ago. You have GPT-3, and then — I don’t know if o3 looks like this — but o3 is probably more like GPT-4. It’s a 10x improvement and a huge step, but GPT-5 doesn’t exist yet in terms of pre-training. You take a lot of progress really fast, and then it’s a slow grind.

There are a lot of dirty things that AI researchers have to do that we haven’t seen proof that AI models can handle. Obviously you can spend more on compute and get more out of AI models — that’s not particularly clever.

Mostly why these aren’t evidence-based discussions is that you and I state different assumptions than the AI 2027 authors would state, and those are really what you have to debate. It’s the assumption of how AI progress is being made and whether algorithmic improvements can translate directly into 2x or 4x or 100x gains.

Jordan Schneider: The other thing is, even though you’re not completely on their trajectory, this stuff is still crazy. All of these Jennifer Lawrence-type things that I’ve been freaking out about over the past 10 minutes are 100% baked in — they already exist and are just waiting to be productized a little bit. That’ll happen in six months, right?

Regardless of whether or not we get flying cars, we are in for a very wild ride. Setting aside the economic impact, just from an interpersonal, day-to-day experience alone, having these things be in our lives much more than they have been.

Nathan Lambert: There’s that side we covered in your personal examples, but also technology companies’ abilities to execute are going to be very honed. It’s obvious that AI companies are trying to do this for AI, but all of these software products where you could scale software much more easily — it’s just way easier to make technology that’s good for the tech industry.

I don’t know if there’s an optimistic vision where that is good for the average person in the world, because there are more things that can exist. At the same time, I expect Google ad placement to get better, Meta ad generation to get better, and Microsoft software products should get better. All these things should just get better using these AI models to write new code.

I listened to the Claude Code podcast on Latent Space, and they use Claude Code a lot at Anthropic — you just tell Claude to do something, it does it in the background, you review the PR. I’ve done it for some basic things. It doesn’t always work, but it’s this autonomous workflow where AI can handle a lot of the tedious work you have to deal with as an engineer quite well. That makes it so you can have fewer teams.

It’s not that software engineers are going to be unemployed — it’s probably going to be more dispersed, and there are going to be more companies. Hopefully it’s a great dissemination of power across the tech landscape, but it’s probably more likely that the big companies get more benefit out of it due to the costs of deploying it and so on. That’s not really a new discussion.

That’s very different from saying that one company rules them all because they’re the one with this model and there are 100,000 AGI agents. The whole argument is based on this happening in one click to make the innovation, and then they’re gone. That’s not something you could prepare for or really argue about because it seems like the root argument is that there’s going to be technological innovations that 2x to 10x the output.

It’s like predicting the inference time scaling with large-scale reinforcement learning — how are we supposed to do that? As a counterargument, it’s unfortunate that it’s not easier, and I think it’s very worthwhile continuing this exercise and trying to push back or push forward.

Open Source at Scale

Jordan Schneider: We’re about six months out from DeepSeek. How’s China doing on the model front?

Nathan Lambert: They’re doing well. DeepSeek R1 exists, and people expect R2 to be released soon, though there’s not really much grounding for that expectation. They released another model thaxt people call a “non-reasoning model” — just a normal ChatGPT-style model — which was very strong. Qwen released their Qwen 3 models, which have great scores.

One of the biggest factors for adoption is that all of these Chinese models have much more permissive licenses than their US counterparts like Llama and Gemma. That’s a subtle but powerful effect for getting more adoption.

At the same time, there’s the license consideration, but there are also stories I hear from medium-sized businesses saying “I won’t let somebody spin up a Chinese open-weight model on my servers for information hazards or tool use code execution risks."

The models they’re releasing now don’t really have PRC influence because they were trained months ago. The PRC is just waking up and following your coverage — they’re learning what a study session is. These political considerations haven’t been translated into policies that impact the models yet.

There’s trepidation out of probably reasonable fear of what future models will be like, leading to very subdued adoption of these Chinese models due to unmeasurable risk. Culture matters, politics matters in technological adoption. I’m regularly surprised when I hear, “Oh, the Chinese model is great, but we can’t deploy it because our IT officer said so.” That matters.

I don’t see this changing as long as this trade war is going on. The actual progress in open models on paper is very good — the best it has ever been. R1 was the first really frontier model with a permissive license, so on paper the open versus closed gap has shrunk a lot.

But if you exclude these Chinese models from the opportunity set, it’s actually kind of the status quo. Llama 4 flopped — it’s solid but flopped in terms of release communications. Google’s Gemma models have been getting better, but they’re actually quite small, so they’re not really at frontier model size if you want to host on-premises. Our models at AI2 have permissive licenses and are about a generation behind Llama in performance, but we haven’t scaled up to frontier model size for pre-training.

If that’s the case, there’s still a big gap between the models that people can actually use for open and closed applications. But I would say the open models have really been smashing through these GPT-4 barriers. Even if they’re smaller, their performance is still well beyond GPT-4’s level. GPT-4 was a step up where so many long-tail things just worked with the models. The open models getting solidly past that is going to be a good reliability point for basic functionality and people trying them.

This is going to take many years to unfold. If you ask people about open source software when it was getting defined, it took over five years to actually reach some sort of agreement. We need to keep showing how to actually use open models, but by no means is it clear how this will play out.

Jordan Schneider: To be clear, the state of models now is: we have these closed offerings from OpenAI, Anthropic, and Google, which are a cut above. Then you have this class of Chinese models which are more impressive from both a licensing and capabilities perspective than what Meta and Google are bringing to the table. But people get the jitters when it comes to actually building businesses on them, as opposed to just tinkering with them.

Which leads us to our question: Meta, trillion-dollar company — why can’t they do this? What’s in the water, Nathan?

Nathan Lambert: There’s a lot of Game of Thrones-level power consolidation happening there. This separates the researchers who can make good output — and they have so many high-quality researchers — from being able to actually execute. There’s a lot of subtlety in avoiding that dynamic.

OpenAI has done a good job keeping their core modeling research to only about 300 people. The people who train the models at OpenAI is around 300 people, while Meta probably has 1,000 with a lot of middle management. That’s the problem — the culture around the technology is just broken there for whatever reason.

I hope they turn it around because the open ecosystem benefits from having multiple players in the right place. But we don’t need to tell complicated stories about how they went down a technical path that didn’t work out or got stuck in some technical debt. They messed up their release either way, and the technical path they’re going down is fine — it’s just that the motivations seem misaligned.

Take the Chatbot Arena situation where they’re calling themselves open models, highlighting a certain result on a model, and then that model with the result was not one that they actually released in their open models. You can’t get away with that — you have to talk about the models that you actually released. If that’s a Zuckerberg-level decision where he’s saying, “Yeah, it’s fine, we accomplished it,” then honestly, that person deserves to go. Whoever signed off on that being okay for a company that can release products — Meta ships a lot of stuff and obviously could get the culture right in some areas.

Some people attribute it to bonus structures — especially these middle managers in generative AI who are going to be making performance bonuses for getting things into production. That’s a lot of money. But it’s more complicated than that because these people were making a lot of money before.

Jordan Schneider: Yeah, you just don’t have resource constraints, and it’s too many cooks in the kitchen. Someone should write the definitive business feature on this — I’ve yet to read the great magazine piece about it, but there should be one. It’s a fascinating corporate case study of a company that has every business reason and every corporate capability necessary to execute on this thing, which smaller teams all around the world have been able to do, and they’re just not there.

Nathan Lambert: It’s more complicated because the Llama models are natively multimodal, which is a big change. As you add more things you’re trying to plug into the model, it really does get complicated.

The dynamic is effectively: you have big egos that did some work, but their work might be slightly subpar, and they’re going to really advocate for their work to be included because they have ego and pride about their work. You have to get them to not do that. I’ve heard somebody not at Meta jokingly say “oh yeah, you should just pay them off. You just pay them more to shut up and get a better model out of it.” That’s something I’ve heard from a more effective organization, so it’s probably something that people have actually done.

Obviously, if that’s happening at another organization, Meta is not the only one dealing with this. It shows how nuanced this is and what a fine line it is — if you literally have to pay people off just to get better results.

Jordan Schneider: Look, Google’s figured it out, which is kind of amazing.

Nathan Lambert: People underestimate AI Overviews, which is an extremely high-usage AI product. All the Gemini stuff might just be a front — the Gemini offerings are a distraction for AI researchers and the people who built the model to show it off, get their peers excited about it, and help with recruiting. All Google needs to do is make AI Overviews extremely profitable and make Google Cloud extremely profitable. All this Gemini app stuff — I don’t think Google was really going to be in the business of making a ChatGPT competitor.

Meta does have the AI app now, which will be interesting to watch relative to ChatGPT. If it’s a sycophancy war, I’m probably going to bet on Meta, but I don’t know.

Jordan Schneider: They’ve got the Meta DNA in them now with Fidji Simo as the new CEO of Applications. Maybe she’ll go even further than Zuck let her in the heady 2010s days.

Nathan Lambert: I don’t know. I’ve read Careless People, the memoir by the Meta Global Affairs Policy person — good read. The TLDR is that Zuckerberg let them push when he didn’t have the reins on growth and other things. Many of your listeners understand the dynamics of a major tech company pushing for growth and influence, primarily internationally and with political power struggles.

Jordan Schneider: We’ll put it on the list. Anything else on the AI stuff before we move on?

Nathan Lambert: When we consider our careers relative to some of these X-risk warriors and takeoff scenarios, there are still many years of very solid progress on raw capabilities and making money off of AI. There will be a bubbly correction when it comes to VC startups. Many of them raised on the assumption of two to three years of runway, and some of those are going to start running out. They could look like big scalps, but most of it is GPU collateral.

It’s this weird situation where they raise a huge round, but most of it is collateral on these GPUs which don’t go away if the company goes away. People will probably overreact to some companies going away, but it was bound to happen. Overall, the AI industry seems super healthy, which is remarkable when you look back two years. People are raising these rounds — it’s not like Mistral raised that round as a seed round where they give away 45% of their company and they’re still functioning.

Jordan Schneider: Nathan, you wrote a nice career reflection post. What do you want to say to the kids out there?

Nathan Lambert: It takes a long time. People don’t want to sit here recording podcasts for five years before finally making any money.

Jordan Schneider: More like seven.

Subscribe now

Nathan Lambert: Yeah, there are two stories. One is that it takes a long time even if you’re at Berkeley. Some of the anecdotes I had were that I was astute enough to pick up that AI was obviously the thing happening in 2017. But even if you’re at a place like Berkeley or Stanford, you can’t just get into one of these research groups. You can’t just pivot. You have to do this long, slow slog and wait for somebody who’s good and sincere to ask to collaborate.

In that post, there were other things I had tried to do, like collaborate with postdocs who didn’t give a shit. That didn’t work. It’s the between-the-lines details that don’t matter, but it just takes a long time. It took me a year to get a real collaborator while being a motivated grad student at Berkeley. If you’re not at a top program, it’s going to take you longer. You might find a collaborator within a year, but it’s not going to be a postdoc in a leading AI lab. It’s going to be a random other person who wants to do AI research.

It just takes a long time and a lot of cycles. People ask me about this with athletics, too. Some random mid-aged tech person who’s losing track of their health asks, “Can I qualify for the Boston Marathon?” I’m like, “Yeah, you absolutely can. You just don’t want to do the 48-week training cycle that I’m going to sign you up for.” It’s not crazy — you’ll be doing seven to ten hours a week of training for 48 weeks, and I can tell you exactly what you need to do. You’ll get injured a couple of times, but people just don’t want to do that work. That’s most of the problem in AI.

It might be a little easier than training for the Boston Marathon because it doesn’t hurt as much physically, but it just takes a long time.

The other side is reflecting on what execution actually means and the difference between a career at Hugging Face before ChatGPT, doing random things, versus a career at Hugging Face after ChatGPT, when you’re doing something that’s actually aligned with the community. It’s very hard to know at the time that you’re doing the right thing. I was committing code and working on teams, and projects would come and go — it all looks very normal. It takes a lot of experience before you know you actually have traction on something.

Now I could see it a bit better just because I have a baseline for when things actually land. That’s something that takes a lot of experience to develop. A lot of companies will make it hard to do that. A lot of careers in big companies involve playing the promotion game that all of your peers are also playing.

The goal is finding a team or area within a company where that’s obviously not the goal — where leadership cares about direct deliverables, so you can skip a bunch of levels and get noticed. Or find work that’s done in public. It’s not always just about promoting within the company; you could also get mindshare outside of the company, which is valuable in both ways.

Internal company work is how you get promoted, but external mindshare is how you always have a job available to you. That kind of power means I can do what I want. I’ll just go get a different job if I want a different job. Once you have that, you can take a lot more risk. You could just write blog posts for six months and say, “I’m just going to ship it."

Jordan Schneider: This is why everyone listening to this needs to start their own Substack and newsletter. It’s particularly important on the China and tech analysis side. If you’re at a consulting firm and you do good work, maybe five people will know about it. If you write something online and you do good work, your entire career cohort and all the hundreds of other future people who could hire you will know about it.

It’s important to do a good job in your work, especially when you’re junior — actually, at any time. But people forget that the number of people in your field and the opportunities out there that are beyond the direct thing you could be promoted for in your company are orders of magnitude larger. Particularly now that we live in the age of the Internet, you can write things even anonymously and people can read them.

Having some sort of public profile or portfolio that you can gain credibility points from is something that everyone should be taught by their career counselor when they’re 21 years old looking for a job.

Nathan Lambert: Jobs where you are very open about what you’re doing and can do other open things provide far more consistent career growth. Even if your output is mid, you’re going to have growth proportional to that mid output. If you have mid output at a closed lab, you probably have no growth. You don’t get promoted, you don’t have any opportunities.

It’s not the only way to do things, but I definitely agree. In this post, I talk about a kind of backcasted strategy: AI is getting really closed, so what if I just keep doing what I’m going to do and don’t go off to some random flashy job? A few years later, nobody could talk about anything, but I’ve established myself as a person who knows how to train AI models and is transparent about it. People associate me with all of AI2’s output, and I’m not responsible for all of this stuff — AI2 doing well takes a lot of people.

Most of it is just work. I’m at the point where I tweet a lot, and on all of these platforms, the number of iterations you put in are proportional to your growth. Growth is easier once you have a bigger following. It’s often just starting from scratch when you have no following to slowly build up — plus one, plus two, plus three. The first thousand Twitter followers are hard, but once you have a thousand, it’s pretty easy to get more Twitter followers.

Getting through that phase in early to mid-career is pretty valuable. Getting to the point where you have some small level of distribution will help open a lot of doors.

Jordan Schneider: This is what young people don’t understand because all they see are LinkedIn job postings and recruiters coming to your college career fair. Once you get to a certain point, the jobs get created for you and there is inbound — that’s where you want to be living. But getting to that place requires people to know you exist and know you do good work. That only happens if you work in public.

Nathan Lambert: I can give an anecdote about how me being public brings a lot of value to AI2. The last three people we’ve hired on my team at AI2 have either been people that I’ve been working on recruiting for years, wanted to recruit for a while, or cold inbound to me. We still filter and read all these applicants, but there’s a lot of lukewarm stuff.

These are people who either wrote a killer email and had work I could easily look at — open source contributions, for example. In one case, someone worked in computer vision but wrote a kick-ass blog post on an RLHF method. I was like, “Okay, that’s cool.” Other people I had known through their writing or being in this open community and talking for years — those are three of the last people I’ve hired onto my team, and I’ve read hundreds of resumes.

There are other academic hires where we hire a random postdoc — that’s an academic thing. But these are engineering roles, and that’s how it works for individuals at AI2. It’s really valuable.

Jordan Schneider: We have this great Confucius line: “At 15 I had my mind bent on learning. At 30 I stood firm. At 40 I had no doubts. At 50 I knew the decrees of heaven. At 60 my ear was an obedient organ for the reception of truth. At 70 I could follow what my heart desired without transgressing what was right.”

It took me until I was 35, but I do feel like I am standing firm now. A lot of that had to do with the fact that I have this career foundation which has been built by doing things out in public.

Anyway, I shouted this out on the episode with a few weeks ago, but write your Substack, write five articles, reach out to me. I’ll give you feedback, we’ll do cross posts, whatever. The world needs more thinking about technology and China and all that good stuff. You can give it to them — I know you can.

Nathan Lambert: The thing about writing that people miss is that there are no new original ideas, really. It’s all just repackaging information in your worldview. The algorithmification of everything actually rewards people with voice. Fifty years ago, just writing the information down was of such value. There are still areas of the world where that is the case, and AI is one of them.

But in terms of building a career, so much of it is your voice and the perspective you put on things and the engagement that adds. Personally, many of my ideas on how I analyze AI model releases are downstream of reading something like Stratechery for multiple years. Ben Thompson’s beat is company earnings and how platforms are evolving. My beat is AI performance and how AI models work. At this point, I just do it and write my state of the world, and it looks like it’s about a specific model. Yeah, there are plots about evaluations, but really it’s about this trend of how models are evolving and how I see things.

There are so many different ways for people to do that. It mostly takes practice.

Mood Music

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【和放学以后永不失联】订阅放学以后Newsletter，每周三收到我们发出的信号：afterschool2021.substack.com 点击链接输入自己的邮箱即可（订阅后如果收不到注意查看垃圾邮箱）。如需查看往期内容，打开任一期你收到的邮件，选择右上角open online，就可以回溯放学以后之前发的所有邮件，或谷歌搜索afterschool2021substack查看。

截至目前，放学以后Newsletter专题系列如下：“在世界游荡的女性”系列、“女性解放指南”系列、“女性浪漫，往复信笺”系列、莫不谷游荡口袋书《做一个蓄意的游荡者》系列、“莫胡说”系列”《创作者手册：从播客开始说起》,播客系列和日常更新等。

大家好，本期Newsletter由西班牙的霸王花木兰轮值。

今天写的这篇文章是源于5月份我做的一个梦。做这个梦大概率是我在现实生活中遇到困扰，白天问题没有解决，晚上压力便在梦境时分得以被释放，一些被隐藏的情绪、记忆便都涌现了出来。那段时间刚好我在阅读《活出生命的意义》《空洞的心》，又和莫不谷在Alicante深度夜聊，我对自己的家庭环境和成长经历，特别是生命最大的课题和挑战有了新的思考和感悟，所以在梦醒时分，我就立刻打开了电脑，把我做的这个梦，和我的想法记录了下来。

鉴于文章涉及较多个人信息，请大家不要以任何方式分享在互联网上，感谢理解和支持！

往期文章：

《莫路狂花2：如何对自己充满爱意和敬意，免于混乱逃避低活力？》

《54 我不要只做世界的承受者，我要对这个世界一顿发起！》

《在5月的尾巴来一些重磅更新！》

（这是我在游荡南美哥伦比亚的时候，在校园的草坪上睡着了的图片）

我昨晚做梦突然梦见我的母亲了。还记得自己在梦里写下了很多文字，想要把内容抄写下来作为我的创作素材。

醒来的时候发现原来在做梦，想要记下来的文字也随着清醒变得有些模糊，只有一些印象深刻的场景还未从脑海散去。

我梦见，我的母亲被中学学校邀请作为学生家长代表站在广场的台子上发言，台下是正在备考奋战坐的整齐的学生，老师们站在台子的一侧。母亲的状态并不太好，记得她控诉我没有感情，不够理解她，养出来的孩子不是她的。细节内容我已经忘记不少，依稀记得我当时正好从台下走过，原本不作理会，却在听她控诉时忍不住想要说几句。于是已经穿过台子的我转回头，从台下老师的手里找到了一个麦克风，站在学生面前说，我也想发言。

手里握着话筒，我想介绍下自己：大家好，我是XXX，毕业于XXX学校，工作于XXX公司，现在已经辞职成为一名无业游民。原本我想这样介绍自己，转念又觉得这未必不是另外一种“包装”，便直接开口和母亲说到：

“你说我为什么会这样？我认为，对于还未成年的我，这些责任要归咎于父母，对于已经成年的我，这些责任我会自己承担起来。”

为什么我不会依赖父母？因为在我最需要依赖亲情的时候，我并没有感受到亲情。小时候父母时常有争吵，争吵时便询问我：“如果她们分开离婚，我会选择跟谁？”那时我做不了选择，也并不想跟谁在一起，如果可以解决经济问题，我会选择谁也不跟。所以当我在小学课堂愣神的时候，终于想到父母分开我该如何活下去的答案时，我的心里就关闭了某些窗户。这个答案是，我只要比父母死的早，就不必担心父母不在，没有经济来源的我活不下去的问题。这个困扰是有解决答案的，想到了答案，我便回过神安心地继续听课。

为什么我要远离父母？看起来我考上了好的学校，有了好的工作，我的母亲还成为了学生家长代表，但是她知道，我也知道，这场有着好的主流评价，令亲戚邻居身边人称赞认可的教育是失败的。因为最近我才深刻意识到，我生命最大的威胁和挑战是，抵抗死亡的诱惑。

为什么我时常想死？或者说，为什么我常常觉得死亡是我面对自己无法解决问题的第一答案，也是终极答案？我想，这与我的成长环境，与我的父亲与母亲脱离不了关系，因为死亡、暴力，是一种会在家庭蔓延的传染病。

我想起小时候的一件事情，一件我被姐姐吓住，不敢真的招惹她的事情。先来介绍下我的家庭背景，我有一个姐姐，比我大一岁，我还有一个弟弟，比我小两岁。记不得是几岁，有一天晚上，姐姐突然向我举起了菜刀。我一时怔住，不知道该怎么办，是她在梦游吗？还是她在生气？那时候的我突然想到看过《圣经》里的一句话，“如果有人打了你的右脸，你就把左脸伸过去让Ta打”。不敢反抗激怒姐姐的我，弱弱地说，你要不要打我另外一边脸？后来这招真的奏效，姐姐放下了她拿起的菜刀。写这篇文章的时候，我找姐姐想询问她那时拿菜刀的原因，却得知她小时候因为吵架好几次拿过刀，但记不清对我拿菜刀的事情。进一步询问原因时，姐姐给了一个答案，模仿母亲。

（我和姐姐的聊天对话）

除了姐姐，我记得自己曾对弟弟做过的一件事。小时候，弟弟聪明却贪玩，沉迷网络游戏和交被我称之为狐朋狗友的朋友。有一天，他的朋友来找他，我知道不免又是去泡网吧，为了拦住我弟不要同流合污，我作势拿起大大的扫帚要把狐朋狗友拦在外面，记得那时候我也曾威胁我弟，如果他一心要出门，我就要拿菜刀阻拦。无计可施，无法控制弟弟的我，也学会了用暴力和威胁强制表达自己认为“绝对正确”的意见。

又记得弟弟高考前发生的一件事情。这件事约莫是大学或者更久以后我从姐姐那里听来的。高考前弟弟还是沉迷网吧，和我每天熬夜努力学习不同，他几乎不怎么做作业，但数理化依旧能考出高分，这使得对他抱有北大清华梦想的母亲被折磨到崩溃。据说第二天就要高考了，弟弟还沉迷在网吧里，母亲无计可施，于是当着弟弟的面要跳楼，听说是父亲拦住了母亲，弟弟当场下跪求母亲不要去死。我没想到弟弟经历过这样的事情，他没和我说过，我也不知如何开口去问。

说到母亲，母亲经常会情绪崩溃，呈现出一忍再忍，无法再忍导致的完全失控。当她无法再忍耐情绪爆发时，就会像火山一样难以抑制，甚至会狠狠打自己耳光，会大哭会完全崩溃。小时候刚开始我们见到这一情形，就会恐惧害怕自责，担心母亲，后来慢慢习惯，便开始麻木，甚至冷冷地沉静地说到：“还有没有新的招数，能别老是这一套吗？”

研究生毕业时，我曾因学业压力一度陷入崩溃，临近毕业还未毕业的我觉得学业不成，人生也没什么可能了。我的状态太差，姐姐知道后告诉了家里，母亲和弟弟便来到北京看我，记得她们来的时候，我内心平静，装作波澜不惊，带她们逛学校，出去吃春饼，烤鸭。其实心里想的是，母亲来了也好，可以替我收尸了。弟弟为什么也来了？平时和弟弟交流不多，没想到他也来了北京。后来我得到另一个角度的分析，让我心下一沉：“因为他不能眼睁睁看着家里第二个人不想活着，他不愿再见到悲剧的发生。”

前两年，我工作已经好几年，在上海有了自己的生活，让母亲来上海和我一同居住方便游玩，刚好7月7日是我的生日，弟弟不远千里请假也要来上海给我庆祝生日。那天我在徐汇滨江提前预定了美丽的法式餐厅，在去往餐厅的出租车里，我还在劝弟弟购买商业保险，特别是经常加班熬夜，万一生病不舒服也有个兜底。弟弟听得不耐烦，“哎呀，不买不买，死了算逑。”听到弟弟这么一说，我心里一惊，怎么他也不想活？接着便释怀了，“嗨，我心里不也是同样不想活吗。”这时，同样在车里的母亲却崩溃了，情绪上来，满脸的悲伤，无奈与愤怒，要求让司机立刻停车，她要下车，崩溃地说：“小时候那么难，吃不上饭的时候都活下来了，现在是怎么了？怎么都不想活？”

弟弟讪讪一笑，站在路边摊手无奈，我一时也不知道该怎么办。母亲不顾劝阻，边走边大哭，路边两位好心女士见状还上前询问母亲是否需要帮助。原本我只担心不会打车的母亲独自一人可能回不了家，但看到母亲距离黄浦江越走越近，我像是预知可怕后果般感到恐惧袭来，我不希望自己的放任导致悲剧无法挽回，便赶紧让弟弟上前劝阻，弟弟不听后我又赶忙上去拉住母亲，好话说尽各种道歉把她劝进了屋里。最终，生日那天，在母亲悲伤不能自已的眼泪中，我们吃完了预定的美丽精致的法餐。

坦白来说，我不知道该如何书写我的家庭，我的母亲，也不知道该如何认识和理解母亲，往事暗沉不可追，很多症结如今已经很难解开。小时候，因为 母亲一出生是女儿，所以她的父亲曾经想要毒死她，也能少一个张口吃饭的。然而她的父亲在食不果腹的艰苦年代太早过世，母亲反而对他念念不忘，常常感念自己从父亲身上学会很多生存和做人的道理。这既让我不解也让我忍不住怒其不争，因为被害者对一个杀人犯念念不忘就是对自己最大的残忍。写到这里也忍不住在想，从小遭受世界残酷对待的母亲又如何习得好好对待自己的方法呢？

小时候，母亲因为家境贫寒，没有上学读书的机会，但她天性喜欢阅读和学习，讨厌做农活，便躲到一处看书，也因此常常被家里和街坊四邻吐槽好吃懒做就是不干活。母亲的一生好像就在争一口气。为了争一口气，坚持让姐姐，我，弟弟读书，砸锅卖铁也要让我们上学。等我们考上大学时，农村老家远亲近邻都夸赞母亲能干，一个大学生就了不起，居然能培养出三个大学生，再也没人会说她好吃懒做。为了争一口气，不被农村亲戚看低，所以妈妈生了弟弟，还因为计划生育东躲西藏，甚至被抓进看守所好几天。我的父亲曾因为只有一个男孩，被有着三个男孩的哥哥，我的三伯当众打了一耳光。母亲知道这件事情后，只身回到了老家，打了我三伯一个耳光，因为不愿意让我的父亲受气。

我的父亲，一个话少沉闷的人，唯一的优点就是吃苦耐劳，很难说这到底是优点还是缺点。作为煤矿工人，他靠着吃苦耐劳拼出性命赚钱养家，供三个孩子读书，他只遭遇了粉碎性骨折、手部受伤此类大大小小的工伤，幸运地活着从岗位熬到退休。由于父亲的性格，胆小谨慎，甚至有些笨拙懦弱，家里大大小小的事情都是母亲出面，小孩外地上学搞不定户籍，父亲被老油条的同事欺负，被领导要求过度加班，农村老家老人赡养，残疾补助申请，亲戚遭遇车祸、纠纷诸如此类问题，都是母亲想办法解决，也因此不论是农村老家，还是父亲职场同事领导，都知道我们家有一位厉害人物，不能轻易招惹会发疯的人物——我的母亲。

现在我的母亲也时常“发疯”，情绪崩溃，不受控制。特别是前两年我的姥姥、奶奶、伯母短时间内陆续去世，母亲几乎双鬓斑白，听说葬礼现场母亲情绪上头，就无人能拦。我的父亲默默承受着这一切。我想，我的父亲母亲在孩子尚小时吵闹离婚，如今历经半载，当姐姐，我，弟弟长大充分理解并同意甚至劝告她们离婚，甚至开玩笑说，正好我学法律可以帮助她们起草离婚协议，却遭到她们正色拒绝时，这段关系已经进入到一种难以分离的畸形状态。所以即使母亲时常发疯崩溃，父亲苦于承担，彼此也有着难以解脱的羁绊。

我怨恨过母亲，同情母亲，我同情过父亲，怨恨父亲。我甚至有些看不清楚束缚母亲的结到底该怎么解，我坦白，这超出我的能力。而我，只能尽力解决我的人生课题。小时候，我给自己找到的答案，“我只要比父母死的早，就不必担心父母不在，没有经济来源的我活不下去的问题。”这既是我的解药，也是我的毒药。当死亡成为解决问题的答案，当死亡的身影频繁在我的生活里出现时，如何拒绝死神的诱惑？

如今，我断了和母亲的联络方式，离开了故土，来到了遥远的异国它乡，小时候从未想过的西班牙。记得从家离开坐上出租车的时候，母亲突然紧紧攥住我的手，面部紧缩皱成了一团，眼泪和话语同时挤出：“我不知道你会离开这么远。”

我曾和一个朋友聊过离开的考量之一，是我想在我这代终结悲剧的命运。在我成长的故土，我的姥姥不幸福，我的母亲不幸福，20世纪出生的我也无法逃离重男轻女、计划生育、疫情封锁等等苦痛。记得之前看过蔡崇达的《命运》，我已经太过熟悉，又太过厌倦这片土地上作为女性，作为人类遭遇的苦痛折磨了。在我目之所及里，我找不到幸免于难可以平安顺遂生活着的人，我也看不到自己有这样的未来可以期待。

所以我决定离开，决定出海找找逃离悲剧的可能性。同时我也早早决定不婚不育，让痛苦的生活终结在我这一代，让我终结痛苦的代际传递。说到为什么不婚不育，小时候我还畅想过，如果我自己有孩子，我要让她过上与我不同的生活。稍大一点，我就没有这种畅想了。先不说我很怕痛，时至今日打针抽血都会嗷嗷哭的我，忍受不了生育对我身体的摧残，想象自己会上手术台，身体遭受种种折磨，这就是我清清楚楚知道我最不愿意过的生活。

另一方面，我知道我无法给予我没有的东西，我的孩子并不会因为我是她的母亲感到幸福，反而更容易陷入痛苦。我自己知道，我温和平静外表下隐藏的控制欲，我的情绪不稳定，我的情感隔离难以建立亲密关系，都会让新的生命受苦（人也无法抑制对比自己弱小的生命的控制欲和暴力倾向）。坦白来说，我的孩子痛恨我，想要逃离我，像我对我母亲一样有着复杂的情感，都是极有可能发生的事情。

我不知道我对父亲母亲的描述，是否过于主观，是否有失偏颇。记得之前看易立竞访谈马薇薇，马薇薇不愿意评价父母，担心自己只记住了不好的一面，忽略了父母客观的付出。我想，尝试记录，尝试描述，或许会有偏颇，但自己记住了什么，感受到了什么，对自己而言很重要，对帮助自己回答生命的意义有帮助，所以先写再说。另一方面，我的母亲现在也常常写作，她有一腔情感和无数文字要倾诉记录，虽然因为阅读她的文章给我带来较大情感压力，我并不愿意去看，但这并不影响她持续表达。女性爬山，我们各自努力。

我也在想，现在这个阶段 ，还在回溯过去的问题是否还有必要。有次我刷到网上一个帖子，说“黄圣依都四十多了，还在围绕在原生家庭的话题。”评论区有个印象深刻的回复是，“现在聊原生家庭都要卡年龄了吗？”说实话，我曾有过发帖博主的想法，想过为何不Let it go？与其沉湎纠结过去，不如向前一步。现在想法有些不同。上大学的时候我就因为单纯对心理咨询感兴趣报了学校的团体心理工作坊，那时候我觉得自己没有任何心理问题，没有痛苦，也没有原生家庭问题。现在，随着我对自己认识越来越敏感，在朋友帮助下，对过去的认识越来越深化。我不得不承认，现在困住我的拦路虎，不是任何其它的阻力，而是我自己，那个被自己困住的我自己。

所以想想觉得人还是挺奇妙，在你挣扎想要生存活下来的时候，身体会调动一切资源帮助你解决生存危机，但当你不再面对生存危机，身体有机会休息时，便会将那个困在童年无力应付世界的小孩释放出来，让成年的你有机会帮助自己。

此前，莫不谷曾经和我说，既然我已经来到了西班牙，就可以尝试重新开始，不要再在意别人的眼光去生活（这里不是引用的原话，可能表述和原话有出入）。听到她这么一说，真是觉得令人振奋，冷静一下，又觉得有些难度。但我还是要写下来，以此共勉。

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）
这世界的辽阔和美好，游荡者知道。使用过程中遇到问题，欢迎联系客服邮箱wanderservice2024@outlook.com.

【放学以后文章&书籍&其它】

解锁放学以后《创作者手册：从播客开始说起》：https://afdian.com/item/ffcd59481b9411ee882652540025c377

解锁莫不谷《做一个“蓄意”的游荡者》口袋书：
爱发电：https://afdian.com/item/62244492ae8611ee91185254001e7c00微信公众号：《放学以后After school》（提示安卓用户可下载“爱发电”app，苹果用户可把爱发电主页添加至手机桌面来使用，目前爱发电未上线苹果商店）

Newsletter订阅链接：https://afterschool2021.substack.com/（需科学/上 网）

联系邮箱：afterschool2021@126.com （投稿来信及合作洽谈）

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）

小红书：游荡者的日常

同名YouTube：https://www.youtube.com/@afterschool2021

同名微信公众号：放学以后after school

欢迎并感谢大家在爱发电平台为我们的创作发电：https://afdian.com/a/afterschool

播客收听平台：【国内】苹果播客（请科学/上网）、爱发电、汽水儿、荔枝、网易云、小宇宙、喜马拉雅、、QQ音乐；
【海外】Spotify、Apple podcast、Google podcast、Snipd、Overcast、Castbox、Amazon Music、Pocket Casts、Stitcher、Radio Public、Wordpress

亲爱的读者六月好，感谢耐心等待～

东德简史第一节：《生于不义 1945-1949》，简单聊一聊东德的“建国大业”。

图片：1949 年 12 月，莫斯科，为斯大林贺寿。斯大林正中，左手边德共乌布利希，右侧中共毛泽东。

帝国两翼

1949 年 10 月 1 日，中华人民共和国成立。

1949 年 10 月 7 日，德意志民主共和国成立。

苏联帝国势力范围 (Sphere of Influence) 在亚欧大陆上方展开双翼，东侧直达南中国海，西侧伸展到易北河畔。

中华人民共和国的成立，是斯大林输出革命和毛泽东权谋机变的协同结果，它是共产主义的巨大成功：共产主义沿着亚欧大陆东缘一路向南突进，东南亚大门洞开，形同一系列多米诺骨牌等待被逐一推倒。

而东德的建立却是一场苦涩的战略失败，苏联迫于无奈接受1/3 个东部德国独立建国，代价却是更庞大的德国——拥有鲁尔区的西德彻底融入西方阵营。冷战边界在中欧固化，苏联影响力被封堵于东欧。西柏林飞地却如同肉中刺扎在东德心脏里，也扎在苏联敏感的神经上。

如果不符合苏联利益，那东德，这个斯大林“并不喜欢的孩子”，为何能够诞生？

二战废墟上，冷战阴云下，横亘着战后欧亚大棋局。苏联美国固然是执棋之手，棋子们同样有自己的意志，他们在有限的活动空间中各寻机遇。

如同毛泽东，乌布利希也是一颗不安分的棋子。

不安分的棋子乌布利希和他的共产党，塑造了东德。

二战废墟上

1945 年 4 月 30 日晚，希特勒在地堡举枪自杀。同一天，乌布利希从莫斯科登机，经停明斯克，降落在老家萨克森林地里一个简易机场，随后换乘苏联红军车辆，一路颠簸，于傍晚抵达柏林郊外的苏军前锋指挥部。

阔别十二年，柏林如此陌生。帝国首都已成废墟，枪炮声在街垒间回荡，呼吸间尽是硝烟的味道。

1893 年乌布利希出生于莱比锡市一户裁缝家庭（同一年，毛泽东出生在湖南韶山冲一户地主家庭）。

1918 年 25 岁的退役士兵乌布利希加入了新创立的德国共产党，1920 前往莫斯科参加共产国际，历经多次培训后返回德国，先在莱比锡党部发展，1928 年以德共议员身份被选进入魏玛共和国国会。

在德国的国会里，乌布利希议员忠实执行莫斯科的指令，坚决斗争社会民主党，他辱骂社民党是“社会法西斯”（莫斯科这一政策，一定程度上源自 1927 中国国共合作试验的失败）。左翼分裂帮助了纳粹登场，1933 年希特勒利用国会纵火案通过“授权法”。共和国告终，乌布利希出逃，一路辗转回归莫斯科。

莫斯科岁月里，乌布利希进一步学会了和总路线一起摇摆。

（插播苏联笑话一则。在调查表上有这样一个问题：在执行总路线时你动摇过吗？拉宾诺维奇回答道：我和总路线一起动摇。）

当其他德共党员诚实地将纳粹上台的责任部分归于莫斯科的错误指导时，乌布利希坚持认为莫斯科都对，只是德共自己执行出了问题，德共才是一群需要做检讨的失败者。

当斯大林 1935 年改变路线，要求共产党和社会民主党、资产阶级右翼组织“人民阵线/统一战线”共同反对法西斯，又是乌布利希第一时间跳出来为新路线辩护：法西斯是最危险的敌人，因此必须和社民党和资本家联合。

当斯大林 1939 年再次大转弯和德国纳粹达成互不侵犯协议并瓜分东欧，乌布利希毫无障碍地辩护说：西方资本主义才是最凶恶的敌人，为此应该和纳粹和解。他似乎已忘记有多少德国共产党员关在纳粹集中营里。

当纳粹德国突袭苏联打到莫斯科城下，这一次狡诈凶恶的纳粹又成了苏联最大的敌人，自然也是乌布利希最大的敌人。

眼睁睁看着各国共产党员死于斯大林大清洗，目送一批批共产党员被苏联的路线转弯刷下车，随着总路线一起摇摆的乌布利希不仅保全了自己的生命，还在德共内部不断高升，获得了苏联信任，建立了属于自己的核心团队。

这便是传说中的坐着苏联坦克归国的德国共产党小组，又称“乌布利希团队”。

分裂的目标

苏联把乌布利希团队投放到柏林，看中乌布利希的忠心，目的是在战后德国维护苏联利益。

图片：德国和柏林被分区占领，苏联占领区中心的柏林市同样也被分成了四块。

什么是苏联利益？

战后德国被苏、美、英、法分区占领，然后如何安排战后德国政治，盟友之间并无一致协议。

赤化整个德国缺乏实现的可能，虽然这是列宁一代的远大理想。

斯大林最好的选择是将整个德国中立化（芬兰化），同时通过德国共产党来保持对中立德国的影响力。这样能彻底消除德国再军事化的可能，毕竟这是一个连续两次攻击俄国/苏联的战争策源地。

为了实现这一目标，需要伪装出民主的样子，搞统一战线，只有如此才能获得美英法的同意让德国恢复全国政治。

苏联做出民主表率，苏占区 1945 年率先允许德国人组织政党，开办媒体，进行多党竞争。但苏联同时给予德共大量便利条件，乌布利希对自己团队成员私下交底：“表面上要民主，实际上我们要掌握一切！”

这套假民主真专制的统战手法能骗到英美吗？能骗到英法美占领区里的德国政治精英吗？

当然不能。纳粹统治从 1933 到 1945，只有十二年。魏玛共和国一代的议员和政治家们多数还活着，他们曾经在魏玛共和国民主体制下和乌布利希们共事，共产党能民主？在远东能骗到人，在德国毫无可能。

乌布利希当然知道斯大林这套手法根本不能起效，但这位机会主义者当然不会违逆斯大林的意志，他反而要坐等这套策略破产。

因为一个统一且中立的德国不符合乌布利希的利益。

统一且中立的德国，如同魏玛共和国，共产党在议会里只能充当左翼反对党或执政联盟里的合作者。乌布利希在魏玛国会里已经坐够了冷板凳，只有把苏联占领区变成自己的独立王国，由共产党独自统治，才符合乌布利希的个人利益。

因此在执行苏联策略时，乌布利希团队对 “表面民主” 敷衍了事，对“掌握一切”十分认真。

1946 年春，德共强制吞并社会民主党。两党合一更名为“统一社会党”（以下行文将继续把统一社会党 SED 简称为德共），反对合并的社民党员批量逃离苏联占领区。

1946 年10 月苏联允许苏占区各州进行半自由选举，结果德共虽然领先，但在所有州得票都没能过半。饱受打压的自由民主党和基督教民主党联合，成功赢得大量选票，和德共分庭抗礼。

统一战线政策不仅难以蒙骗西方，连在苏占区都无法保证德共“控制一切”。

冷战阴云下

1946 年，马歇尔将军调停失败，中国内战白热化，冷战首先在远东爆发。

1947 年 1 月马歇尔返回美国，随后出任杜鲁门政府的国务卿。6 月 5 日马歇尔在哈佛大学演讲，正式提出欧洲复兴计划，史称“马歇尔计划”。

美国应该尽其所能帮助世界经济恢复正常健康，否则就不可能有政治稳定和有保障的和平。

我们的政策不针对任何国家或任何学说，而是针对饥饿、贫困、绝望和混乱。 旨在复兴世界经济，为自由制度提供提供存续所需要的政治和社会条件。

……

任何借机阻碍其他国家复苏的政府不会得到我们的帮助。更有甚之，那些寻求使人类苦难永久化、以便从中获得政治或其他利益的的政府、政党或团体，将遭到美国的反对。

《马歇尔哈佛演说》

斯大林不准所有东欧共产党政府参加马歇尔计划，捷克不服从，被召到莫斯科管教：“参加马歇尔计划就是反对苏联”。捷克外交部长哀叹：“去莫斯科前我是外交部长，回来时我是苏联的奴仆。”

冷战在欧洲彻底爆发。

英美法占领区陆续合并，从州级开始组建民主政府，筹备成立联邦政府。1948 年西德开始经济复苏和币制改革，德国分裂已成定局。

1948 年 6 月，苏联以封锁西柏林要挟西方放弃西德建国方案，美国派出大批运输机向柏林250 万居民空运粮食及各种日用品，一年间飞行次数 27 万余次。

四年前美军轰炸机带来炸弹和死亡，这一次带来粮食、孩子的糖果和自由的希望。西柏林儿童亲呢地把飞机叫做“棒棒糖轰炸机”。

1948 年 9 月西德各州代表齐聚西南小城开始制宪，以尊重人权、自由选举和联邦制原则进行民主重建。同期乌布利希开始清洗德共，撕破统一战线的民主伪装。

1949 年 5 月 23 日，联邦德国基本法签署，西德成立，阿登纳带着 2/3 个德国毫不犹豫地加入西方阵营。作为莱茵兰的天主教徒，阿登纳宁愿带着一部分德国人先拥抱自由，也不愿意选择在苏联管控下的所谓“统一”。

1949 年 9 月 16 日，乌布利希带领德共高层飞往莫斯科，等待斯大林批准东德建国。斯大林陷入长时间的思考，将德共团队晾在莫斯科郊区十余日。

拥有鲁尔区煤铁资源的西德彻底倒向了美国，迟早会再次武装成为苏联的大敌，封锁柏林失败，斯大林的战后德国安排全部付诸东流……一切已无法挽回，苏联只能接受最坏结果，东德建立国家，自由的西柏林却如同肉中刺插在它的心脏里。

斯大林无奈放行。

1949 年 10 月 9 日，德意志民主共和国成立。乌布利希得偿所愿。

扫清废墟，重建家园
面向未来展雄翅，
为了你的崇高事业，
统一的祖国德意志。
……

幸福前程，和平生活
降临祖国德意志。

……

我们耕种，我们建筑，
学习、创造干劲足，
自由的一代茁壮成长，
相信自己有力量。
德国青年不要辜负
来自人民的期望，
你们就是祖国的未来。

东德国歌 《从废墟中崛起》

双重危机

虽然国歌里写满了“统一，幸福，自由”，可这个新建立的国家既不统一，又不幸福，更无自由。

东德的产生，既无斯大林的祝福，又不符合德国人民自己的期望。它的存在本身只符合德共的利益：在 1/3 德国独享权力，以一党专制模式建设一个金字塔体制，党高居于人民之上，领袖乌布利希高居于党之上。

这一权力稳固吗？它从诞生的第一刻起就内蕴双重危机：

1，苏联可能出卖东德来换取德国中立

2，德国人会顺从共产党的严酷统治吗？

名义上的东德成立了，但建国未立国。

下一章《反抗的国度》 写动荡的 50 年代：斯大林的 1952 纸条、贝利亚试图出卖东德、1953 年的六一七起义，一路到东德真正的立国斗争：1961 年建立柏林墙。

直到德共建立一堵墙把人民关起来，东德才真正立国。

闲话几句

我写东德史首先用的是跨国视角，希望呈现冷战背景下亚欧大陆东西两侧命运的关联和互动。

其次受后修正主义冷战史学的影响，我强调本地角色所起的作用。传统史学里往往过度强调美苏，似乎本地政治人物只是一些被动的棋子。

对乌布利希的分析，主要来自于 Wolfgang Benz. 苏联策略的失败，出自 Ulrich Mählert. 斯大林并不愿意分裂德国，观点来自 Wilfried Loth.

再次感谢你们的耐心等待，下一章七月更新。写长文不易，我慢慢写，您慢慢读……感谢包容。

如果喜欢请推荐给家人朋友订阅，谢谢：

改变的气息，总是催人振奋。我们这代人不管后来变成什么样，人生的成年好象是从1989年那个夏天开始的。那时候，我们向往民主，但没有见过民主，以为拼命抗争，就可以一蹴而就。一蹴而就的是那场惨剧，民主却离我们越来越远了。只有事件，没有进程。但民主不是一个事件，而是一个进程。近代以来，中国不缺少大大小小的追求民主的事件，但反复难产，孩子生不出来，也无所谓成长。

六四是中国现代文明化反复难产中一场母婴双亡的惨剧。三十六年后，被那场惨剧塑造的一代人也逐渐进入暮年。很多曾参与其中的年轻人和大妈大叔背着六四的行李箱逃出来，或被赶出来，三十多年再也没有离开过那个行李箱。

六四那代人背出来的行李箱有多沉重？去年六四之际，有自称搞民主的人士，把美国大学校园的哈马斯拉拉队跟天安门的抗议学生相提并论，甚至暗示美国警察像解放军一样镇压学生。这是一群苟活者对在六四镇压中失去生命的年轻人的羞辱。向往自由的人用抗议争取民主，像天安门的学生；独裁势力发动有用的蠢货抗议破坏民主——中华民国在大陆被推翻就有这类有用的蠢货的功劳。这是基本的是非。

但有些六四的苟活者不辨是非，突破不了低矮的认知天花板，举旗抗议成了他们的巴甫洛夫铃铛，只要听到抗议的铃声，不管是争取民主还是破坏民主，他们就会兴奋地往前凑。可悲又可怜。

疫情最后一年的四月，崔健在网上开演唱会，说该变的都没变，不该变的都变了。平日大家看到的是变了的那些，产生出全都已经改变了的错觉。但决定人群命运的那些东西一点都没变，野蛮的照旧野蛮，疯狂的照旧疯狂，而且野蛮和疯狂的能量更大了，不只是制度层面，而且是人群的精神世界层面。

三十六年中，国民大都被暴力和利益驯化了，屠民治国成了上上下下的信仰。他们或者主动选择相信，或者半推半就地相信，高压加镇压才能发展经济，只要生活水平能提高，其他都无所谓，只要自己的生活水平能提高，牺牲一下其他人的利益、健康、未来、生命……都是可以接受的。

1990年代的最初两年，6月3日夜，北大宿舍楼的酒瓶子就像冰雹一样密集地落在楼下的石棉瓦自行车棚上。大约是第三年，突然就平静了。那代学生离开了校园，驯化的一代成了新生主流。往事随流水，转眼三十六年。

人群被暴力和利益驯化之后，社会就变成一座疯人院。

一位国内做出版的朋友说，我们好象全都被这几十年的“奇迹”驯化了，说出了大部分中国人精神世界的现状。看他们相信的东西：弱肉强食、权利不能当饭吃、武统台湾、欧洲完了、美国养懒人、民主没效率、东升西降……这是个奇异的精神世界梦幻组合。

很多年前，每逢春夏之交，北大校园就盛开着结不出果实的红罂粟，绰号“虞美人”。中国曾发生的每一个争取民主的事件，都是刚开始就结束了，生生死死，花开花落，像结不出鸦片的罂粟。

说这些话会被认为是对六四先烈不尊重。那个悲惨的夜晚，很多年轻人失去了生命，很多父母失去了孩子，包括我十分敬佩的蒋培坤、丁子霖老师和他们17岁的儿子。后来在北京见到蒋老师，他长发及肩，时已丧子三年。如果儿子还在，都上大学二年级了。三十几年间，每想及此，沉抑悲痛如初。对于蒋老师夫妇和我不相识的失去孩子的父母，对于那个夜晚被屠杀的年轻人，我只有敬仰和痛惜，没有一字微词。

死者长已矣，但我们有幸或不幸能苟且偷生，总要对自己诚实一点，才对得起那些死难者和他们的亲人。对自己诚实，就免不了面对真实的自己。不管当初的理想多么美好，内心多么勇敢，冲动多么迫切，三十多年后， 我们这代人——就是已经熬成大妈大叔的这一代，除了数钱算计，头脑清楚的不是太多，而且很多头脑清楚的人选择做沉默的少数。

欺骗自己往往是欺骗别人的预演。作为写作者，一个基本的从业道德就是要诚实，尤其是对自己诚实，不说自己不相信的话。该被误解就被误解，该被诅咒就被诅咒，竹杖芒鞋，一蓑烟雨，随他去吧。

我们年轻的时候，改变的风吹遍了世界，吹到中国，发生了六四。三十六年后，蓦然回首，改变的风吹过之后，自由在有些地方生根发芽，长成了大树，在有些地方，像盐碱地的庄稼，半死不活；在那个叫中国的时空，却是连种子都死了。

该变的都没变，不该变的都他妈变了。这么大一个国家，十几亿人，不是离现代文明越来越近，而是成了反现代文明的堡垒。还有脸喊什么5000年文明，还有脸喊什么民族复兴。一些六四的苟活者成了川粉，有些成了哈马斯啦啦队的啦啦队，还有脸喊什么自由，喊什么民主——借用崔健开音乐会时的那句京骂：“去你大爷的！” 别逗了。

对于个人来讲，六四可以是人生的天花板，年轻时经历一次，轰轰烈烈，不枉此生，值得每年纪念，怀旧一生。但作为一个人群争取民主的诉求，作为一个国家的民主愿景，如果六四成了天花板，民主在那个叫中国的时空就没有希望了。

乐见更多具有民主诉求的年轻人把六四作为上一代人铺下的地板，这才是希望。不是一蹴而就改变现状的希望——那是异想天开，而是从自己做起，一点一滴改变贫瘠的民主土壤的希望。毕竟，抗争的意义，不在于它能立即改变现状，而在于抗争者在立志做人，做主宰自己命运的人。这样的人多了，独裁就难以为继，民主才有希望。

莫路狂花，让女性在旷野中围炉夜话。这是一个放学以后创造的崭新的专题，由莫不谷和霸王花来进行的关于女性自我解放可能性的坦诚对话。我们本来以为这个专题只会有且只有一期，万万没想到它在本年度迎来了第二期。

这期也来得非常突然，源于我和霸王花进行了好几个小时的电话通话，我偶然发现了霸王花一个很关键的财务漏洞，好几张信用卡一直在被持续地扣钱，以及霸王花应对日常生活所有关键问题可怕的冗余和混乱：持有着12张银行卡，每天带着2个手机和两个充电宝，很多时候什么也没干就累了，记忆力持续衰退，生命的活力被消耗殆尽，遇到很多问题只能选择逃避，常常因为微小的问题把自己逼入绝境……我们在电话聊天嘻嘻哈哈的惊呼中，不断发现很多问题的连锁效应和根本原因。

因此在已经电话聊天好几个小时后，我们决定就这个问题录一期播客，以更坦诚的态度去深入畅谈这些表象背后的根源，和解决它们的答案。在毫无准备的情况下，我们发现这期播客因为对真问题进行了真面对，所以随着聊天的深入，它带来的震撼和aha moment也像菜刀砍电线一样：一路火花带闪电。

之前面对霸王花呈现的这些种种问题和混乱，发生的当下，我们都以一种“天选综艺人”的幽默搞笑的态度来反应，也是我们几个小时电话的主旋律。而王小波有一句话是：“任何一种负面的生活都能产生很多乱七八糟的细节 , 使它变得蛮有趣的，人就在这种有趣中沉沦下去 , 从根本上忘记了这种生活需要改进”。在几个小时嘻嘻哈哈的电话结束，我们打开录音，决定录一期播客，就是想要改进这种生活。所以这期播客本来还有一个备选标题：《莫路狂花2：给混乱开刀，给痛苦施药，给自我一次自救的真疗效》。

在这期播客录制的过程中，随着对问题根源的探寻和enligtenment的获得，霸王花当听到自己所有痛苦真正的原因的时候，觉得获得了真正的纾解：终于知道自己为什么那么痛苦了，为什么生活还在继续，但是自己就是不能开心和舒服起来的原因了。而在录完这期播客好久之后，霸王花从同样糊涂混乱逃避生活的朋友身上看到了堪称惨痛的教训，立刻发消息和我说：“这期播客实在太重要了。糊涂糊弄的人生不会舒服，只会上一个又一个当，进入一个又一个可怕的坑”。

这期深入真问题，找到真根源，探索真答案的播客将近三个小时，我们希望能通过这期播客给大家带来非常多的安慰，提供很多的启发并激发真正的疏解和惊人的解放。我也希望大家能通过这期播客找到生命给自己的那个礼物，通过它来完成自救。从虚无，混乱，各种深坑和骗局，生存危机和存在主义危机的虚无中完成真正的自救，建立起对自己的爱意和敬意。

说明：本期播客设置10元付费门槛，国内在网易云音乐首发上线，海外同步上线Newsletter、Spotify（Newsletter和Spotify是月度会员解锁所有付费播客和文章的方式），两周后6月14号在爱发电和喜马拉雅上线。

【播客音频】

网易云音乐：http://163cn.tv/FpQ0eVP

Newsletter：https://open.substack.com/pub/afterschool2021/p/2-f98?r=pilpv&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Upgrade subscription

Spotify：https://creators.spotify.com/pod/show/afterschool2021/episodes/2-e33d0vj

想要单期解锁，可以在网易云解锁收听，Newsletter和Spotify付费订阅会员不仅可以直接收听本期付费播客，还可以解锁所有付费播客和文章。

爱发电链接：https://afdian.com/p/2c77401e3a6d11f0a5d852540025c377（6月14号凌晨上线）

喜马拉雅链接：https://www.ximalaya.com/sound/859029300（6月14号凌晨上线）

（本期封面为莫不谷使用Canva设计，设计理念是：“这期封面我搞了一个纷繁复杂的视觉效果，点题‘混乱’，同时里面的元素：手术刀（打开谜团和真正的内心），拥抱，insight等等也都对应着这期播客的主题”）

【Timeline】

06:39 意外发现霸王花财务管理漏洞：人怎么能如此荒谬糊弄混乱又有综艺效果？

21:00 由财务管理问题的举一反三：你在日常生活中有没有非常核心坚定的人生原则？

29:31 莫不谷：很多人想要找到生命的意义，但是你有回答生命向你提出的问题吗？

33:03 为什么莫不谷建议霸王花少用Chatgpt？很多事情看似是果，但其实是因

44:53 为什么我们这一代又一代的悲剧重复在上演，一直被复刻？

54:49 疫情和封控结束了，大家反而陷入更可怕的境地？一个非常非常残酷的原因

59:37 模糊的正确感会让我们得救，而精准的错误会让我们陷入绝境

67:32 人唯一不应该选择的道路，就是接受别人残酷地对待自己

78:26 莫不谷：我从宗教学习中获得的enlightenment：是什么让个体拥有活力？

81:30 警惕破窗效应：正确的事情究竟是什么？人究竟应该怎样对待环境和自身？

98:11 人最大的主观能动性在于选择环境，选择一个能够滋长理性、诗性的环境

102:28 一个决心爱自己的人，无论人生出现什么问题，她都能够面对和解决

107:23 人如何对自己充满感激和敬意，爱自己究竟要体现在哪里？

109:12 道理我都懂，可是做不到，为什么从假知识到真知识会这么难？

120:00 生命给每个人都赠送了礼物，你收到的生命的礼物是什么？

139:00 莫不谷：我们太容易高估天赋和才华，太容易忽视那些比才华更重要的东西

143:00 这世界每一个活着的人都要面对的命题：生存危机和存在主义危机

157:00 从娱乐圈八卦聊起：播客和各种各样的自媒体如何为女性赋权？什么样的态度面对女性的选择才会更深地解放女性和我们自身？

【播客提及的相关内容】

播客：

放学以后往期播客《莫路狂花今夜不设防：人如何不糊弄和痛恨自己，并找到自己的渴望呢？》

放学以后往期播客《（无码版）42 吐槽八卦大会：今天谁也不能阻止我快乐开怀！》

放学以后往期播客《54 我不要只做世界的承受者，我要对这个世界一顿发起！》

放学以后往期播客《53 在诗意匮乏的时代，我们一起来读读诗》

放学以后往期播客《52 美妙人生的关键呀，让我们一起扭一扭它》

放学以后往期播客《51 我们一起Gap一个冬天，再用这些火焰召唤一下2025年！》

影视：

《人体切割术》《世界上最糟糕的人》《娱圈恶女传|白百何：人生易如反掌》《韩国忠武路老登学》

放学以后小红书《游荡采访创作者101：世界文化遗产小镇画家》

书籍：

《活出生命的意义》；《空洞的心》；《祈祷之海》（推荐其中一个短篇《我学习成为我》）

文章：

莫不谷游荡者文章《Run的800种可能：各国拿永居或入籍的方法一览》注册游荡者即可查看www.youdangzhe.com

霸王花游荡者文章《低成本低难度Run的可能性：西班牙非盈利签证申请DIY全攻略!》《为什么来欧洲生活？到西班牙后惊奇我的事情》《为什么来欧洲生活?西班牙生活成本大揭秘，丰俭由人!》注册游荡者即可查看www.youdangzhe.com

莫不谷游荡者文章《女性的声音是如何让这个时代听见的——播客为女性之声赋权》（Newsletter及爱发电也可查看）

莫不谷游荡者文章《第3章 定位与选题：如何拥有个性还能触发共鸣？》（Newsletter及爱发电也可查看）

莫不谷游荡者文章《莫不谷闹学记（每周连载追番-4月番）》（目前已经更新到5月番）注册游荡者即可查看www.youdangzhe.com

莫不谷爱发电文章《价值投资：在中国或欧洲投资美股美债的原因及方法》（Newsletter及游荡者网站也可查看）

莫不谷爱发电文章《从《财富自由主义》到比特币，自由的上限是我们持续学习的能力》（Newsletter及游荡者网站也可查看）

莫不谷爱发电文章《莫不谷的滔滔生活和金龟换酒》（Newsletter及游荡者网站也可查看）

金钟罩的Newsletter《我在散步时被一道天雷击中，陷入狂喜》《大口吸入春天，为热爱的生活再摸三张牌》

《Spotlight Interview on Writer Wei-Ning Yu (THE DROPOUT, SEVERANCE)》 美国电影学院AFI对编剧魏宁·余的采访，她曾参与《Severance/人体切割术》的创作。

【播客音频】

网易云音乐：http://163cn.tv/FpQ0eVP

Newsletter：https://open.substack.com/pub/afterschool2021/p/2-f98?r=pilpv&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Upgrade subscription

Spotify：https://creators.spotify.com/pod/show/afterschool2021/episodes/2-e33d0vj

想要单期解锁，可以在网易云解锁收听，Newsletter和Spotify付费订阅会员不仅可以直接收听本期付费播客，还可以解锁所有付费播客和文章。

爱发电链接：https://afdian.com/p/2c77401e3a6d11f0a5d852540025c377（6月14号凌晨上线）

喜马拉雅链接：https://www.ximalaya.com/sound/859029300（6月14号凌晨上线）

【为全球华人游荡者提供解决方案的平台】

游荡者（www.youdangzhe.com），注册完成后可免费阅读由莫不谷和霸王花撰写的三篇文章（Run的800种可能、语言攻略和全球签证攻略），目前游荡者平台已更新上线文章分区功能（游荡区、学习区、欢愉区和闲聊搭子区），欢迎大家注册完成后开启内容创作并在游荡者游荡愉快！找到同类！交易自由！手机端用户可把新网址添加桌面，便于日常使用。在使用新网址期间如果有任何注册、支付、退款等需求，欢迎给我们客服邮箱wanderservice2024@outlook.com发送邮件。

放学以后Newsletter《新的一年会好吗？答案在这期播客和这些祝福中》

【延伸信息】

永不失联Newsletter订阅链接：https://afterschool2021.substack.com/（需科 学/上 网）

为全球华人游荡者提供解决方案的平台：游荡者（www.youdangzhe.com）

联系邮箱：afterschool2021@126.com（投稿来信及合作洽谈）

同名YouTube：https://www.youtube.com/@afterschool2021

同名公众号：放学以后after school

小红书：游荡者的日常

欢迎并感谢大家在爱发电为我们的创作发电：https://afdian.com/a/afterschool

片头曲：Susan Sarandon As ALouise

片尾曲：Wild Night

播客封面：莫不谷用Canva制作

放学以后表情包：微信表情包搜索“放学以后”，感谢萝卜特创作。

播客收听平台：

【国内】苹果播客（请科学/上网）、爱发电、汽水儿、荔枝、网易云、小宇宙、喜马拉雅、QQ音乐；

【海外】Spotify、Newsletter（Newsletter和Spotify是月度会员解锁所有付费播客和文章的方式）、Apple podcast、Google podcast、Snipd、Overcast、Castbox、Amazon Music、Pocket Casts、Stitcher、Radio Public、Wordpress

莫路狂花，让女性在旷野中围炉夜话。这是一个放学以后创造的崭新的专题，由莫不谷和霸王花来进行的关于女性自我解放可能性的坦诚对话。我们本来以为这个专题只会有且只有一期，万万没想到它在本年度迎来了第二期。

这期也来得非常突然，源于我和霸王花进行了好几个小时的电话通话，我偶然发现了霸王花一个很关键的财务漏洞，好几张信用卡一直在被持续地扣钱，以及霸王花应对日常生活所有关键问题可怕的冗余和混乱：持有着12张银行卡，每天带着2个手机和两个充电宝，很多时候什么也没干就累了，记忆力持续衰退，生命的活力被消耗殆尽，遇到很多问题只能选择逃避，常常因为微小的问题把自己逼入绝境……我们在电话聊天嘻嘻哈哈的惊呼中，不断发现很多问题的连锁效应和根本原因。

因此在已经电话聊天好几个小时后，我们决定就这个问题录一期播客，以更坦诚的态度去深…

Read more

We’re having a meetup on Monday in DC. RSVP here!

This edition is brought to you by Pelanor, the AI-powered FinOps startup letting companies make sense of their cloud spend.

Founded by alumni of Israel’s elite cyber intelligence unit, Pelanor untangles the chaos behind nine-figure AWS and Azure bills. Most companies have no idea how their compute spend ties into business outcomes. Pelanor solves this by building a dynamic graph of your cloud environment—tracking which microservices talk to which databases and how AI workloads are actually being used. Even non-technical team members can ask complex questions in plain English and get straight answers:

• “What’s driving our OpenAI bill?”

• “Which services are talking to unused databases?”

• “Where can we save?”

Reach out to founder Matan Mates on LinkedIn or email him directly at matan@pelanor.io.

Oskar Galeev is a PhD researcher at Johns Hopkins SAIS working on AI history and the politics of the US-China tech race. Previously, he was a Yenching Scholar at Peking University and a Winter Fellow at the Centre for the Governance of AI.

Girish Sastry is an independent AI policy researcher specializing in technical aspects of AI governance. Before this, he spent 4 years at OpenAI where he worked on research related to AI misuse, compute governance, and capability evaluations.

Modern computing export restrictions have deep historical roots that extend far beyond recent headlines. Today’s AI chip denials represent just the latest chapter in a decades-old American strategy of technological containment — one that began long before the Biden administration’s AI diffusion framework or the Trump-era Huawei sanctions. This approach of strategically limiting adversaries’ computing capabilities traces back to the earliest days of the Cold War, when computational power first emerged as a geopolitical asset.

The Coordinating Committee for Multilateral Export Controls (CoCom), established in 1949, was America’s immediate response to a Soviet Union newly armed with nuclear capabilities. This American-led multilateral alliance, led by the United States, established a comprehensive technology embargo system. CoCom controlled the export of strategic goods and technologies — particularly computing equipment — that could enhance Soviet military and economic power. More than just a regulatory body, it represented a coordinated Western effort to maintain technological superiority throughout the Cold War. For 45 years, CoCom carefully managed what computing technologies could cross the Iron Curtain. It also unintentionally taught the Soviets to excel at smuggling and reverse-engineering computers.

Chinese and American think tanks alike have drawn parallels between CoCom and the modern Sino-American semiconductor competition. Guided by historical memory, Chinese policy conversations tend to focus on ChinCom, the specialized control system for the PRC. ChinCom had different goals than CoCom — as one CIA memo explained in 1952, “The problem of security controls respecting the China area is significantly different from that respecting the rest of the Soviet bloc”. While CoCom focused on long-term strategic competition with another superpower, ChinCom implied a much stricter punitive embargo as a direct reaction to “Chinese aggression in Korea.” The USA’s current view of China is similar to the perception of the USSR in the 1950s, but back then, the young PRC was not viewed as a geopolitical competitor or a party in the tech Cold War.

CoCom is the single longest case study on the impact of high-tech export controls in geopolitical competition — and for all intents and purposes, it worked. So what lessons can be learned from the history of the CoCom? And how are they likely to guide both the US and Chinese approaches to limiting each others’ compute capabilities?

New Cold War, Same Challenges

The first and most important parallel is the difficulty of enforcement. We often forget just how much the Cold War policy community complained about CoCom simply “not working.” Even though it was a multilateral regime, like today’s US semiconductor controls, it was not based on any treaty or binding agreement. And out of 17 member states of CoCom, only the US imposed re-export controls.

Sample of smuggled mainframes, industrial control computers, and supercomputers (in bold) during the CoCom era:

Throughout its history, CoCom’s effectiveness was reduced by overt non-compliance, differences between individual member nations, the overall secretive regime, and the financial bottom line of tech exporters. The Eastern Bloc was a giant developing market, and tech companies didn’t want to lose access. The export lists themselves often did not make any strategic sense — in the early years, CoCom even prohibited exports of items like typewriters (for an overview of control lists, check the collection by Sam Weiss Evans). But when it came to truly strategic technologies, control evasions were simply overlooked in most cases. Despite the Bruchhausen Semiconductor Smuggling in 1977-1980 and VAX Supercomputer Diversion of 1983, which directly boosted Soviet missile and aerospace design, only one control evasion case led to a strong policy response from the US. That was the Toshiba-Kongsberg case, the main geopolitical tech scandal of the Cold War.

In the early 1980s, the Soviet KGB received computer numerical controls (CNCs) clearly restricted under CoCom from the Japanese Toshiba Machine and Itochu Corporation as well as the Norwegian state-owned Kongsberg Vaapenfabrik Company. The suppliers were even updating and fine-tuning software on the Soviet facilities between 1982 and 1984. For export license purposes, the Norwegian Trade Ministry listed the items as spare parts for a civilian facility, while Japan’s MITI (Ministry of International Trade and Industry) registered the machines as older models not included in CoCom lists. The result was twofold — Soviet submarines got reduced acoustic signatures, increasing their stealth against American anti-submarine warfare capabilities, while the White House went on a punitive campaign, imposing sanctions on the Norwegian supplier and fully banning imports of Toshiba products. These repercussions made the Toshiba-Kongsberg case unique. The harsh consequences were partly motivated by the rising narrative of tech competition with Japan. It was not only the USSR that the American security establishment was worried about, but also about winning the competition against the Japanese tech sector. But the shamelessness of this case also motivated consequences — no effort was made to send the equipment through third countries, like everyone else did.

Soviet techniques for bypassing CoCom1

Despite extensive multilateral coordination under CoCom, enforcement remained its Achilles heel — a challenge that persists in today’s semiconductor controls. As the Heritage Foundation lamented in 1983: «Terrible Fact. To be sure, Washington would not have to control the re-export of U.S. items from its allies if the allies actually were effectively controlling the re-exports. The problem is that they are not».

On the Soviet side, a critical structural problem of technology competition lay not in acquisition but in efficient diffusion and allocation of restricted computing resources — that is, what happened after restricted tech made it through the blockade. Soviet archives reveal a dysfunctional internal competition that severely limited the effectiveness of their technology transfer efforts. Various intelligence departments, ministries, and industrial enterprises routinely competed for the same technological products, often outbidding each other and creating artificial scarcity for domestic users. The same IBM computer would be simultaneously pursued by a truck manufacturing plant, military production facilities, the state tourism operator Intourist, and even the Soviet Olympic Committee. This fragmentation extended to collection channels themselves, with Soviet industrial ministries frequently requesting identical Western computers through multiple intelligence agencies. Such uncoordinated efforts not only wasted resources, but also critically impaired the USSR’s ability to strategically deploy compute.

This pattern of inefficient resource allocation has modern parallels, such as the inefficient stockpiling of GPUs by Chinese companies and local governments, as covered by ChinaTalk. Chinese policy conversations are increasingly focused on this, arguing that the US-led export control regime should be addressed through “Construction of a Unified National Market” 全国统一大市场建设, a policy aiming to coordinate provincial governments to gain leverage over foreign firms and eliminate local protectionism and administrative monopolies. While there are structural similarities to the supply-side restrictions of the CoCom era, China’s effort to streamline resource allocation means those same control strategies might not yield the same results today.

PRC≠USSR

China’s economic leverage creates a formidable counterweight to future export control efforts. Beijing possesses retaliatory capabilities along the semiconductor supply chain in a way the Soviets never did, especially through critical resources like rare earth minerals. More importantly, China actually has an export market. The Soviet Union’s technological autarky meant it never developed computing giants capable of competing globally. Archival evidence shows that Soviet-made computers rarely crossed even Eastern Bloc borders, let alone captured a share of Western markets. And when transistors first revolutionized telecommunications in the 1950s, CoCom had already been established, reinforcing America’s first-mover advantage. Today, the United States faces the much more complex challenge of building a coalition mid-race.

The second critical difference between the political era of CoCom and that of the 2020s lies in access to human capital. Throughout its technological competition with the United States, the USSR was severely constrained by its limited access to international talent, particularly the tacit knowledge transfer from American computer engineers. While exceptions like Joel Barr and Alfred Sarant enabled rare technological breakthroughs, such cases were anomalies rather than the norm. The Soviet Union could not practically implement anything like the Thousand Talents Program (千人计划). By contrast, Chinese recruitment efforts have successfully targeted specialists from companies like ASML and TSMC alongside top graduates from universities worldwide.

The current competition for global STEM talent contrasts sharply with the CoCom era, when the United States served as the unchallenged center of gravity for international technical expertise. In a world where knowledge transfer often proves more valuable than hardware access alone, human resources can play a substantial role in circumventing export restrictions. The US Foreign Direct Product Rule effectively restricts American talent, but any realistic export control framework must also incorporate talent retention.

Three lessons from CoCom

1. Effective technology denial requires multilateral enforcement.

There may be inherent limitations to a unilateral enforcement of the export control regime. Despite its structural challenges, CoCom at least regularly coordinated the leading technological and manufacturing powers in restricting access to critical technologies for the United States’ strategic adversary — an alignment that does not exist today. Both CSIS and Carnegie suggested that unilateral controls on frontier technologies are ineffective, a point made many times before, including by Eric Hirschhorn, the former head of BIS, who compared it to “damming half the river.”

2. Tracking and verification systems are a prerequisite to effective enforcement.

Where CoCom frequently failed was its inability to track controlled technologies once they left manufacturer facilities, creating enforcement gaps that were systematically exploited through transshipment, diversion, and falsified end-user declarations. Erich Grunewald and Michael Aird proposed a chip registry idea, noting that “a key problem for AI chip export enforcement is that BIS does not know where exported AI chips are.” Without visibility into the movement and end-use of restricted technologies, even the most comprehensive control frameworks will ultimately collapse under their own enforcement limitations. Implementing something like the chip registry would significantly increase the odds of technology denial achieving the desired effect.

3. CoCom operated under nearly ideal institutional conditions, and still could not be enforced perfectly.

CoCom was poised for success, operating under an ideologically bifurcated international system with established multilateral coordination structures and decades of institutional development. Despite these favorable conditions, CoCom still struggled. Today’s landscape offers none of these advantages to the US — the world today has incomparably more interconnected trade networks, fragmented alliance structures with competing economic interests, and a technologically sophisticated competitor deeply embedded in semiconductor supply chains. If CoCom’s results were mixed even under optimal containment conditions, expectations for current export control effectiveness should probably be tempered.

Is the glass half-full or half-empty?

CoCom suggests that export control regimes have always faced practical challenges. Perhaps the most significant challenge is reliable enforcement. To the extent that CoCom was successful, it was through the cooperation of other countries in a more multilateral regime. Today, deteriorating alliance cohesion and tariff wars could undermine the verification and enforcement mechanisms necessary for effective controls. As the intelligence leaks and scandals of the Cold War era show, crucial partners like South Korea and the Netherlands did not always have incentives to comply with the US-led tech regime. Today, these partners simultaneously host critical semiconductor infrastructure while also maintaining substantial economic ties with China. In fact, despite export restrictions, TSMC may have already manufactured millions of controlled high-end AI chips for Huawei.

What does this mean for US AI policy? First, policies that improve verification systems and bolster multilateral institution building would be very effective. Verification and compliance measures should address not only chip smuggling but also other potential circumvention methods such as the use and operation of data centers in countries like Malaysia, remote access to large amounts of compute through cloud services, use of TSMC manufacturing, and other avenues to skirt US export controls.

Second, as policymakers navigate the current “chip war,” they should recognize that export controls represent just one element of a comprehensive technology strategy — one that must be balanced against both economic interdependence and the reality that innovation often flourishes in response to constraints. The most sustainable technological advantage will likely come not from restriction alone, but from accelerating domestic innovation while selectively managing the most critical chokepoints in the AI supply chain.

在北半球的夏天马上就要到来之际，我们来发布一些重磅更新！

一、播客更新

5月31日（也就是东八区今晚0点之后）将会在spotify，newsletter和网易云更新发布《莫路狂花2：如何对自己充满爱意和敬意，免于混乱逃避低活力？》。爱发电和喜马拉雅将于6月14号更新，以下是收听链接！希望给大家的痛苦，混乱和倦怠都带来巨大的纾解和解放！

（今晚零点过后更新）网易云链接：http://163cn.tv/FpQ0eVP

（今晚零点过后更新）付费版Newsletter链接：https://open.substack.com/pub/afterschool2021/p/2-f98?r=pilpv&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

（今晚零点过后更新）spotify链接：https://creators.spotify.com/pod/show/afterschool2021/episodes/2-e33d0vj

（6.14更新）爱发电链接：

https://afdian.com/p/2c77401e3a6d11f0a5d852540025c377

（6.14更新）喜马拉雅链接：https://www.ximalaya.com/sound/859029300

此外还将有3-5期**“终身学习”系列播客**做成专辑系列更新，我们已经录完一期，其余几期将会在6月会发布征稿，敬请期待！

二、游荡者平台

1.付费咨询功能上线

开发了很久终于这个新功能终于要上线了！目前是先上线了图文付费咨询的功能，之后还会上线视频语音付费咨询的功能。各位游荡者平台的创作者们可以查看游荡者平台右上角的消息，看看自己是否已经满足开通付费咨询的条件。

倘若已经收到通知满足条件就可以点击“下一步”设置自己的接收付费图文咨询的单价啦！

各位游荡者朋友们也可以在阅读首页文章时向自己感兴趣可以解答自己问题的创作者进行提问。

以我为例，在游荡者平台点击你想咨询的创作者的头像，即可看到“付费咨询”的按钮（倘若对方已经符合条件，完成设置，开通该功能够），点击该按钮就可以进行咨询，共有一次提问和一次追问的机会。

一个好的提问的公式是：

“我是谁（不用暴露隐私） + 我遇到什么问题 + 在什么背景下 + 我想要什么帮助。”

希望提问的朋友们都能利用这个公式提出自己真正想提的“真问题”，不辜负自己花的钱，也帮助创作者更切中肯綮地回答你和帮助你！

目前游荡者的开发安排是：

（1）图文付费咨询（已完成）-（2）视频语音咨询（待完成）-（3）问答专区搭建（待开发，图文付费咨询的问题可由提问者选择一键分享到该专区，其它游荡者付费解锁该问答后，提问者和创作者将持续获得收益；同时也可在该问答专区进行直接无定向提问）

（4）社区广场搭建（待开发）：含活动发起专区（各种workshop和大家主动发起的活动都在这里进行）；aha moment专区（能给人情绪带来最正向感受的内容）；同游搭子专区（给自己的游荡和学习以及各种想要同游者的活动找搭子）

这是我们初步的安排，针对此大家倘若有建议和期待也可以在评论区给我们留言，我们还计划在7/8月份开一次线上游荡者大会，和大家connenct起来，和大家聊一聊大家对这个社区的期待，建议和各种反馈！

2.导游区开始“每周一游”更新！

（“导游区”就在首页的最上方的居右位置）

各位全球的游荡者朋友们！大家好，来自“游荡者号”游轮的一则重要通知：从本周起，游荡者平台线上游轮将每周启航，我们作为“导游区”成员（莫不谷、霸王花木兰，粽子、金钟罩）将每周轮值担任船长，在国内时间每周四更新一篇【每周一游】，分享内容包括但不限于各种花花万物和生命体验的推荐和避雷！在临近周末打工人即将解放的周四，和大家一起驶向一些海域打发时间，度过无聊，对抗虚无！欢迎各位游荡者每周四定期登船！

本期游荡者游轮【每周一游】轮值船长是莫不谷，祝大家游览愉快，在活力不高的时间吃一些不耗费精力，能抚慰自己的细糠。我已经于昨日周四第一次启航！

希望大家能在这即将要和未来要更新的播客，已经更新和正在开发本年内会更新的游荡者的新功能中，游荡愉快！

Read more

本篇闹学记写于2025年5月23号，我刚考完一个艰难的试的当晚。《莫不谷闹学记》系列正在游荡者平台(www.youdangzhe.com）的莫不谷首页独家免费更新，已经更新了4月番和5月番的好多篇和好多字。看完这篇对更多内容感兴趣的朋友也可前往兔子洞一跃而入！

我今天刚考完了一场非常艰辛的荷兰语考试，我考了8.8分，比我更努力这几天天天复习到凌晨三点多的同学考了4.4分，TA一直在问我怎么会这样？是不是因为自己是个愚蠢的人？我在一直安慰它和消解它的自我攻击的时候，找到了TA问的这个问题真正的的答案。

霸王花听到我同学持续问我的这个问题，想起了她在剪辑《莫路狂花2：如何对自己充满爱意和敬意，免于混乱逃避低活力？》(5月31号本周六凌晨spotify和newsletter，以及国内平台网易云上线，两周后6月14号爱发电和喜马拉雅上线）时，发现她想问的问题，和我拼命努力却考了4.4分的同学一样。我说你的是什么问题呀？

她回复说：“我一直在问，怎么有确定的人生原则和念头，且不会摇摆和变动，就像《活出生命的意义》这本书里作者找到了答案，你读了这本书，就变得与众不同，我读了，可以与众不同三天，我就在问，怎么能长时间与众不同？”

就像我同学问我是TA不够努力和聪明，才考了4.4分吗这个问题一样，对于霸王花问我的这个问题，我都觉得是假问题。

因为TA明明知道自己很努力，且很聪明，考试结果的差异完全不来源于TA攻击自己的这些原因。

而对霸王花这个问题，首先想要“与众不同”我是我俩任何一个人的真正渴求。同时“与众不同”也是一个无法追求的事情，一个人一旦刻意追求“与众不同”，就会刻奇虚伪矫饰堕入比与众相同更可怕的造作。所以对霸王花这个问题真正的理解究竟是什么？很多时候唯有正确理解了问题，才有可能找到真正的答案。所以我想花一篇文章的篇幅仔细讲述一下两个数字差距背后的答案，和假问题背后的真解答。

在开始给出答案之前，我必须介绍一下我们的考试。我从去年年底开始正式高强度上课学习荷兰语，和来自世界各国的难民朋友们一起线下上课学习。因为难民朋友们有快速融入开始用荷兰语上学，生活和工作的需求，所以可想而知我们的强度有多大，简直是秋风扫落叶，摧枯拉巨朽。

今年3月我已经高分通过了0-A1-A2的课程（天知道我有多努力用心激情澎湃），4月开始了荷兰语B1的课程。0-A2上课的内容和强度已经够吓人了，来到了B1简直是每天巨浪拍打在脸上。A2还是荷兰幼儿园小孩的水平（能达到母语者幼儿园水平也很艰难），B1已经来到了高中生。高中生意味着什么：它们已经可以使用成语，并开始谈论各种抽象，形而上学的东西！

所以我们B1的课文又长又复杂，之前还能正常说话表达：他了解了各项政策。到了B1就是：他可以高屋建瓴地把握各项政策。对于刚学中文没多久的人在脑子里把“高屋建瓴”四个字分别记住且黏在一起真的是难之又难。而几十篇课文通篇充斥着这些复杂的东西，我家地板上全是我掉落的头发，感觉上面附着着我死亡的脑细胞的冤魂。

今天我们进行了第二次考试，要考23篇巨长课文的填空和听写，无比复杂的课文里随机挖一个洞让你填，真的难如徒步登山。我们也没有任何额外的时间复习，这一周每天都还要为第二天的课自学准备，没有一天空隙。所以周三的课，大家周二晚上都在老师不在的群聊里说，决定明天生病，在家复习考试。当我学到晚上12点多打开群聊时，真的眼前一黑：因为群里大家都在说不去上课，有一个人问所有人都不去吗，另一个我压根不熟悉的同学说：应该除了Echo都不去上课。

不知道为什么，我树立了一个所有人都不去上课我也会去上课的形象。当然我也果然去了，当天陆续还来了另外两三个同学，我觉得这么小班的教学也甚好，能和老师多做问答练习。

其它时间我每天都在学到晚上12点以后，来应对周五这场考试，中间还帮我之前考得不好的同学解答问题，把我总结到的一些规律分享到群里。文章开头提到的这位同学之前在A2第三场考试没有及格，在第四场考试的时候我和它一起花了一天的时间一起梳理，分享了我的学习方法和策略，它第四场考试就通过了。所以今天这场考试前我俩学到崩溃了就互相安慰，还一起讨论各种知识点。它比我学得更努力，因为我常常第二天早上看到它凌晨三点多给我发的关于学习的消息。

结果今天考试结果出来，我考了8.8分（应该是全班最高分），它考了4.4分（荷兰是5.5分及格）。它就完全崩溃了，因为它如此努力，做了所有能做的一切，我发了好多信息安慰它，它还一直说一定是因为自己就是个蠢货。

在我一遍一遍安慰无效的时候，我引用了我俩之前的聊天记录，我说我找到了结果差异的根本原因。

在前两天我俩讨论一个知识点的时候，

（看不懂别害怕，下面有翻译）

它说：“Echo，现在不是把一切搞懂的时刻，是为考试而学的时刻”。

我说：“只有我搞懂了，我才有可能通过考试。”

的确，在几乎没有什么时间复习的时候，看起来已经火烧眉毛了，不该再花时间去搞明白课文里的各种细节问题，而应该是一股脑把文章反复读和背，才能应对考试。

但是我就是放不过自己，我搞不明白的东西我就是不想记住，所以即使火已经烧到了眉毛，我还要趁着火苗的微光照一照问题，把不清楚的搞清楚。

但通过荷兰语一次次的考试，比我更努力，比我花时间更久的同学得到的结果却让它怀疑自己，而我却一次次不断地验证：学习（包含考试）就是一场搞明白的过程，搞得明白才可能真的学得好和考得好。

今天考试前老师还问我们有没有问题要问，大家都没有问题，就我举起了手说：“我有一堆问题要问！”。然后打开我的手机一条条读我在复习课文的时候发现的课文里前后不一致，逻辑有问题，语法不正确的地方。

我每问一个问题老师都很震惊，说：“你的眼睛也太好了（其实我高度近视），你怎么发现这些的，你问的问题也太好了，大家有没有能解答的？”

然后大家和老师都解答不了的，老师和我说书里这部分的确是stupid，如果考试考到了你说的地方，你按照你的方式写，一定给你算对。

所以最后23篇文章，我只有时间精细复习了12篇，剩余11篇呢？我虽然前一天晚上几乎没睡，但是还是一大清早起来了爬起来去学校把这11篇全部通读了一遍，又把23篇我之前所有犯错的地方重看了一遍，虽然我只睡了两个小时，但是上课和考试的时候我都精神奕奕，充满信心同时还非常平静，因为我知道：我已经尽了自己能尽的努力了，而且我也满足了自己一定要搞明白的欲望。我没有什么可担忧和可不满足的了。

即使考得不好我也不会自我攻击，觉得自己愚蠢或者不够努力。更何况我发现，有了“一定要搞明白”这种劲头，你很难fail任何学习和考试。

而“为考试而学”的态度和方法，首先会fail学习，其次很大概率也会fail考试。

解答了第一个问题，我们来到霸王花的那个问题：

：“我一直在问，怎么有确定的人生原则和念头，且不会摇摆和变动，就像《活出生命的意义》这本书里作者找到了答案，你读了这本书，就变得与众不同，我读了，可以与众不同三天，我就在问，怎么能长时间与众不同？”

我觉得首先这是一个假问题，因为霸王花和我都没有“与众不同”的真心渴求，我们不能假装我们有这个渴求，来去寻找答案。但是通过《莫路狂花2》这期播客的长谈和无数个电话以及当面深夜长谈，我理解的霸王花的真问题是：

为什么你学到了一些好的东西，就能真心相信它，把它变成自己恒久的人生原则，不因为各种大象和小蚂蚁动摇？然后还能把这些人生原则真的应用于生活？

为什么我学到了很多“好道理”却做不到，即使能被好道理激发3天，后面还是会偃旗息鼓，被自己的逃避和恐惧击溃回到原处。我的学习并未在我身上发挥真正的作用。

在这一点上，霸王花和我那位自我攻击的同学，所问的问题的确是一致的：为什么拼命学了，还是无法考好这场试，过好这人生的日复一日？

那我的答案归根结底也是一致的：对于我不明白的问题，我穷尽心力地追问了。所以对于我终于搞明白的答案，我才能深深记住和牢牢坚守。

在我分享完我的答案后，霸王花喜欢从MBTI里找答案。于是她去搜索我的mbti（INTJ）和学习的关键词后发给了我一张截图。

紧跟着这个对我MBTI的解释截图，霸王花说：“从‘但凡排除不掉一个，都难受’中看到了你。而我：知道答案是哪个就够了”。

而我也的确如此，我一定要把4个选项都搞明白才舒服。甚至在上班工作的时候，都一直想搞明白为什么做某件事，如果领导或者客户无法说服我这件事，我的激情就会消弭。这对公司来说不一定是好事，因为并不是一个合格的工具人。但是对我自己，对人类而言，都是好事。不追寻原因，理由和答案合理性的人类，就一定会被工具化。

不仅要知道正确答案为什么是对的，还要知道错误的答案为什么是错的，这会拖慢我为考试复习的时间，但是会加快我真正学习到东西的进度。真正学到了东西就很难考得糟糕。

而对于个人人生那些真正的问题？你有每年和每月都追问吗？我每天都追问。

或许你会觉得追问让人头大，让自己疲乏，无力，停滞不前。

可答案恰恰相反：不追问背后的恐惧和逃避，才会吸食你所有的生命精力和活力。不敢追问自己生活中真正存在的问题，不敢去理解生命对自己提出的问题和挑战，无论看多少书，走多少路，假装为生命的考卷拼命攒积分，都会陷入那所谓的俗语。

俗语能成为俗语，是因为它适配绝大多数人。绝大多数人都在为试卷背答案。

而唯有你，为自己的学习，为自己的人生，去理解问题，排除错误的选项，追寻属于你正确的答案，俗语才无法适配你，老话才无法圈禁你。

以及，以及，背别人的答案，一定会落入俗套的人生啊！（我像千寻一样冲着正在猛吃的妈妈爸爸呐喊！）

Read more

Read more

在“欢乐”打工了一周以后，兔姐、莫甘和犀犀终于能聚在一起、吐槽自己被工作硬控的无奈经历。

👩 作为外来打工女，我们的睡眠、饮食、作息和生活方式好像都逃不脱资本主义的魔爪；我们的性别、种族和文化身份更是让我们在职场举步维艰。

虽然身处三个完全不同的行业——学术界、航空业、金融界——我们却都不禁想问：

为什么女性想要融入职场总是难于登天？为什么我们总被挑剔“不够专业”？为什么女人只能内卷，而男人却总是偷偷升职？为什么月经、孕肚、乃至眼泪都好像是职场的禁忌？

不过，比起吐槽和抱怨，我们更想做的还是积极探索实际的解决办法：在工作带来的高压之下，我们有可能找回生活的主动权吗？职场女性究竟怎样才能调节心态、摆脱工作带来的窒息感？我们，能逃出生天、重获自由吗？

本期播客已上传至小宇宙、Spotify、苹果播客、喜马拉雅以及荔枝fm。大家在各个平台搜索「陌生女人帮」就能找到我们啦。欢迎姐妹们订阅和收听，期待在评论区看到大家的留言！！

🎙️ 本期主播：

兔姐（陌生女人1号）

莫甘（陌生女人2号）

林犀（陌生女人4号）

📒 Shownotes

被工作“硬控”的女人们

03:18 被工作反复折磨后，我梦见自己错过早上六点的会议

09:04 我和盒饭里的肉有本质区别吗？

10:42 做资本主义的螺丝钉和我的女性主义信仰背道而驰

15:46 白天给公司打工，晚上给各种APP打工

在职场，我们因多重身份遭受过哪些歧视？

18:51 在香港工作，学不会广东话就是你的问题

21:25 文化背景差异让我沦为职场“边缘人”

26:03 父权制下，最理想的打工人画像是「精英白男」

27:06 我经常是会议室里唯一的女性，更是唯一的亚裔女性

28:51 学系开会时，我好像误入了“兄弟会”

33:08 职场容不下身体，更容不下女性的身体

37:17 怀孕、生病、来月经就像“犯罪”

41:27 所谓“专业”（professional）是客观中立的吗？

44:39 职场不相信眼泪？💧

重重压迫之下，我们怎样才能短暂解放自己？

47:52 我不再赋予工作具体的意义

52:19 用成长型思维和多元化的标准衡量自我价值

54:37 警惕老板那些包装成「正反馈」的PUA

58:42 大家都说“职场没有真正的朋友”，我却偏要和同事交朋友

为了逃出资本主义体系，我们该如何关照自己？

01:01:37 户外跑步是我关怀自己的小小仪式

01:03:24 去山野、森林徒步是最贴近自然的冥想

01:08:23 Self-care的本质是Others-care：有时候，关怀她人就是关怀自己

01:13:42 快乐本身也可以是一种反抗

01:16:05 建立双向、互惠的关系 ≠ 利「他」

🎵 片尾曲：

Labor by Paris Paloma

👧 关于我们

陌生女人帮（Let it Out）是一档聚焦女性生命体验的谈话节目。你可以在微信搜索「两个陌生女人的来信」，或在小红书搜索「DearSisters」关注我们，还可以在苹果播客、Spotify、喜马拉雅听书等平台收听我们的节目。

商务合作/投稿等：dearsisters2022@gmail.com

P.s. 我们已经把本期博客上传至小宇宙、Spotify、苹果播客、喜马拉雅以及荔枝fm。大家可以点击上文链接、或在各个平台搜索「陌生女人帮」就能找到我们啦。欢迎姐妹们订阅和收听，期待在评论区看到大家的留言！！

We’re looking for part-time help with the podcast and newsletter! Spend ~10 hrs/week editing shows and putting together enhanced transcripts starting at $35/hr. All you need is enthusiasm for our content—no prior audio experience required (we use this software called Descript which is easy to learn and a pleasure to use).

Respond to this email if you’re interested with a little background on yourself.

This edition is brought to you by Pelanor, the AI-powered FinOps startup letting companies make sense of their cloud spend.

Founded by alumni of Israel’s elite cyber intelligence unit, Pelanor untangles the chaos behind nine-figure AWS and Azure bills. Most companies have no idea how their compute spend ties into business outcomes. That’s because the people managing infrastructure and the people using it rarely talk, let alone share ownership.

Pelanor solves this by building a dynamic graph of your cloud environment—tracking which microservices talk to which databases and how AI workloads are actually being used. Even non-technical team members can ask complex questions in plain English and get straight answers:

• “What’s driving our OpenAI bill?”

• “Which services are talking to unused databases?”

• “Where can we save?”

Reach out to founder Matan Mates on LinkedIn or email him directly at matan@pelanor.io.

As Meta aggressively consolidated the US market for augmented reality glasses, it scared off what could have been a competitive AR ecosystem in the West. Only now with Google’s recent AR announcements at I/O is a second serious player back in the game. In the words of Leap Motion and Midjourney founder David Holz:

“VR and AR really needed 12 companies basically making products… [I]n two generations of 12 companies, we would have been way closer to something that was really compelling for everyone. But instead, we got one product, maybe two.”

In China, however, there has been no such consolidation, and a “Hundred Lens War” (百镜大战) has instead produced a vibrant AR ecosystem where small startups, rather than tech giants, lead. But does it matter that there are no American analogues to China’s “Five Little Dragons” of augmented reality (AR眼镜五小龙)?

Skepticism toward AR glasses is understandable (especially after the highly publicized failure of Google Glass), but the premature consolidation of the US market could have dramatic consequences one day. Real-time translation via a wearable product could be game-changing in linguistically diverse places like India; some consumers could be compelled by the prospect of using AR glasses as a real-time conversation guide; and maybe hyper-immersive Wii sports will prove to be even more fun than the low-tech version we enjoy today. The point is, we can’t really sit here in 2025 and say with certainty that AR glasses won’t have any compelling use cases.

If manufacturers eventually overcome technical challenges and get the public on board, AR glasses could generate tons of real-world audio and video data valuable for training AI, much like humanoid robots. This route to profitability could help buoy AR investment, even if the public isn’t sold on AR products yet. That could also partly explain why Xi tried on some AR glasses during his recent tour of Shanghai’s AI ecosystem.

Today’s article will explore China’s market for AR glasses and introduce China’s Five Little AR Dragons. It will also include my personal review of a Chinese-made AR headset that I was able to try in Taipei.

Terminology and the State of the Art

Augmented reality (AR) is distinct from virtual reality (VR) — AR headsets are designed to overlay digital features onto the user’s interaction with the real world, and can theoretically be worn for extended periods. VR products are designed for shorter periods of fully immersive use (Apple’s Vision Pro, for example, does both VR and passthrough AR). The umbrella term for both AR and VR is extended reality, or XR for short.

Chinese XR glasses manufacturers rely on microdisplays, primarily micro-LED and micro-OLED displays. While traditional OLED displays use a backplane made of glass, the pixels of micro-OLED displays are etched directly onto a silicon wafer (hence their alternative name, OLED on Silicon, or OLEDoS). Thanks to supply chains created for the Apple Vision Pro, the cost of producing these displays has dropped rapidly in China since 2023. In March, Chinese LEDoS manufacturer JBD cinched a deal with Meta to become the exclusive supplier of LEDoS displays for Meta’s Orion prototype.

Another key component is the waveguide. A waveguide is a transparent optical component that guides light from the display to the user’s eyes while allowing them to see the real world. The waveguide makes it possible to overlay digital content onto physical environments. As is the case with microdisplays, China’s leading suppliers of waveguides are based in Shanghai.

Five Little Dragons

What can these glasses do today? Traditionally, the industry has emphasized entertainment features (e.g., gaming, streaming movies and music, and shooting photos and videos) while touting the potential future benefits in education, medical care, and delivery logistics.

But after Meta announced new AI features for their Ray-Ban smart glasses in April 2024, China’s AR companies have been eager to capitalize on the “AI+AR” hype. According to Li Hongwei 李宏伟, CEO of the AR dragon RayNeo:

“Smartphones do not have the display features or capabilities of spatial perception interaction that AR glasses have. AI+AR glasses encompass three categories: mobile phone applications, AI smart assistants, and virtual reality integration. The latter two are opportunities for disruptive innovation. For this reason, more than half of the successful companies in the future AR market may not be traditional giants, but emerging companies.”

Still, less than 20% of designs in the Chinese smart glasses market had AI functions by the end of 2024. AI integration is difficult in part due to dependence on external computing power — suitable processors are simply way too big and energy-hungry to fit into the frames of the glasses. That’s why AR glasses often rely on split compute to preserve battery life — tasks like translation are offloaded to the user’s cellphone, and then the result is transferred to the glasses through WiFi to save power. This also means that AR companies benefit substantially from locking down partnerships with phone companies that control the ecosystem.

China's VR/AR industry reportedly raised 2.948 billion RMB (~US$340 million) across 30 investment and financing events from January to November 2024. While tech giants like Huawei and Xiaomi have been dabbling in AR, innovation has primarily been driven by the so-called “Five Little Dragons” of the AR industry: XREAL (优奈柯恩), RayNeo (雷鸟创新), Rokid (灵伴科技), INMO (影目科技), and the Xi-endorsed Meizu (星纪魅族).

Thanks to fierce domestic competition, most models of AR glasses currently available in China are in the price range of 2,000-4,000 RMB (~US$275-$550). By comparison, Meta’s Ray Ban glasses, which don’t include displays in the lenses, start at US$299, and their full-service Orion prototypes cost US$10,000 per unit to produce.

We’ll briefly highlight each company’s quirks below.

XREAL 优奈柯恩

XREAL was founded in Hangzhou by Xu Chi 徐驰, but the company has recently relocated its headquarters to Shanghai. While other AR startups primarily use Snapdragon processors made by Qualcomm, XREAL uniquely uses self-developed chips. XREAL’s X1 processors can reportedly achieve a latency of 3 milliseconds (compared to 12 milliseconds for the Apple Vision Pro). Xu explains:

“The X1 chip equipped in the XREAL One not only successfully resolves technical challenges like 3DoF spatial anchoring and ultra-low latency, but also fundamentally overcomes the longstanding issue of inconsistent cross-device experiences for AR glasses. In the past, to provide a consistent experience across different operating systems, we had to develop separate software for Windows, macOS, Android, and iOS platforms. Yet even then, it was difficult to fully ensure uniformity in the user experience. With XREAL One, however, we’ve finally found a more elegant solution. By integrating computing power directly into the glasses themselves, we’ve fundamentally changed the game — delivering a truly meaningful “AR for all” experience.”

Xu also noted that 65% of XREAL’s smart glasses components are developed and manufactured in-house, and the company is aiming for 100% hardware independence within the next two years. But this hardware-focused approach has tradeoffs — XREAL’s products don’t currently come equipped with out-of-the-box AI features, although the company claims that LLM integration will be included in a future software update. Here’s Xu’s justification for not engaging with the AI+AR hype:

“AR glasses are always centered around user experience. We won’t blindly build AI glasses just for the sake of having AI. Instead, we start from the user experience and work backwards to identify what kinds of technical innovation are necessary.

In the R&D process, we have to wisely understand the limits of current technology — knowing what’s realistically achievable and what isn’t, at least for now. Blindly chasing breakthroughs can lead to disastrous outcomes.

Although XREAL is a relatively small company, our strength lies in the smart use of limited resources. We’re very clear about where to allocate funds to keep our product logic sharp and focused.”

Consumer reviews of the XREAL glasses have been quite positive despite the lack of AR+AI branding. One reviewer from Hong Kong wrote in April of 2025:

These are the best AR smart glasses, perfect for entertainment whether you're out and about or relaxing at home (such as in a small apartment or lying in bed). Once connected to a compatible device via USB-C, users can watch their favorite shows, movies, or games on a large virtual screen, like having a private cinema.

The glasses feature a Full HD 120Hz OLED display with a brightness of up to 600 nits, delivering vivid images and excellent contrast for viewing dark scenes. The image quality is further enhanced by electrochromic dimming lenses, allowing users to adjust the lens brightness with a switch. At the clearest setting, the lenses let users easily see their surroundings, while at the darkest setting, they effectively block out external light, providing an almost perfect viewing experience.

In addition, the glasses boast impressive speakers, with sound quality tuned by Bose, offering rich audio across highs, mids, and lows. While using separate headphones can still enhance the audio experience, this is the first pair of smart glasses where users feel that headphones are an optional accessory rather than a necessity.

XREAL has also prioritized business partnerships under Xu, betting that the experience will translate to market share once AR technology is mature enough to garner widespread consumer interest. XREAL has inked deals with BMW, T-Mobile, AT&T, Bose, and Google.

RayNeo 雷鸟创新

RayNeo is a subsidiary of TCL, an electronics manufacturing giant that is partially state-owned. The company’s Mandarin name translates literally to “Thunderbird Innovations,” and they are notable for marketing AR products aggressively outside of China. CEO Li Hongwei 李宏伟 explains in an interview from January 2024:

“We wanted to establish a foothold in overseas markets first, so we started by working on distribution channels. For example, we launched on Amazon, and in November last year, we achieved strong results by ranking first on both the new arrivals chart and the bestsellers list in the smart glasses category.

Right now, we’re also selling our products in some boutique stores across Europe… and the sales performance there has been fairly good as well.”

RayNeo recently announced a partnership with the International Olympic Committee, so we’ll probably see promotional videos for Los Angeles 2028 shot from the perspective of athletes.

The company also partnered with Alibaba Cloud to develop a multimodal AI model specifically for AR glasses — now, the average response time for AI queries on RayNeo glasses is reportedly 1.3 seconds.

Finally, RayNeo was the first company in the Chinese market to sell AR glasses for less than 2,000 RMB. Here’s how they keep costs low, according to Li:

“If Meta’s Orion is the Vision Pro of AR glasses, then the RayNeo X3 Pro is more like the “Vision.” …It does not use the most cutting-edge technology in all technical indicators like Orion…. For example, in the selection of optical waveguides, although RayNeo has a silicon carbide wafer version internally, our commercial products do not use this material. Instead… we use photolithography machines and chip etching processes to make waveguides on glass, so as to better balance costs and product experience. As the company's strategy, RayNeo will not choose to pile up all the industry's most advanced technologies in listed products for the sake of showing off its skills, but will develop cutting-edge technologies and products internally, and eventually launch leading, pragmatic, and mass-producible products to the market.”

Rokid 灵伴科技

Rokid was founded in Hangzhou in 2014 by Zhu Mingming 祝铭明, who left his job at Alibaba to start the company. In 2024, Rokid received nearly 500 million RMB and 100 million RMB across two rounds of financing. Their AI-equipped glasses use Alibaba’s Qwen models for real-time translation.

Among the AR dragons, a uniquely large portion of Rokid’s investors are SOEs. Here’s Zhu’s explanation from an April 2025 interview:

We have significant influence on the B2B side — the cultural and museum market alone is worth around a hundred million yuan annually. Our ecosystem partners bring in tens of millions, and each year, several million people visit museums. Rokid is the only player in this sector. …

B2B operations serve as a bridge for interacting with the government. Even in local governments with no overlapping business opportunities, there are always departments for cultural tourism and museums.

Rokid partnered with the PLA to build custom AR glasses for use on China’s space station. Their website also highlights B2B partnerships with oil, gas, and mining operations. But for now, Rokid’s consumer products receive mixed reviews.

I was able to try the Rokid Max AR glasses at the Guanghua Digital Plaza in Taipei. I watched a clip from Avengers: Endgame with the glasses, and the image quality was quite nice. You can correct for nearsightedness without customized lenses if your prescription is between 0.00D and -6.00D. I expected my eyes to feel strange switching between far-away objects and the close-up digital projection (as is reported by many reviewers), but I didn’t notice any such feeling during short-term use.

Unfortunately, they are very ugly and sit weirdly far and high up on your face. They’re also not wireless (unless you buy a wireless adapter, which adds latency), and the Rokid Max glasses I tried don’t support AI features like translation (Rokid’s AI-equipped glasses are a separate product line).

My impression is that, for now, the most tangible use case for these glasses is avoiding neck strain.

INMO 影目科技

INMO’s Chinese name could be translated as “Image Eye Technologies.” As a newer entrant into the Chinese AR market, their strategy has focused on affordable consumer AR with basic features, advertising use cases like cycling directions, translation, and taking notes in meetings. The company marketed their INMO GO glasses as “the first consumer-grade AR glasses with deep AI integration,” highlighting translation capabilities in 11 languages, smart notifications, and AI assistant features supported by ChatGPT. Here’s INMO founder and CEO Yang Longsheng 杨龙昇 on his vision for popularizing AR glasses:

“We hope that AI glasses can be like a personal assistant in the future, helping me order takeout, order a cup of coffee, etc. at any time. …

In the past, social interaction was generally between real people, but based on the emergence of AI and the improvement of technologies such as virtual humans, I believe that in the near future, perhaps within four or five years, people will be able to socialize with these virtual intelligent entities.

I can create a virtual image of my dreams, infuse it with the personality I want, and then interact with it in this entire virtual-real world. …

[AI companies] also need to find some landing points for these intelligent entities. Glasses are undoubtedly the best form at present, which also encourages them to try more content on glasses.”

INMO’s content partners include Baidu (China’s Google analogue), TanTan (a dating app), and game developers NetEase and 37Games. Chinese reviewers seem to appreciate the INMO software ecosystem — one user wrote,

“The INMO AIR 3 uses the IMOS 3.0 operating system, an OS specifically designed for AI+AR terminals. The AIR 3 can project the equivalent of a 150-inch giant screen. With the help of a 3DoF smart ring, the AIR 3 supports screen-space hovering functionality. IMOS 3.0 provides more efficient applications and a more immersive experience through intelligent interaction and spatial display capabilities. IMOS 3.0 not only supports native AR applications but is also compatible with most Android apps on the market, offering a relatively rich content ecosystem.”

INMO also recently announced a partnership with China Mobile to integrate their AR products with China Mobile’s Jiutian LLM ecosystem.

Meizu 星纪魅族

Meizu is owned by Geely, a conglomerate that primarily manufactures automobiles and holds a controlling stake in Volvo. Meizu’s primary business ventures are smartphone manufacturing and developing the FlyMe Auto operating system used by many Geely-owned car brands. As such, Meizu’s AR glasses are compatible with FlyMe-equipped vehicles.

At the 2025 Shanghai Global Investment Promotion Conference, Meizu founder Li Shufu 李书福gave a highly-publicized speech while wearing a pair of Meizu glasses, which acted as a teleprompter.

The company has developed its own FlyMe LLM, but Meizu glasses also support integration with third-party AI models, including DeepSeek, Qwen, and ByteDance’s Doubao. In early 2024, the company signed a partnership with Malaysia's Juwei Group to expand sales of Meizu glasses in Southeast Asian markets.

Looking Forward

China’s “Hundred Lens War” is a live experiment in hardware innovation under pressure. While consumer interest remains tepid, China’s Five Little Dragons are constantly launching new products in search of the ideal combination of design choices — wired vs wireless, with lens displays vs without, AI vs traditional translation, and many more. If AR glasses eventually succeed in locking down compelling use cases, it’s likely just as likely their decisive breakthroughs will come from not from Silicon Valley but Shanghai.

Thanks to Mike G. and Benjamin Reinhardt for offering feedback on previous drafts of this article.

Can studying Mao Zedong help explain Donald Trump?

To find out, ChinaTalk interviewed the legendary sinologist Orville Schell, who visited China during the Cultural Revolution and is currently at the Asia Society.

We discuss…

• Mao Zedong’s psychology and political style,

• Similarities and differences between Mao and Trump,

• How Mao-era traumas reverberate in modern China, including how the Cultural Revolution has influenced the Xi family,

• How Zhou Enlai and Deng Xiaoping survived the Cultural Revolution, and which of their tactics could be useful in modern America,

• What civil society can do to defend democracy over the next four years.

Co-hosting is Alexander Boyd, associate editor at China Books Review and former ChinaTalk intern.

Listen on Spotify, iTunes, or your favorite podcast app.

Culture War x Cultural Revolution

Alexander Boyd: Orville, you wrote the most prominent essay on Trump’s Cultural Revolution, comparing Donald Trump’s behavior in office and personal style to Mao Zedong. To start, what is Maoism, and how would you describe Mao Zedong?

Orville Schell: Mao Zedong, of course, was the great progenitor of the Chinese Communist Revolution. He was a Marxist-Leninist, and he liked control as any autocrat would. However, one of the hallmarks of Mao Zedong was also an abiding interest in throwing things off balance as a way to gain even greater power.

In this regard, he became a great fan of Sun Wukong. This golden-haired monkey was one of the heroes of the classic Chinese novel, Journey to the West, which tells the story of Buddhist scriptures being brought from India to China. One of the most prominent features of this monkey king was his love of disorder. His sort of watchword was “dànào-tiāngōng” (大鬧天宮), to make great disorder under heaven. Mao Zedong actually ended one of his poems with that line, and he always loved this novel, Journey to the West.

When the Cultural Revolution arrived, I think this was a real consummate expression of Mao’s affection for chaos. He did feel that not only did Chinese society need to be upturned, but the whole political structure of China needed to be upturned. Everything in effect needed to be “fānshēn" (翻身), as he said, “turned over.” He adopted many expressions similar to the Monkey King that expressed his love of contradiction and disorder. Class struggle, of course, which became the hallmark of the Cultural Revolution, was a form of deep and disturbing disorder.

Alexander Boyd: Where did Mao’s need for constant chaos and rebellion come from? In your Project Syndicate essay, you posit that it came from his troubled relationship with his father. Can you describe the parallels to the case of Donald Trump, who had a famously domineering father himself?

Orville Schell: The one place where we really get chapter and verse on Mao’s relationship to his father growing up in Hunan at the end of the 19th century was in Red Star over China by Edgar Snow. Mao told Snow that he had a very adversarial relationship with his father. He even said he hated his father, that his father was a tyrant, and they were constantly battling. On a number of occasions, Mao Zedong actually ran away from his home.

He did have a very sort of Buddhist-inclined, loving mother who made a lot of difference. But the relationship with his father clearly set off the notion of the world as an adversarial place. He tells Snow that he learned only by standing up to his father could he survive, and then his father would come to heel in some sense and not just overwhelm him with his sort of tyrannical paternal role.

That sort of characterizes Mao, and in fact Trump too, as we learn from his niece (the daughter of his older brother), who is a psychologist. Trump had a father who was very preemptive, very tyrannical and very judgmental.

Wallowing off into the bog of psychobabble here, any human being who reads literature knows that a father’s relationship to his son and vice versa is a profoundly influential relationship. As a young man grows up, that relationship forms him.

Alexander Boyd: We know that Mao loved to speak in allegory. He often would speak about one thing but mean another. How much can we believe his stories to Edgar Snow? And is it just a metaphor for American paternalism?

Orville Schell: I think at the period when Edgar Snow was doing these interviews with Mao was in the 1930s. We hadn’t started all the rectification movements and Mao had not been in power with all his mass campaigns. He had just arrived in Shaanxi province. I think that we get a pretty unalloyed representation of his early years. I don’t see there’d be any reason for him to be setting traps or deceiving Snow.

Throughout his life, Mao easily took umbrage at things when he felt sat upon or disrespected. Another famous example, of course, is when he finally went to Beijing in the late 1920s and he became a senior intern of some sort at the Library in Beijing University. He used to feel incredibly disdained by Chinese intellectuals who’d come in and sneer at his hickey Hunan accent.

If you‘ve ever heard Mao Zedong speak or seen a film in which he speaks, he’s almost unintelligible to Mandarin speakers. Like most Chinese people of his generation, including Deng Xiaoping and Chiang Kai Shek, he preferred to speak his local dialect.

This laid the track for Mao’s antipathy towards intellectuals, just like Trump hates intellectuals and universities. He thinks they’re arrogant, they’re elitist, they look down on people like him and down on working people, et cetera. Mao felt very much that no matter what he did in his formative years, he was disdained and disrespected by the intellectual class.

Of course, in that sense, I think he is a metaphor for sort of the whole ‘kultur’ of China in a sense that it is aggrieved, it’s been humiliated, it has been looked down on, kicked around, exploited, you name it.

Alexander Boyd: Another similarity between Trump and Mao is that both men had a lot of wives, often tumultuous ends to those relationships with women, and a propensity to start new relationships before the last one ended.

Yet another parallel is Mao’s exposure to Hunan secret societies, specifically Gēlǎohuì (哥老會), the Red Gang, and his time as an organizer at Anyuan. This mimics Trump’s early mob ties, especially in the New York of John Gotti and then Rudy Giuliani, who was a big mob fighter back in his day.

Their styles of governance are also similar — Mao made frequent trips outside of Beijing, and he loved to launch campaigns while on the move, much like Trump’s Mar-a-Lago golf club trips.

Are these comparisons substantial, or is it just the case that when you have so much information about two people, you can always draw connections between them?

Orville Schell: It’s true that both Mao and Trump had a lot of wives and a lot of ladies. What that suggests is that they need to have that kind of affirmation and signs that they can beguile people and win people over, which speaks to me of a fundamental lack of self-confidence. Both of them derived a certain measure of prowess from their ability to attract beautiful women.

Trump still talks about women as objects and as adornments. Mao clearly was the same. There were other leaders in China who were not. Chiang Kai-shek had a lot of ladies (and ladies of the night as well in his youth), but after he got married, he was quite faithful. I don’t quite know what to make of Zhou Enlai, but he had one wife. Whether he was gay or not, it’s a question people do care about.

Mao was somewhat special, and he arrogated that special right to poach on women himself. But when revolution came home to roost after 1949, it was not something he found acceptable in ordinary people or even in his acolytes. He was very puritan. But he was not just like Trump. Trump purports to be a Christian and yet doesn’t abide in any meaningful way by the notion of loyalty implicit in marriage.

Jordan Schneider: Shall we discuss the deep state example?

Alexander Boyd: Trump famously vowed to drain the swamp and railed against a deep state that he perceives as having both frustrated his attempts to exercise power in his first term and prevented him from regaining office in 2020. Now he’s engaged in a campaign of revenge against all those purported deep state agents.

Many people, and you chief among them, have made that comparison to the Cultural Revolution where a frustrated and suspicious Mao unleashed these animal forces within China to take down a party in a state leadership that he felt was shackling his own ambitions to remake China. Is that an accurate comparison in your mind?

Orville Schell: One of the hallmarks of Mao’s revolution was a sense that somehow the party (which he himself had helped build) and the state (which was the handmaiden of the party) were ultimately the refuge of rivals.

He had a great antipathy against bureaucratism. This also speaks of his love of disorder as a creative force. When he started the Cultural Revolution, one of the first things he did was to issue a wall poster that said “bombard the headquarters” (pào dǎ sīlìngbù, 炮打司令部). What he meant by that was that he felt the party had become ossified. It had become the refuge of bureaucrats who were living high on the hog but didn‘t want to make revolution anymore and found class warfare too disruptive.

He felt that it was time to destroy it. How did he do that? He gave permission to young people, idealistic young people, to attack. Almost all of the leaders who were potential adversaries of Mao either died or were purged. Xi Jinping’s father was one of them, a vice premier who had a very bitter purging.

In that regard, Mao Zedong took, I think, great satisfaction in overturning even his close revolutionary comrades. Here Trump is not too dissimilar. He seems to be invigorated by the idea of destroying institutions that he views as refuges of people who might be against him and of firing people he’s hired, turning against people who he perceives as potential rivals. He demands complete fealty and loyalty or you’re in trouble.

Here too, I think there’s a kind of a similarity in the way the two relate to other leaders and to institutions of government, the deep state. It’s the equivalent for Trump of what the party and the state that he himself helped build were in China. He saw them as slowing down his revolution, as harboring his adversaries and as being overly bureaucratic, what he called “sugar-coated bullets of the bourgeoisie” (资产阶级的糖衣炮弹). In other words, they’d given up their revolutionary vigor in favor of staid bureaucratic forms of government.

Alexander Boyd: So if there’s an attack on the bureaucracy, does that make Elon Musk and DOGE a new Kuai Dafu 蒯大富 and the Red Guards?

Orville Schell: Musk is older. Kuai Dafu was one of the earliest Red Guard leaders at Tsinghua University when Mao issued his order to bombard the headquarters.

Elon Musk should know better, but I think he too has a kind of innate impulse that chaos is a creative element. It’s one step away from the Silicon Valley mantra, “failure is positive.” But I think he does share with Trump this idea that somehow you need to clean out the Augean stables of the government. I don’t know why Musk might feel disrespected or disdained when he’s been so successful and the richest man in the world.

Here, too, I think we have to remember that these leaders are human beings. They’re not just rational creatures who look at the national interest of the country or read reports to make rational decisions. Some leaders are completely crazy.   We know from Euripides and Shakespeare that leaders are completely crazy and they do astounding things.

You can read Stephen Greenblatt’s book, Tyrant: Shakespeare on Politics, which is about six great plays that Shakespeare wrote about tyrannical leaders. There’s nothing new here. It’s just that policy people and I think many academics are loath to recognize that we’re also dealing with something very human here, namely, leaders with deep and tragic flaws, which in Euripides are hubris, arrogance, and overreach.

When Croesus — we‘ve derived the expression “rich as Croesus” — went to the oracle of Delphi, he governed the state of Lydia, about whether he should go to war. He got the diction back from the oracle that if he went to war, he’d lose his kingdom. What did he do? He went to war and lost his kingdom.

There’s a little of that going on here. I think we need to factor the human dimension into the equation of understanding big leaders like Putin, Orban, Xi Jinping and others. Look back at their formative years.

Here, I highly recommend a wonderful new book that’s coming out by Joseph Torigan on Xi Jinping’s father, Xi Zhongxun 习仲勋. You see what nightmarish experience Xi Jinping went through as the son of a man who was twice purged during Xi Jinping’s teenage years, and was purged the second time as a counter-historical, counter-revolutionary. And what travails Xi Jinping as a young man went through to have a father who was in the five black categories.

This is a little bit beyond the mandate of so-called China socialist. But I do think that here’s where literature, drama, some of these other representations of leadership help us understand what’s going on.

Alexander Boyd: Let’s stay on Xi Jinping here for a second and Xi Zhongxun as well. What is Xi Jinping’s view on the Cultural Revolution today? Obviously it’s opaque, but I agree Torigan’s book is incredible and I read his section on Xi Jinping’s Cultural Revolution.

Orville Schell: I think it’s unfair to say that Xi Jinping is like Trump. Xi Jinping does not like disorder. He does not want to create great disorder under heaven, unlike Mao. The part of Mao he does bond with and did grow up with and appreciate is the Leninist part — the organized state, organized party control, autocracy. But he has no fascination for the part which we’ve been talking about.

This is why I suggested if Xi Jinping wants to come to some better understanding of who Trump is, ironically he has a homegrown example in Mao Zedong. He lived through it. He knows what people like that can do to a society and to even the global order. I think maybe he has thought this way. I don’t know, but I think he uniquely, unlike many Americans who’ve never been through this kind of disturbance, shouldn’t be so surprised by Trump.

Why do we think as Americans, when Italy had Mussolini and Germany had Hitler and Russia had Stalin and Spain had Franco, Salazar in Portugal and on and on, why do we think that we are somehow immune from these kinds of aberrant, overreaching, arrogant, and finally incredibly destructive leaders?

Jordan Schneider: China from 1949-1967 was a very different place with a very different governance system than America circa 2016 or 2025. Shall we discuss some of the differences here?

Alexander Boyd: Well, I think the first place that we should start is rise to power. Trump, you know, for better or for worse, won two elections, and Mao won power through civil war and various other means. Orville, what would you say is the biggest difference between Trump and Mao?

Orville Schell: Trump is more like Hitler, who came to power by being elected Chancellor of the Reichstag, whereas, as you point out, Mao Zedong came to power through insurgency and civil war.

Obviously, you can‘t completely compare these people, but I think in trying to understand the leaders of the present, it does behoove us to look back at leaders in the past who also created disorder, one kind or another — a world war, economic crisis, a revolution, whatever. That might help us understand a little bit what it is they’re offended by. What do they want? What would propitiate them? How do you deal with them? Can you deal with them?

I don‘t mean to compare Xi Jinping with Trump. But only to say that China’s historical experience of having undergone probably the most tectonic, catastrophic, and destructive revolution in history might help Xi understand what animates a leader like Trump and how best to deal with it.

Jordan Schneider: How best to deal with Mao is kind of not something people really figured out. Can you talk about what the antibodies were over the course of his reign, and highlight some examples of successful and unsuccessful pushback against his craziest ideas?

Orville Schell: Well, of course, the best, the biggest antibody of all to Mao was death. Many autocrats are very disruptive. Hitler died in his bunker, Stalin died, and wildly began to change. I don’t know what the antidote to Mao at that period would be, but I will say this, that if we look at our own country, there is a hint of similarity among the way people come to heel when big leader culture gets rolling.

The Republican Party is completely supine now. We do have the Democratic Party still raging against the storm. But one of the lessons I think that is quite striking about the Chinese communist revolutionary period was the way in which everybody finally was neutered. Those few who did speak out, and there were some, had very bitter ends. All know what happened. This is another hallmark of powerful and effective autocratic leaders is that they manage by one way or another — and one way is disorder — to intimidate people, frighten people into submission and silence.

Alexander Boyd: I think in a lot of histories of the 1970s though, everyone points to 1971 and Lin Biao’s death.

Jordan Schneider: Why don’t we do the Great Leap Forward and response to that? Because Mao had to do with self-criticism, right? This was a real brushback moment for him where after killing eight figures of his own citizens, there did end up being some pushback from the top that forced him down a policy path he wasn’t really excited to take.

Orville Schell: After the Great Leap Forward, many leaders like Deng Xiaoping and Zhou Enlai and Peng Dehuai felt it was too excessive. Forty million people dead, starvation, agriculture in ruins. They did for a brief period of time prevail. What was Mao’s answer?

Mao’s answer was the Socialist Education Campaign, which is a prelude to the Cultural Revolution, which ended up labeling people like Peng Dehuai at the Lushan conference, put out of business very quickly, ended up in jail. The president of China, Liu Shaoqi, Deng Xiaoping, were sent down. In other words, almost all the leaders, veteran revolutionaries that had accompanied Mao in the Long March in the Yan’an period, ended up in the doghouse or dead.

That was China’s experience. That made it very difficult for any voices of dissent to find any purchase. I remember being there myself during the end of the Cultural Revolution and just thinking, well, this is the way it is. There were no voices. They wouldn’t even talk to you as a foreigner. Nobody had permission even to interact in a normal human way with anybody foreign or an outsider because they were afraid. Mao had brought complete submission down onto society.

That didn‘t mean he suffocated all of the impulses that had built up over previous decades — they remained latent and dormant and they arose again when Mao died.

Alexander Boyd: Basically the biggest argument against Trump’s effort to remake America in his image, to bring manufacturing onshore, is that Mao, with more power at his disposal, more party at his disposal, a whole society cowed, actually failed. Even more recent scholarship, like Odd Arne Westad and Chen Jian’s new book The Great Transformation: China’s Road from Revolution to Reform argues that opening and reform really began in the 1970s. If Trump is Mao in this comparison, is the Trumpist effort, you know, this great “cultural revolution” in America, is it doomed to fail right from the beginning?

Orville Schell: Well, I think we could be doomed to something even worse. Mao failed in the Great Leap Forward. Then in order to regain and maintain power, he brought on the Great Proletarian Cultural Revolution. It’s hard to know. We can’t predict history and we don’t know whether American democracy will survive.

What we want to say is that we know this archetype of leader and whether it’s a communist, a Leninist, a so-called democrat, Nazi, fascist, whatever. That’s why I wrote some of these essays, to just remind people that there are examples throughout history and literature of these kinds of people. We need to better understand because now it’s America’s turn to have one.

Jordan Schneider: I think another one of the big differences that Tanner Greer pointed out on a show earlier where I tried to compare Trump to Hitler was the ideological commitment level that Trump has versus someone like a Mao or a Hitler who deeply believed in his bones in class struggle or Lebensraum or international Jewish conspiracy. Yes, Trump’s got some views on trade policy, but he lifted them after the bond market changed. It’s the sort of the level of focus which he can bring or has shown, the level of focus that he’s shown he can bring to policy stuff versus some of the other sort of totalitarians that you reference who maybe have some personality traits in common, but also kind of have a real agenda behind them. Whereas our current president, not so much.

Orville Schell: I agree with that, Jordan. Mao had a very highly evolved ideological agenda and analysis of how the world worked, where history was going, knowing what the dynamics would be. They did hew to that in various ways, sometimes rather opportunistically or in a utilitarian fashion. That’s communism.

But fascism is a very different animal. If you read Robert Paxton’s book on fascism, you begin to understand people like Mussolini. He had no ideology. He was sort of inventing himself as he went along. I think that’s one of the big differences between Trump and Mao. Mao was a very intelligent man and actually a good writer, good poet. He thought.

I don’t think Trump thinks — he acts almost like an animal. He feels this today and he acts, he responds. He certainly has no ideology or no sort of political commitments to principles that guide him in what he does. It’s more what he feels like doing. He feels someone doesn’t like him. He feels threatened. Whatever. It’s almost animal-like in his responses, which are completely irrational.

Alexander Boyd: This is actually not so much a continuation of this question, but it’s a different tack. I’m curious about Mao’s foreign policy and Trump’s foreign policy. A curious similarity is that Mao stated, “I like rightists.” He met with Richard Nixon. He found them easy to deal with in general, perhaps because he understood the ideological motivations of, or he perceived himself to understand the ideological motivations of capitalist right-wing politicians.

Then, Trump himself meeting today in the Oval Office with Carney, the Prime Minister of Canada, and earlier saying something like “I like the left” in reference to him not wanting Pierre Poilievre to win in Canada. Why do you think that Mao liked rightists or claimed to like rightists? What sort of insights does that give us into Trump’s foreign policy?

Orville Schell: In many ways, Mao could be a rightist, but I don’t think he liked anyone who opposed him. He viewed the right as opposing him and as opposing his ability to control thinking and ideas. This is why you get the whole idea of sīxiǎng gǎizào (思想改造), thought reform, that there’s a correct way to think and Mao helps limit that, describe that, lay the boundaries for that out. If you don’t want to think that way, then you’re a rightist or maybe even a leftist and then you deserve to be defenestrated.

I think Mao says he likes rightists because he thinks they’re practical and he can deal with them. It’s a bit of a capo-to-capo business. You can deal with a thug, even if you call him a rightist, but I don’t think he had any affection for rightists or leftists. These are categories of convenience into which he put people when he needed to get them off the stage.

Jordan Schneider: Yeah, I feel like the capo-to-capo thing is it’s less him liking Mark Carney, which I just truly do not believe, and more the ease and excitement where he gets talking to big, powerful, authoritarian leaders, as opposed to democratically elected ones.

Orville Schell: I think that, Jordan, if I may say, when Nixon and Kissinger went to China in ’72 and ’71 to set it up, I think there was a certain thrill for Zhou Enlai and Mao to have these people come hat in hand to Beijing to talk to them. Because remember that even though they were opposed to imperialism, colonialism, capitalism, America, et cetera, there is, I think, in my experience, at least amongst Chinese leaders representing their country, a deep and abiding wish to be respected.

They speak about that all the time, mutual respect and understanding, as if to say, “Can’t you just respect us as a dictatorship? Show us some respect. We’ve got a good economy, we’ve done a lot of amazing things,” which is true. Again, autocrats, one of the things that really riles them is that they’re cast out, they’re disdained, not considered proper company for liberal democratic states.

When China is the aggrieved, humiliated, kicked-around sick man of Asia, they want nothing more than to be respected. That is a complete contradiction because if you don’t act respectably, it’s very hard to be respected. Yet that is something they deeply crave, even though they would never acknowledge it.

Jordan Schneider: Yeah, I think it kind of works the other way as well, where you just see how big Trump’s smile was in his 2017 visit to China.

Orville Schell: I’ve been on several presidential trips, like for instance, with Clinton, which was completely different, very informal and open and cheerful. But Trump’s trip, it was all about pomp, circumstance, awe, ceremony and ritual. Both leaders were trying to impress each other and Xi Jinping wanted to impress Trump. You remember he took him to the Forbidden City and they had a banquet and all the rest of it.

But there were no moments of bonhomie, of personal smiling and back-slapping and saying, “Hey, we’ll work this out.” No, it was all about sort of who is the bigger dog with the most impressive marching band and most impressive Great Hall of the People. I think that’s very characteristic of both Trump and Xi. Trump wants to have a parade, just like Xi gets to have parades with a lot of tanks and missiles going by.

There’s an element of similarity, I think, of both deeply insecure men fundamentally, and nothing like a good parade to puff up insecure leaders’ egos.

Jordan Schneider: Can you talk about the sort of red versus expert dilemma, which we’re seeing play out with Laura Loomer in this administration?

Orville Schell: Remember that during the Cultural Revolution there were many, many struggles going on. One of them was the struggle between being Red and an expert. Experts, of course, were people who knew how to do something. They were the intellectuals, they were the scientists, the technicians, people who ran institutions and could be accused of bureaucracy.

The Reds were the people who wholeheartedly embraced Mao Zedong and were dedicated to overthrowing institutions of the experts. In that period, not only was expertise disdained and diminished in standing as a societal avocation, it was a hallmark of those first people being insubordinate. It was a different kind of loyalty — not to Mao, not to Maoism and Marxism-Leninism, but to rationalism, to scientific experiment, or these other things that had a kind of a logic that defied complete and total loyalty to the Great Helmsman, whatever he represented at the time.

Whether the party was intact or not during the Cultural Revolution, it was not loyalty to Mao and the revolution versus loyalty to whatever other thing you might be — a scientist, a businessman loyal to profit, and policymakers who are loyal to trying to figure out the national interest. That was a huge divide.

We see that now with Trump. What he wants is not people who know something in the FDA or the FAA. He wants people who are loyal. So you got Laura Loomer, like Kang Sheng or Chen Boda or someone running around firing people and putting people in prison for Mao.

Jordan Schneider: What was the upside for Mao of getting rid of the experts?

Orville Schell: Liu Binyan, the great writer of the 1980s, wrote a book called A Higher Kind of Loyalty. This was an analysis of people who felt a loyalty, whether it was through religion, technology, science, intellect, or just ideals, to something other than the revolution and the leader.

What Mao rejected and recoiled from was professionals who were experts, who said, “No, this revolution does not make sense economically, scientifically or in any other way. It’s mayhem.” That put them immediately on the enemies list. That’s why intellectuals, and they categorized them into many different categories, were pilloried because they couldn’t be totally loyal as religious people were, because religious people owe a different loyalty to their God and to their principles and to their morality, not to the leader.

Mao couldn’t stand that, so he waged war against those people. We see sort of, I think, hints of that happening now in America where people like Fauci in the first administration of Trump were not respected at all. He was a very good scientist, very devoted public servant. We see attacks on vaccines, Robert F. Kennedy Jr., science has proven measles vaccines work. We see a lot of things like this looming up again where what’s important is not logic, but it’s loyalty to the leader.

Alexander Boyd: But Mao also resented his need for these bureaucrats, for these technical experts. And chief among them was his own Confucius, Zhou Enlai. Even though he was fighting against these empiricists and scientists, cultural leaders and everything during the Cultural Revolution, probably the number one empiricist of all was right next to him. Although at the same time many Chinese spoke about him like we spoke about John Kelly in the first term. “Oh, he’ll be able to restrain Mao’s worst instincts.” How do you read Zhou Enlai?

Orville Schell: Zhou Enlai was a restraining influence. He was also a complete factotum. When you see what he was put through, like in the rectification campaign in 1942 where he sat before the Politburo for five days, wrote self-criticism, self-immolations, humiliating, pusillanimous kind. He paid a bitter price as a human being to keep in Mao’s good-enough graces. Still, Mao endlessly tormented him and humiliated him — very smart man. And he took it. Why? That’s an interesting question.

Did he think he was doing good like Matt Pottinger in the first administration of Trump, that if he just kept his head down and tried to do a good job, he could restrain the leader? But there was also probably a lust for power, an urge towards power which kept him there. He once got that needle in his arm, it was very hard to get out or he’d end up like his friend Liu Shaoqi — imprisoned or dead.

These kinds of leaders demand not one-time propitiation declarations of fealty and loyalty, but continuous. The leader keeps ramping up the ask. If the lieutenant wants to stay in their graces, they have to keep becoming more and more genuflective. We see an awful lot of people left the first administration of Trump and now he’s already lost all the people he’s lost. I mean, Rubio, everybody. He’s taken over positions of Waltz and others who drop like flies because it’s very difficult to satisfy the demands of autocrats who require 200% loyalty.

Alexander Boyd: Is part of this the “Coalitions of the Weak” thesis put out by Victor Shih about how Mao would often rehabilitate disgraced cadres? You saw that with Deng Xiaoping, you saw that with Zhou Enlai, he’d constantly send people down, bring them back, criticize them, humiliate them, purge them, restore them, and it ended up necessitating their loyalty. They created psychological dependence, but also political dependence.

Especially with Victor Shih, he’s talking about the Fourth Front Army in the Long March, I believe Zhang Guotao. Is that a similar coalition with Trump and Mao? And how did Mao’s coalition of the weak work? And is that an effective governance tool? We might be skeptical of it, but Rubio, who has no basis left in the GOP, has basically been entirely kneecapped, was humiliated in his run — he’s Zhou Enlai exactly. But he could also be our Deng Xiaoping.

Orville Schell: If you want some good reading, go back and read Deng Xiaoping’s self-criticism during the Cultural Revolution. He just abased himself and said over and over again, “I didn’t know Chairman Mao. I didn’t appreciate the brilliance of Chairman Mao.” Even he went through the ultimate humiliation but survived intact.

I was in Washington and went on the whole trip when Deng arrived in 1979 to normalize relationships. And you did get the sense of somebody who had his own sense of gravity about him, wasn’t a deeply insecure person just craving slavish loyalty. Deng Xiaoping was different. I would say Zhao Ziyang 赵紫阳 and Hu Yaobang 胡耀邦 were also different. When you look through the different leaders, you want to get back and judge their character. Yes, times were different, but Deng Xiaoping was special because he had seniority. He did get cashiered twice, but he never lost his fundamental sense of himself, which I think many other people did.

Alexander Boyd: Let’s be optimistic here and say there’ll be a post-Trump GOP. How did those Chinese survivors like Deng Xiaoping, Zhao Ziyang, etc. make it through Mao’s Cultural Revolution? And what could that tell you for an aspiring politician today hoping to make it through Trump 2.0 and still have a political career?

Orville Schell: Xi Jinping made it through and he’s done alright. There’s no simple answer. In the good old days when you also had imperial autocracies in the form of emperors, if you ran afoul of the government, you could run up into the mountains and become a Buddhist monk or a Taoist priest, and mind your own business. But that wasn’t possible under Mao.

We’ll see about what happens in America. I suspect we won’t get to such a state in America, but who knows? The question is, during autocracy, authoritarian rule, what should good people do? If you stick your head up, it gets chopped off. You can run abroad, you’re just in waiting. There’s not much you can do. It is a good question, and I’m not sure I know the answer to it. Keep saying what we need to say as we are here today. In China, that was not possible under Mao. It was not possible under Mao ever since the early 1940s when he began lowering the boom, wanting to create a new man and a new era, bringing on thought reform, rectification, and all the rest of it.

Subscribe now

Jordan Schneider: Let’s stay on this dilemma of the officials who know that they are living in crazy times but still want to help the people. What’s the right way to kind of look at what Deng and Zhou and others did in the Mao period?

Orville Schell: They’re always dancing on a razor’s edge. You know, it's not a, a dance I would care to know how to get out of. Fair enough. Some of these people — and Zhou Enlai had a measure of this, I wager — you want to do good by the people but the cost of staying in the game is very high.

The people in the Trump administration, in the first go around, there were a few, some quit and they did in some significant measure keep their integrity intact. And they did do some good restraining things. I think this administration is much harder. He’s bringing in the — Elon Musk is like a leader of Red Guards and the Proud Boys are Red Guards equivalent.

It’s a very difficult human dilemma to know if you want to be in government and you are drawn to political power, how do you do it now? Can you do it? Or should you just become a Buddhist monk or a Taoist priest and just go up on your mountain and wait? I don’t know the answer to that. Us who are writers? Who have not been in political power, don’t want to be in political power. We’re not drawn to that flame. So we do what we do.

Jordan Schneider: Comparing America 2025 to anytime in Mao’s reign, the downsides of recording a podcast like this are much lower.

Orville Schell: For now Jordan, but people have long memories and there are archives and there are a lot of people. The way Xi Jinping’s father fell the second time was over a book about a big leader in northwest Shaanxi province, Liu Zhidan 刘志丹, that he allowed to be published. Mao said, “Well, you’re trying to put too much emphasis on him as the hero, not me.” Anyway, it’s a long, complicated story, but simply to say that sometimes small things done in past, to autocratic regimes like China, are grounds for you being pilloried in the future.

Jordan Schneider: We’ll get into that arc with Joseph later this summer. I’m still feeling okay about our freedom to podcast.

Orville Schell: I’m glad you’re doing it. My virtue is that I’m a little older. I don’t need to be so worried about my future.

Alexander Boyd: Michael Berry talks about this in his writing about Fang Fang — the term míng zhé bǎo shēn (明哲保身) means, “Don’t speak out in order to preserve yourself.” I personally think that in the United States, we have a great privilege to be able to speak out, and we should exercise that privilege.

Orville Schell: We still do. The government in sort of in the shape it’s in, it puts all the more burden on the institutions of civil society. Universities, think tanks, libraries, and community organizations do not owe fealty to the central government, but owe fealty to what they do. Media would be another very important example — cultural organizations, orchestras, operas, whatnot.

Alexander Boyd: Trump has shown an immense fascination with the Kennedy Center in D.C., which is where I’m based. I think it was Cats that was his favorite.

Bach and Bloodlines

Alexander Boyd: Let’s talk, Orville, about bloodline theory. What was bloodline theory during the Cultural Revolution? Why did it matter? Mao himself wasn’t an endorser of bloodline theory, but it did have a lot of influence.

Trump always talks about genes. “It’s all in the genes.” Quite recently, he weighed in on the NFL draft about a quarterback who’s sliding, Shedeur Sanders, and saying, “He has phenomenal genes. They should have picked him because his dad was such a good player.” Is bloodline theory another parallel with the Cultural Revolution ?

Orville Schell: During the Cultural Revolution, the notion of bloodlines worked like this: if your father was a hero, so you were good to go. But on the other hand, if your father was someone of questionable background, then you bore that stigma. You were placed in that class category because families were categorized based on their class background.

As you all remember, Mao had this notion that certain classes had rights and were revolutionary, while certain classes — like the bourgeoisie and landlords — didn’t have rights. The bloodline concept was very pernicious because it meant that if your father was labeled as a counter-revolutionary, a rightist, a capitalist roader, or a bourgeois element, the children inherited that stigma through blood.

That’s why it’s fascinating to delve into familial relationships in any Chinese family. Xi Jinping is the most important case here because he’s now the leader. But I should also mention that in my experience — and this may be better explored in literature than in nonfiction — there’s a cascading effect. All the harm, damage, and attacks that occurred throughout the fifties, sixties, and seventies in China have endured across generations in the Chinese families I know, going from grandfather to father to son to grandson. They persist like microplastics in the ocean — they’re forever chemicals in a way.

We’ve paid no attention to this phenomenon. The way these experiences deranged families, destroyed people’s ability to respect and love their parents, caused betrayals of friendships, and led to the savage attacks that people inflicted on one another — Red Guards attacking their teachers — all of this continues to reverberate.

This trauma isn’t something you get over the next day, and it lives on in ways that are very difficult to analyze. There’s no data, and China doesn’t have a vigorous psychoanalytic tradition to help people understand what influences may have been passed down to them through their experiences with parents who suffered.

The Cultural Revolution was deep, and its consequences are enduring. That’s why, when Deng Xiaoping came to power and waved his wand to rehabilitate people, saying it was a new era, I felt incredibly skeptical. I believed there was a whole residue of impact deeply embedded within society and human beings.

There had been so much damage — not just Mao and the party treating people badly, but people being forced to treat their spouses badly, their children badly, their relatives, friends, and colleagues. This is something that endures.

Alexander Boyd: This endures in China to this day, you argue. When was your most recent trip back to China, and how do you see it enduring today?

Orville Schell: My most recent trip was just as the COVID pandemic hit. All you have to do is talk to your friends. I have a friend who went to Harvard, had a very difficult time with her parents, grew up in China, and she set up a group for Chinese women similar to her to discuss this. I found that incredibly interesting.

Some of the things that they stumbled upon as they were trying to analyze the relationships they have with their parents — how are they influenced by the relationship their parents had to their parents and to society, and power. Very few people have wandered into this field.

Robert J. Lifton, a wonderfully brilliant psychoanalyst who wrote Thought Reform and the Psychology of Totalism in the 50s and then Revolutionary Immortality about Mao’s quest to make himself immortal so that his legacy would live on. There have been very few people — Lucien Pye, Richard Solomon — who’ve actually looked into the human element. That’s why I wrote a novel, because I felt I couldn’t touch it as a nonfiction writer. I didn’t get to the question of the role of religion, music, culture, love, family. All of these things are abiding human concerns.

Alexander Boyd: You titled that novel after a Lu Xun 魯迅 essay. How come?

Orville Schell: I love that essay, My Old Home (故鄉). It’s a very wistful essay about returning home after things changed. My novel was about a classical musician and what happened to him when he returned back to China in the 50s as a lover of Bach.

If I may say so, there is no human being whom I think is more antithetical to Chairman Mao than Johann Sebastian Bach. In fact, I want to write a play called “My Dinner with Johann,” where they have a conversation. Because Bach was all about religion. Mao Zedong was all about the external. Something’s wrong? It’s out there, not in here.

Yes, Confucianism did have a notion of self-cultivation, but it’s not like Christianity.

Jordan Schneider: Well, we have to end with the ChatGPT imagined conversation between Bach and Mao.

Orville Schell: There was a show Henry Kissinger went to, and Robin Williams started wandering down the aisle afterwards. He passed Kissinger, and he was saying things to people as he went. He said, “Oh, Henry, love all your wars.” I could imagine Bach starting off by saying to Mao, “Love all your revolutions.”

Jordan Schneider: This is how we’re going to start. Rewrite with Bach saying sarcastically to Mao, “Love all your revolutions. ” Alex, you’re Bach. Let’s go.

Alexander Boyd:  Love all of your revolutions, Chairman. Tossing the world upside down seems to be your favorite key signature.

Jordan Schneider: Upside down is where history finds its balance, Herr Bach. The masses must turn the old order on its head to set it upright.

Orville Schell: Now you’re talking like a robot, like a propaganda minister. I think Mao would say, “Tell me, Johann, what’s all this about Jesus? Why are you so obsessed with Jesus?” That would get Bach rolling. You remember when Clinton was in China, where he went into the Great Hall of the People for the press conference. At one point, Jiang Zemin, completely sui generis, said to Clinton, “Mr. President, I have a question. Why are so many Americans so interested in the Dalai Lama and Tibetan Buddhism?” He was speaking in Chinese. Of course, Clinton went on a tear. But I thought that was a sort of interesting question to ask. You can’t imagine Xi Jinping asking such a question.

Alexander Boyd: What did Clinton say?

Orville Schell: Clinton said something like, “Chairman Jiang, I think if you had a chance to meet the Dalai Lama, you’d really like him.” Jiang, who’d been off script and bantering in a very nice human way with Clinton, grabbed the podium and, as I recollect, he said, “With your permission, Mr. President, shall we close this section?”

Jordan Schneider: What are you reading right now?

Orville Schell: I’m reviewing this Torigan book for Foreign Affairs. I’ve also been reading William Shirer’s The Rise and Fall of the Third Reich, Timothy Ryback’s book about 1931 and ’32 in Germany, and a Robert Paxton book on fascism.

Jordan Schneider: Let’s talk about Rise and Fall of the Third Reich for a second. What stuck out to you about that book?

Orville Schell: I’m very curious where we are on this sort of scenario — how Germany headed off into fascism and the Third Reich. It’s pretty frightening when you look back at the various steps, at what happened and who didn’t say anything, who just shut up.

There’s a wonderful diary of Victor Klemperer, who was the cousin of a famous conductor. He kept a daily record of what happened. He was Jewish, his wife was Catholic, and he lived in Potsdam. They keep saying, “Surely something will happen, surely someone will come, and that can’t be it. Surely the Allies will come in.” Of course, they didn’t. We ended up with Hitler being elected Vice Chancellor, then we’re off to the races.

I’m very interested in how things slide into this state where you end up with an autocracy. Remember that Germany was the highest form of European civilization, and yet you ended up with Hitler.

Alexander Boyd: I’m also reading Hitler-specific these days. I’m reading Ian Kershaw’s two-part, two-volume biography of Hitler. I just finished Hubris and now I’m onto Nemesis.

Jordan Schneider: Orville, do you know Ian Kershaw? I’ve been trying to find his email address.

Orville Schell: I don’t know him but he would be great to get on and just walk him through the steps. There are some wonderful, wonderful books about that period that we need to know more about. Because you see how a slow erosion step by step, step by step with a kind of charismatic crackpot leader leading the charge and how it happens and how people just don’t rise to the occasion to stop it. They think, “Oh the courts will do it, oh something will do it,” but sometimes they don’t.

This is why I think comparing Trump to Xi is interesting and worthwhile doing. Although some of your more rigorous scholars may think there’s no data, no theoretical constructs, but for me it’s the heart of the matter. It has a lot to do with how people grow up. Autocratic leaders write themselves as very large — democratic leaders don’t have that opportunity as much. When you’re in big leader culture land, whether Putin, Kim Jong Un, Xi Jinping, Orban, whoever, it really matters who they are and where they came from and what their sort of operating system is, who installed it and when.

You can say fairly safely, although there are a lot of amazingly wonderful people in China — I have to say, and I married one — but the Cultural Revolution created massive amounts of personal, psychological, intellectual damage. It wasn’t just people got killed, people got in jail for a little while, and then Deng waved his wand and it was all over. That’s not how historical trauma works.

That’s why I find Torigan’s book so interesting. To his credit, he doesn’t do what I’ve just done, which is draw conclusions or try to draw gratuitous conclusions. He just tells the story. It’s a monumental job of research. You can draw your own conclusions, and that’s what I intend to do in Foreign Affairs.

Alexander Boyd: Any hints on those conclusions that are coming out soon?

Orville Schell: I want to make some surmises about what growing up in the Cultural Revolution meant to the formation of Xi Jinping, his form of governance today.

Alexander Boyd: According to the book, Xi Zhongxun, upon hearing of the Cultural Revolution, actually asked for his soul to be lit afire by it, which I found to be incredible research, obviously, on Torigan’s part to get this. Does that indicate that Xi Zhongxun, for whom the Party always came first, was unable fundamentally to connect with Mao because the Cultural Revolution was ongoing? He was already purged, but he yearned desperately for this. It’s kind of like a priest who doesn’t hear God’s voice calling. Is that a correct analysis?

Orville Schell: The Party — and Zhou Enlai suffered from this too — they all did. Some of them did have a sense that something was deeply awry. But there was no other show in town except the Party and the Revolution. They were veteran revolutionaries.

Xi Zhongxun, no matter what they did to him, and what they did to him was pretty horrendous, though not the worst, he never lost his belief in the Revolution and the Party. That’s what he imbued his son with. Yes, bad things happen. We can’t have chaos again. But the party is fundamentally right. The revolution cannot be questioned.

Share

It’s a classic case of where people have no other place to turn except run off into the hills if you can. And we see this in our own government now. People desperately wanting to be in the limelight, in power, in government. And they have all kinds of rationalizations. Rubio, my God, he used to think Trump was a buffoon. Now he’s sold his soul. Read Doctor Faustus.

Jordan Schneider: My favorite line with Rubio is there’s an old New Yorker profile of him where he reads The Last Lion, the Churchill biography. And he said he read it twice and that he saw himself as Churchill, like warning about the Nazis. The analogy was Iran getting the bomb or something. But to go from that to where we are today is something.

Orville Schell: Power is an incredible intoxicant. Once you get that needle in your arm, that’s your currency and that’s your realm. It’s very hard to imagine what else you’re going to do with yourself. That’s why as a writer, I’ve always said, “No, not going there.” I’m just going to stay a lowly scribe. I don’t even particularly yearn to go to China now because I know if I did yearn to go, that would circumscribe me, it would make me feel I couldn’t say certain things because I’d know there’d be consequences.

I told you this, Jordan. In 1991, I did a year-long project with 60 Minutes on forced labor and the Laogai system. It aired. It was incredible. We got into prison camps. I kept a diary of it, and I edited it and sent it to The New Yorker. They edited it and were about to go into print and I looked at it and said, “I can’t publish this.” I was a younger man, I had a Chinese wife, and I had parents-in-law in China. I threw it in a box.

I pulled it out two years ago and thought, “My God, the question of forced labor in Xinjiang is more relevant than ever.” I took it out. That will be the end of me in terms of grace from the Chinese Communist Party. But that’s okay. I’d rather that than I can’t write and say what I think. I think I was right to put it in a box then. But that’s not a healthy tendency for any society. You remember chōutì wénxué (抽屉文学), “drawer literature,” things that people could only write and put in a drawer.

Jordan Schneider: Benjamin Nathan just won the Pulitzer Prize for this really awesome book To the Success of Our Hopeless Cause: The Many Lives of the Soviet Dissident Movement, which folks should also read.

Orville Schell: Perry Link is sort of the avatar of the Chinese version of that. I recommend his book on Liu Xiaobo, and he’s just written another book, The Anaconda in the Chandelier. I think that’s a really important question.

 You know, intellectuals are poor, weak creatures, and those who stand up — read Blood Letters. Alex, will you send Jordan the Elaine Pagels program we did on the comparison of Jesus and Lin Zhao? We had them both on stage talking about the role that faith plays in adversity and revolution. We started off with Bach, a beautiful aria. I wanted Bach as the avatar of being in the mix. You’ll enjoy this, Jordan.

Jordan Schneider: All right, well, we’ll put it in the show notes as well.

Orville Schell: It’s hard to explain to people, but if you watch it, you’ll understand.

Alexander Boyd: On Orville’s note on publishing and not publishing, we just published an excerpt from Perry Link’s forthcoming book, The Anaconda in the Chandelier.

Jordan Schneider: All right, thanks so much for being a part of ChinaTalk, Orville.

Orville Schell: As always, it’s a great pleasure. You have a great program, Jordan.

of the substack did an excellent job writing up the key dynamics of the deal. I’ll be running excerpts from his post with some comments of mine interspersed.

Our government, having withdrawn the new diffusion rules, has now announced an agreement to sell massive numbers of highly advanced AI chips to UAE and Saudi Arabia (KSA). This post analyzes that deal and that decision.

It is possible, given sufficiently strong agreement details (which are not yet public and may not be finalized) and private unvoiced considerations, that this deal contains sufficient safeguards and justifications that, absent ability to fix other American policy failures, this decision is superior to the available alternatives. Perhaps these are good deals, with sufficiently strong security arrangements that will actually stick.

Perhaps UAE and KSA are more important markets and general partners than we realize, and the rest of the world really is unable to deploy capital and electrical power the way they can and there is nothing we can do to change this, and perhaps they have other points of strategic importance, so we have to deal with them. Perhaps they are reliable American allies going forward who wouldn’t use this as leverage, for reasons I do not understand. There are potential worlds where this makes sense.

The fact remains that the case being made for this deal, in public, actively makes the situation seem worse. David Sacks in particular is doubling down and extending the rhetoric I pushed back against last week, when I targeted Obvious Nonsense in AI diffusion discourse. Even within the White House, the China hawks are questioning this deal, and Sacks responded by claiming to not even understand their objections and to all but accuse such people of being traitorous decels wearing trench coats.

I stand by my statements last week that even if accept the premise that all we need care about are ‘America wins the AI race’ and how we must ‘beat China,’ our government’s policies, on diffusion and elsewhere, seem determined to lose an AI race against China.

The Central Points From Last Week

The point of the diffusion rules is to keep the AI chips secure and out of Chinese hands, both in terms of physical security and use of their compute via remote access. It is possible that the agreements we are making with UAE and KSA will replace and improve upon the functionality, in those countries in particular, of the diffusion rules.

It’s not about a particular set of rules. It is about the effect of those rules. Give me a better way to get the same effect, and I’m happy to take it. When I say ‘something similar’ below, I mean in the sense of sufficient safeguards against the diversion of either the physical AI chips or the compute from the AI chips. Access to those chips is what matters most. Whereas market share in selling AI chips is not something I am inclined to worry about except in my role as Nvidia shareholder.

I do not consider them reliable allies going forward, and there are various reasons that even the best version of these agreements would make me deeply uncomfortable, but it is possible to reach an agreement that physically locates many data centers in the Middle East and lets them reap the financial benefits of their investments and have compute available for local use, but does not in the most meaningful senses ‘hand them’ the compute in question. As in, no I do not trust them, but we could find a way that we do not have to, if they were fully open to whatever it took to make that happen.

I also would highlight the implicit claim I made here, that the pool of American advanced AI chips is essentially fixed, and that we have sufficient funding available in Big Tech to buy all of them indefinitely. If that is not true, then the UAE/KSA money matters a lot more. Then there is the similar question of whether we were going to actually run out of available electrical power with no way to get around that. A lot of the question comes down to: What would have counterfactually happened to those chips? Would we have been unable to deploy them?

[Jordan: See ChinaTalk’s recent coverage of an excellent IFP report that explored what a policy agenda to rapidly increase the energy available for AI deployment would look like]

With that in mind, here are the central points I highlighted last week:

1. America is ahead of China in AI.

2. Diffusion rules serve to protect America’s technological lead where it matters.

3. UAE, Qatar and Saudi Arabia are not reliable American allies, nor are they important markets for our technology. We should not be handing them large shares of the world’s most valuable resource, compute.

4. The exact diffusion rule is gone but something similar must take its place, to do otherwise would be how America ‘loses the AI race.’

5. Not having any meaningful regulations at all on AI, or ‘building machines that are smarter and more capable than humans,’ is not a good idea, nor would it mean America would ‘lose the AI race.’

6. AI is currently virtually unregulated as a distinct entity, so ‘repeal 10 regulations for every one you add’ is to not regulate at all building machines that are soon likely to be smarter and more capable than humans, or anything else either.

7. ‘Winning the AI race’ is about racing to superintelligence. It is not about who gets to build the GPU. The reason to ‘win’ the ‘race’ is not market share in selling big tech solutions. It is especially not about who gets to sell others the AI chips.

8. If we care about American dominance in global markets, including tech markets, stop talking about how what we need to do is not regulate AI, and start talking about the things that will actually help us, or at least stop doing the things that actively hurt us and could actually make us lose.

Diffusion Controls Have Proven Vital

Diffusion controls on AI chips we’ve enforced on China so far have had a huge impact. DeepSeek put out a highly impressive AI model, but by their own statements they were severely handicapped by lack of compute. Chinese adoption of AI is also greatly held back by lack of inference compute.

China is competing in spite of this severe disadvantage. It is vital that we hold their feet to the fire on this. China has an acute chip shortage, because it physically cannot make more AI chips, so any chips it would ship to a place like UAE or KSA would each be one less chip available in China.

Whenever you see arguments from David Sacks and others against AI diffusion rules, ask the question:

1. Is an argument for a different set of export controls and a different chip regime that still protects against China getting large quantities of advanced AI chips?

2. Or is it an argument, as it often is, that to preserve our edge in compute we should sell off our compute, that to preserve our edge in tech we should give away our edge in tech?

   1. As in, that what matters is our market share of AI chips, not who uses them?

   2. This is not a strawman, for example Ben Thompson argues exactly this very explicitly and repeatedly. Ben Thompson’s recent interview with Jensen Huang, CEO of Nvidia, made it clear both of them have this exact position. That to maintain America’s edge in AI, we need to sell our AI chips to whoever wants them, including China, because ‘China will not be held back’ as if having a lot more chips wouldn’t have helped them. And essentially saying that all Nvidia chips everywhere support the ‘American tech stack’ rather than China rather obviously turning around and using them for their own tech. He explicitly is yelling we need to ‘compete in China’ or else.

   3. Complete Obvious Nonsense talking of his own book, which one must remind oneself is indeed his job, what were you really expecting him to say? Well, what he is saying is that the way we ‘lose the AI race’ is someone builds a CUDA alternative or steals Nvidia market share. That his market is what matters. It’s full text. Not remotely a strawman.

I would disagree with arguments of form #2 in the strongest possible terms. If it’s arguments of form #1, we can talk about it.

It’s a Huge Deal

We should keep these facts in mind as we analyze the fact that the United States has signed a preliminary chip deal with the UAE. There is a 5GW AUE-US AI campus planned, and is taking similar action in Saudi Arabia. The deals were negotiated by a team led by David Sacks and Sriram Krishnan.

Lennart Heim: To put the new 5GW AI campus in Abu Dhabi (UAE) into perspective. It would support up to 2.5 million NVIDIA B200s.

That's bigger than all other major AI infrastructure announcements we've seen so far.

In exchange for access to our chips, we get what are claimed to be strong protections against chip diversion, and promises of what I understand to be a total of $200 billion in investments by the UAE. That dollar figure is counting things like aluminum, petroleum, airplanes, Qualcomm and so on. It is unclear how much of that is new.

The part of the deal that matters is that a majority of the UAE investment in data centers has to happen here in America.

I notice that I am skeptical that all the huge numbers cited in the various investment ‘deals’ we keep making will end up as actual on-the-ground investments. As in:

Walter Bloomberg: UAE PRESIDENT SAYS UAE TO INVEST $1.4T IN U.S OVER NEXT 10 YEARS

At best there presumably is some creative accounting and political symbolism involved in such statements. Current UAE foreign-direct-investment stock in the USA is only $38 billion, their combined wealth funds only have $1.9 trillion total. We can at best treat $1.4 trillion as an aspiration, an upper bound scenario. If we get the $200 billion we should consider that a win, although if the deal is effectively ‘all your investments broadly are in the West and not in China’ then that would indeed be a substantial amount of funds.

Nor is this an isolated incident. The Administration is constantly harping huge numbers, claiming to have brought in $14 trillion in new investment, including $4 trillion from the recent trip to Arabia, or roughly half of America’s GDP.

Jason Furman (top economic advisor, Obama White House): That’s nuts and baseless. I doubt the press releases even add up to that. But, regardless, press releases are a terrible way to determine the investment or the impact of his policies on it.

Justin Wolfers: Trump has claimed a $1.2 trillion investment deal from Qatar. Qatar’s annual GDP is a bit less than $250 billion per year. So he’s claiming an investment that would require every dollar every Qatari earned over the next five years.

UAE’s MGX will also be opening Europe’s largest data center in France, together with Nvidia, an 8.5 billion Euro investment, first phase to be operational in 2028. This has been in the works for a while.

Do You Feel Secure?

Not that the numbers ultimately matter all that much. What does matter is: How will we ensure the chips don’t fall literally or functionally into Chinese hands?

It comes down to the security provisions and who is going to effectively have access to and run all this compute. I don’t see here any laying out of the supposed tough security provisions.

Without going into details, if the agreements on both physical and digital security are indeed implemented in a way that is sufficiently tough and robust, if we are the ones who both physically and digitally control and monitor things on a level at least as high as domestically, and can actually have confidence none of this will get diverted, then that goes a long way.

We don’t yet have enough of that information to say.

The public explanations for the deal, and the public statements about what safety precautions are considered necessary, do not bring comfort.

I very much do not like comments like this, made in response to the Bloomberg piece above.

David Sacks (US AI Czar): If the concern [about the deal] is about diversion of advanced semiconductors to China, that’s an important policy objective but one that is easily addressed with a security agreement and a “trust but verify” approach. Modern AI supercomputers are 8 feet tall and weigh two tons; these are not like diamonds smuggled in a briefcase. To verify that chips are where they’re supposed to be, an auditor can just visit the data center and count server racks.

Semianalysis: As such, physical inspections are key to ensuring the GPUs are where they are supposed to be. The White House is aware of the risks and is prepared to ensure the chips are where they are supposed to be. As Bloomberg reported (quotes David Sacks’s statement above).

I’m sorry, what? You can ‘just visit the data center and count server racks’?

It terrifies me to see so many people arguing for the deal explicitly saying that this is a good statement, rather than a terrifying statement that we are hoping no one involved in the actual work believes to be true.

[Jordan: In 2022 I did a show with Kevin Wolf, the former head of BIS, who mentioned off-hand that there were literally only two people Commerce had in China to do site inspections. Since then BIS’ budget has barely budged and faced DOGE cuts in the early days of this administration. There is not the capacity to do datacenter inspections in a serious way today.]

Jordan Schneider: Let's start with the last thing first. This 60-day unverified list: is there a potential that the impact of this could actually be a lot bigger than what we currently see, if some of these big names on the unverified entity list end up graduating to full entity-list status?

Kevin Wolf: It's really up to the Chinese government. If they allow access, give the information that the Commerce Department is seeking, and allow the two people in China that do inspections and site visits to do their jobs…

Jordan Schneider: Whoa, wait, two? There are, like, 30 firms on there!

Kevin Wolf: Yeah, I know, but no other country has any, and the Commerce Department is the only export control organization on the planet with its own enforcement agents who are subject matter experts in export controls.

I think diversion via remote access is far more likely than physical diversion, and this response does not address the remote access issue at all, but even simply treating this as a physical diversion issue, o3, is the quoted statement accurate?

o3 (full one shot, full response quoted, in response to literally quoting Sacks without attribution and asking ‘is this accurate?’):

Short version: Mostly wrong. Chips are tiny and high-value, so diversion is hard to stop with “count the racks.” Physical audits help but they are far from sufficient.

Why “just count server racks” fails

1. High value-to-weight ratio.
   At ~$30 k per H100, a single 20 kg carry-on full of GPUs is > $1 m. That’s an attractive target. Racks are heavy; the payload isn’t.

2. Modularity.
   GPUs slide out in minutes. An auditor sees the enclosure; they cannot tell if the trays were swapped yesterday or will be tomorrow. Dummy blanks cost <$10.

3. Front companies & trans-shipping.
   Chips often leave the compliant data-center before any inspection window. Paperwork shows them installed; reality changes after the auditor flies home.

4. Remote lease risk.
   Even if the hardware stays in the U.S., Chinese entities can simply rent time on it. Export-control rules treat “deemed exports” (remote access by restricted users) as violations.

5. Limited audit frequency & competence.
   Meaningful verification requires random, intrusive, technically savvy inspections plus cryptographic attestation of firmware. Counting racks is neither.

Bottom-line probability estimate

My best estimate: ≤ 30 % that “security agreement + occasional rack counting” alone keeps advanced GPUs out of China for > 3 years. ≥ 70 % that significant leakage continues absent tighter controls (HW tracking, cryptographic attestation, and supply-chain tagging).

So the quoted claim is misleading: rack-level audits are helpful but nowhere near “easily addresses” the diversion problem.

When I asked how many chips would likely be diverted from a G42 data center if this was the security regime, o3’s 90% confidence interval was 5%-50%. Note that the G42 data center is 20% of the total compute here, so if we generously assume no physical diversion risk in the other 80%, that’s 1%-10% of all compute we deploy in the UAE.

Is that acceptable? The optimal amount of chip diversion is not zero. But I think this level of diversion would be a big deal, and the bigger concern is remote access.

I want to presume, for overdetermined reasons, that Sacks’s statement was written without due consideration or it does not reflect his actual views, and we would not actually make this level of dumb mistake where they could literally just swap the chips out for dummy chips. I presume we are planning to use vastly superior and more effective precautions against chip diversion and also have a plan for robust monitoring of compute use to prevent remote access diversion.

But how can we trust an administration to take such issues seriously, if their AI Czar is not taking this even a little bit seriously? This is not a one time incident. Similar statements keep coming. That’s why I spent a whole post responding to them.

David Sacks is also quoted extensively directly in the Bloomberg piece, and is repeatedly very dismissive of worried about diversion of chips or of compute, saying it is a fake argument and an easy problem to solve, and he talks about these as if they were reliable American allies in ways I do not believe are accurate.

Sacks also continues to appear to view winning AI to be largely about selling AI chips. As in, if G42, an Abu Dhabi-based AI firm, is using American AI chips, then it essentially ‘counts as American’ for purposes of ‘winning,’ or similar. I don’t think that is how this works, or that this is a good use of a million H100s. Bloomberg reports 80% of chips headed to the UAE would go to US companies, 20% to G42.

I very much want us to think about the actual physical consequences of various actions, not what those actions symbolize or look like. I do think, despite everything else, it is a very good sign that David Sacks is ‘urging people to read the fine print.’ This is moderated by the fact that we do not have the fine print, so we can’t read it. The true good news there requires one to read all that fine print, and one also should not assume that the fine print will get implemented. Nor do we yet have access to what the actual fine print says, so we cannot read it.

Semianalysis Defends the Deal

Dylan Patel and others at Semianalysis offer a robust defense of the deal, saying clearly that ‘America wins’ and that this benefits American AI infrastructure suppliers on all levels, including AI labs and cloud providers.

They focus on three benefits: money, tying KSA/UAE to our tech stack, and electrical power, and warn of the need for proper security, including model weight security, a point I appreciated them highlighting.

Those seem like the right places to focus, and the right questions to ask. How much of their money is really up for grabs and how much does it matter? To what extent does this meaningfully tie UAE/KSA to America and how much does that matter? How much do we need their ability to provide electrical power? How will the security arrangements work, will they be effective, and who will effectively be in charge and have what leverage?

Specifically, on their three central points:

1. They call this macro, but a better term would be money. UAE and KSA (Saudi Arabia) can make it rain, a ‘trillion-dollar floodgate.’ This raises two questions.

   1. Question one: Was American AI ‘funding constrained’? The big tech companies were already putting in a combined hundreds of billions a year. Companies like xAI can easily raise funds to build giant data centers. If Google, Amazon, Apple, Meta or Microsoft wanted to invest more, are they really about to run out of available funding? Are there enough more chips available to be bought to run us out of cash?

   2. Semianalysis seems to think we should be worried about willingness of American companies to invest here and thinks we will have trouble with the financing.

   3. I am not convinced of this. Have you seen what these companies (don’t have to) pay on corporate bonds? Did we need to bring in outside investors? Should we even want to, given these investments look likely to pay off?

   4. This is a major crux. If indeed American big tech companies are funding constrained in their AI investments, then the money matters a lot more. Whereas if we were already capable of buying up all the chips, that very much cuts the other way.

   5. Question two: As we discussed earlier, is the trillion-dollar number real? We keep seeing these eye-popping headline investment numbers, but they don’t seem that anchored to reality, and seem to include all forms of investment including not AI, although of course other foreign direct investment is welcome.

   6. Do their investments in US datacenters mean anything, and are they even something we want, given that the limiting factor driving all this is either constraints on chip availability or on electrical power? Will this be crowding out other providers?

   7. If these deals are so positive for American tech companies, why didn’t the stock market moves reflect this? No, I will not accept ‘priced in.’

2. They call this geopolitical, that UAE and KSA are now tied to American technology stacks.

   1. As they say, ‘if Washington enforces tight security protocols.’ We will see. David Sacks is explicitly dismissing the need for tight security protocols.

   2. Classically, as Trump knows well, when the bank loans you a large enough amount and you don’t pay it back, it is the bank that has the problem. Who is being tied to whose stack? They will be able to at least cut the power any time. It is not clear from public info what other security will be present and what happens if they decide to turn on us, or use that threat as leverage. Can they take our chips and their talents elsewhere?

   3. This can almost be looked at as a deal with one corporation. G42 seems like it’s going to effectively be on the UAE side of the deal, and it is going to have a lot of chips in a lot of places. A key question is, to what extent do we have the leverage on and control over G42, and to what extent does this mean they will act as a de facto American tech company and ally? How much can we trust that our interests will continue to align? Who will be dependent on who? Will our security protocols extend to their African and European outposts?

   4. Why does buying a bunch of our chips tie them into the rest of our stack? My technical understand is that it doesn’t. They’re only tied to the extent that they agreed to be tied as part of the deal (again, details unknown), and they could swap out that part at any time. In my experience you can change which AI your program uses by changing a few lines of code, and people often do.

   5. It is not obvious why KSA and UAE using our software or tech stack is important to us other than because they are about to have all these chips. These aren’t exactly huge markets. If the argument is they have oversized effect on lots of other markets, we need to hear this case made out loud.

   6. Seminanalysis points out China doesn’t even have the capacity to sell its own AI chips yet. And I am confused about the perspectives here on ‘market share’ and the implied expectations about customer lock-in.

3. They call this infrastructure, I’d simply call it (electrical) power. This is the clearly valuable thing we are getting. It’s rather crazy that ‘put our most strategic asset except maybe nukes into the UAE and KSA’ was chosen over ‘overrule permitting rules and build some power plants or convince one of our closer allies to do it’ but here we are.

   1. So the question here is, what are the alternatives? How acute is the shortage going to be and was there no one else capable of addressing it?

   2. Also, even if we do have to make this deal now, this is screaming from the rooftops, we need to build up more electrical power everywhere else now, so we don’t have this constraint again in the future.

Semianalysis also raises the concern about model weight security, but essentially think this is solvable via funding work to develop countermeasures and use of red teaming, plus defense in depth. It’s great to see this concern raised explicitly, as it is another real worry. Yes, we could do work to mitigate it and impose good security protocols, and keep the models from running in places and ways that create this danger, but will we? I don’t know. Failure here would be catastrophic.

Understanding the China Hawks

There are also other concerns even if we successfully retain physical and digital control over the chips. The more we place AI chips and other strategic AI assets there, the more we are turning UAE, Saudi Arabia and potentially Qatar into major AI players, granting them leverage I believe they can and will use for various purposes.

David Sacks continues to claim to not understand that others think that ‘winning AI’ is mostly not about who gets to sell chips, who uses our models and picks up market share, or about superficially ‘winning’ ‘deals.’

He not only thinks it is about market penetration, he can’t imagine an alternative. He doesn’t understand that many, including myself, this is about who has compute and who gets superintelligence, and about the need for proper security.

David Sacks: I’m genuinely perplexed how any self-proclaimed “China Hawk” can claim that President Trump’s AI deals with UAE and Saudi Arabia aren’t hugely beneficial for the United States. As leading semiconductor analyst Dylan Patel observed, these deals “will noticeably shift the balance of power” in America’s favor. The only question you need to ask is: does China wish it had made these deals? Yes of course it does. But President Trump got there first and beat them to the punch.

Sam Altman: this was an extremely smart thing for you all to do and i’m sorry naive people are giving you grief.

Tripp Mickle and Ana Swanson (NYT): One Trump administration official, who declined to be named because he was not authorized to speak publicly, said that with the G42 deal, American policymakers were making a choice that could mean the most powerful A.I. training facility in 2029 would be in the United Arab Emirates, rather than the United States. [Jordan: this is a wild line to be giving to the NY Times—I thought all the China hawks already got purged!]

But Trump officials worried that if the United States continued to limit the Emirates’ access to American technology, the Persian Gulf nation would try Chinese alternatives.

The hawks are concerned, because the hawks largely do not think that the key question is who will get to sell chips, but rather who gets to buy them and use them. This is especially true given that both America and China are producing as many top AI chips as they can, us far more successfully, and there is more than enough demand for both of them. One must think on the margin.

Given that so many China hawks are indeed on record doubting this deal, if you are perplexed by this I suggest reading their explanations. Here is one example.

Tripp Mickle and Ana Swanson (NYT): Mr. Goodrich said the United States still had the best A.I. engineers, companies and chips and should look for ways to speed up permitting and improve its energy grid to hold on to that expertise. Setting up some of the world’s largest data centers in the Middle East risks turning the Gulf States, or even China, into A.I. rivals, he said.

“We’ve seen this movie before and we should not repeat it,” Mr. Goodrich said.

Sam Winter-Levy, a fellow at the Carnegie Endowment for International Peace, said the huge chip sales did “not feel consistent with an America First approach to A.I. policy or industrial policy.”

“Why would we want to offshore the infrastructure that will underpin the key industrial technology of the coming years?” he asked.

This does not seem like a difficult position to understand? There are of course also other reasons to oppose such deals.

Here is Jordan Schneider of China Talk’s response, in which he is having absolutely none of it, explicitly rejecting that either America or China has chips to spare for this. rejecting that UAE and KSA are actual allies, not expecting us to follow through with reasonable security precautions, and saying if we wanted to do this anyway we could have held out for a better deal with more control than this, I don’t know why you would be confused how someone could have this reaction based on the publicly available information:

Jordan Schneider: It’s going to cannibalize US build-out and leave the world with three independent power-centers of AI hardware where we could’ve stuck to our guns, done more power generation at home, and only had China to deal with not these wild-card countries that are not actual allies. If this really is as important as we believe, why are we letting these countries and companies we deeply distrust get access to it?

The Gulf’s BATNA wasn’t Huawei chips, it was no chips. Whatever we’re trying to negotiate for, we can play harder to get. BIS can just say they can’t buy Ascends and it’s not like there’s enough capacity domestically in China to service global demand absent the TSMC loophole they charged through. Plus, we’re offering to sell them 10× the chips that Huawei could conceivably sell them anytime soon even if they use the TSMC-fabbed wafers.

Where’s the art-of-the-deal energy here? Right now I only see AMD and NVDA shareholders as well as Sama benefiting from all of this. I thought we wanted to raise revenue from tariffs? Why not charge 3× the market rate and put the premium into the US Treasury, some “Make America Great Again” industrial-development fund, use it to triple BIS’ budget so they can actually enforce the security side, put them on the hook for Gaza…I don't know literally anything you care about. How about a commitment not to invest in Chinese tech firms? Do we still care about advanced logic made in America? How about we only let them buy chips fabbed in the US, fixing the demand-side problem and forcing NVDA to teach Intel how to not suck.

Speaking of charging through loopholes, all of the security issues Dylan raises in his article I have, generously, 15 % confidence in USG being able to resolve/resist industry and politicians when they push back. If it’s so simple to just count the servers, why hasn’t BIS already done it / been able to fight upstream industry lobbying to update the chips-and-SME regs to stop Chinese build-outs and chip acquisition? What happens when the Trump gets a call from the King when some bureaucrat is trying to stop shipments because they see diversion if they ever catch it in the first place?

Why are we doing anything with G42 again? Fine, if you really decide you want to sell chips to the UAE, at the very least give American hyperscalers the off-switch. It’s not like they would’ve walked away from that offer! America has a ton to lose in the medium term from creating another cloud provider that can service at scale, saying nothing of one that has some deeply-discomforting China ties pretty obvious even to me sitting here having never gotten classified briefings on the topic.

Do the deal’s details and various private or unvoiced considerations make this deal better than it looks and answer many of these concerns? Could this be sufficient that, if looked at purely through the lens of American strategic interests, this deal was a win versus the salient alternatives? Again: That is all certainly possible!

Our negotiating position could have been worse than Jordan believes. We could have gotten important things for America we aren’t mentioning yet. The administration could have limited room to maneuver including by being divided against itself or against Congress on this. On the flip side, there are some potentially uncharitable explanations for all of this, that would be reasonable to consider.

Rhetoric Unbecoming

Instead of understanding and engaging with such concerns and working to allay them, Sacks has repeatedly decided to make this a mask off moment, and engage in a response that I would expect on something like the All-In Podcast or in a Twitter beef, but which is unbecoming of his office and responsibilities, with multiple baseless vibe and ad hominem attacks at once that reflect that he either is willfully ignorant of the views, goals and beliefs of those he is attacking and even who they actually are, or he is lying and does not care, or both, and a failure to take seriously the concerns and objections being raised. Here is another illustration of this:

David Sacks (May 17): After the Sam Bankrun-Fraud fiasco, it was necessary for the Effective Altruists to rebrand. So they are trying to position themselves as “China Hawks.” But their tech deceleration agenda is the same, and it would cost America the AI race with China.

There are multiple other people I often disagree with on important questions but whom I greatly respect who are working on in administration on AI policy. There are good arguments you can make in defense of this deal. Instead of making those arguments in public, we repeatedly get this.

Everything Sacks says seems to be about vibes and implications first and actual factual claims a distant second at best. He doesn’t logically say ‘all so-called China hawks who don’t agree with me are secret effective altruists in trench coats and also decels who hate all technology and all of humanity and also America,’ but you better believe that’s the impression he’s going for here.

Could China Have ‘Done This Deal’?

Would China have preferred to ‘do this deal’ instead? That at best assumes facts, and arguments, not in evidence. It depends what they would get out of such a deal, and what we’re getting out of ours, and also the security arrangements and whether we’ve formed a long lasting relationship in which we hold the cards.

I’m also not even sure what it would mean for China to have ‘done this deal,’ it does not have what we are offering. Semianalysis says they don’t have similar quantities of chips to sell, and might not have any, nor are their chips of similar quality.

I do agree China would have liked to ‘do a deal’ in some general sense, where they bring UAE/KSA into their orbit, on AI and otherwise, although they don’t need access to electrical power. More capital and friends are always helpful. It’s not clear what that deal would have looked like.

[Jordan: the fact that the administration is citing some old news of Huawei promising to gift just $75m in chips to the Malaysian government (and not even today! Over two years!) is policy malpractice. Plus, this administration just made clear that using Huawei Ascend chips violates US export controls, giving America an enormously powerful lever to dissuade fence-sitting countries not to buy into what is already an inferior tech offering.]

Tyler Cowen Asks Good Questions

Here’s Tyler Cowen being clear eyed about some of what we are selling so cheap. The most powerful AI training facility could be in the UAE, and you’re laughing?

Tyler Cowen: Of course Saudi and the UAE have plenty of energy, including oil, solar, and the ability to put up nuclear quickly. We can all agree that it might be better to put these data centers on US territory, but of course the NIMBYs will not let us build at the required speeds. Not doing these deals could mean ceding superintelligence capabilities to China first. Or letting other parties move in and take advantage of the abilities of the Gulf states to build out energy supplies quickly.

Energy and ability to overcome NIMBYs is only that which is scarce because America is refusing to rise to this challenge and actually enable more power generation. Seriously, is there nowhere in America we can make this happen at scale? If we wanted to, we could do this ourselves easily. We have the natural gas, even if nuclear would be too slow to come online. It is a policy choice not to clear the way. And no, I see zero evidence that we are pulling out the stops here and coming up short.

I think this frame is exactly correct - that this deal makes sense if and only if all of:

1. The security deal is robust and we retain functional control over where the compute goes.

2. We trust our friends here to remain our friends at a reasonable price.

3. We counterfactually would not have been able to buy these chips and build data centers to power these chips.

As far as I can tell China already has all the power it needs to power any AI chips it can produce, it is using them all, and its chip efforts are not funding constrained.

So for want of electrical power, and for a few dollars, we are handing over a large amount of influence over the future to authoritarian powers with very different priorities and values?

Tyler Cowen: In any case, imagine that soon the world’s smartest and wisest philosopher will soon again be in Arabic lands.

We seem to be moving to a world where there will be four major AI powers — adding Saudi and UAE — rather than just two, namely the US and China. But if energy is what is scarce here, perhaps we were headed for additional AI powers anyway, and best for the US to be in on the deal?

Who really will have de facto final rights of control in these deals? Plug pulling abilities? What will the actual balance of power and influence look like? Exactly what role will the US private sector play? Will Saudi and the UAE then have to procure nuclear weapons to guard the highly valuable data centers? Will Saudi and the UAE simply become the most powerful and influential nations in the Middle East and perhaps somewhat beyond?

Yes. Those are indeed many of the right questions, once you think security is solid. Who is in charge of these data centers in the ways that matter? Won’t they at minimum have the ability to cut the power at any time? Who gets to decide where the compute goes? What are they going to do with all this leverage we are handing them?

Is this what it means to have the future be based on American or Democratic values? Do you like ‘the values’ of the UAE and Saudi Arabian authorities?

Tyler Cowen: I don’t have the answers to those questions. If I were president I suppose I would be doing these deals, but it is very difficult to analyze all of the relevant factors. The variance of outcomes is large, and I have very little confidence in anyone’s judgments here, my own included.

Few people are shrieking about this, either positively or negatively, but it could be the series of decisions that settles our final opinion of the second Trump presidency.

The administration thinks that the compute in question will remain under the indefinitely control of American tech companies, to be directed as we wish.

Sriram Krishnan: Reflecting on what has been an amazing week and a key step in global American AI dominance under President Trump.

These Middle East AI partnerships are historic and this “AI diplomacy” will help lock in the American tech stack in the region, help American companies expand there while also building infrastructure back in the U.S to continue expanding our compute capacity.

This happens on top of rigorous security guarantees to stop diversion or unauthorized access of our technology.

More broadly this helps pull the region closer to the U.S and aligns our technological interests in a very key moment for AI.

It’s a very exciting moment and a key milestone.

I hope that they are right about this, but I notice that I share Tyler’s worry that they are wrong.

Saudi Arabia Also Made a Deal

Similarly, Saudi Arabia’s Humain is going to get ‘several hundred thousand’ of Nvidia’s most advanced processors, starting with 18k GB300 Grace Blackwells.

The justification given for rescinding the Biden diffusion rules is primarily that failure to do this would have ‘weakened diplomatic relations with dozens of countries by downgrading them to second-tier status.’

But, well, not to reiterate everything I said last week, but on that note I have news.

One, we’re weakening diplomatic relations with essentially all countries in a series of unforced errors elsewhere, and we could stop.

[Jordan: that feeling when you treat Dubai better than Canada]

Two, most of the listed tier two countries have always had second-tier status. There’s a reason Saudi Arabia isn’t in Five Eyes or NATO. We can talk price about which countries should have which status, but no our relations are not all created equal, not when it comes to strategically vital national interests and to deep trust. I don’t share Sacks’s stated view that these are some of our closest and most trustworthy allies. Why does this administration seem to always want to make its deals mostly with authoritarian regimes, usually in places where Trump has financial ties?

Tripp Mickle and Ana Swanson (NY Times): The announcements of the two deals follow reports that $2 billion has flowed to Trump companies over the last month from the Middle East, including a Saudi-backed investment in Trump’s cryptocurrency and plans for a new presidential airplane from Qatar.

There’s always Trust But Verify. The best solution, if you can’t trust, is often to set up things so that you don’t have to. This can largely be done. Will we do it? And what will we get in return? What is announced mostly seems to be investments and purchases, that what we are getting are dollars, and Bloomberg is skeptical of the stated dollar amounts.

At Best A Second Best Solution

This deal is very much not a first best solution. It is, at best, a move that we are forced into on the margin due to our massive unforced errors in a variety of other realms. Even if it makes sense to do this, it makes even more sense to be addressing and fixing those other critical mistakes.

Electrical power is the most glaring in the context of this particular. There needs to be national emergency level focus on America’s inability to build electrical power capacity. Where are the special compute zones? Where are the categorical exemptions? Where is DOGE with regard to the NRC? Where is the push for real reform on any of these fronts? Instead, we see story after story of Congress actively moving to withdraw even the supports that are already there, including plans to outright abrogate contracts on existing projects.

The other very glaring issue is trade policy. If we think it is this vital to maintain trade alliances and open up markets, and maintaining market share, why are we otherwise going in the opposite direction? Why are we alienating most of our allies? And so on.

The argument for this deal is, essentially, that it must be considered in isolation. That other stuff is someone else’s department, and we can only work with what we have. But this is a very bitter pill to be asked to swallow, especially as Sacks himself has spoken out quite loudly in favor of many of those same anti-helpful policies, and the others he seems to be sitting out. You can argue that he needs to maintain his political position, but if that also rules out advocating for electrical power generation and permitting reform, what are we even doing?

If we swallow the entire pill, and consider these deals only on the margin, without any ability to impact any of our other decisions, and only with respect to ‘beating China’ and ability to ‘win the AI race,’ and assume fully good faith and set aside all the poor arguments and consider only the steelman case, we can ask: Do these deals help us?

I believe that such a deal is justifiable, again on the margin and regarding our position with respect to China, if and only if ALL of the following are true:

1. Security arrangements are robust, the chips actually do remain under our physical control and we actually do determine what happens with the compute. And things are set up such that America retains the leverage, and we can count on UAE/KSA to remain our friends going forward.

2. This was essentially the best deal we could have gotten.

3. This represents a major shift in our or China’s ability to stand up advanced AI chips, because for the bulk of these chips either Big Tech would have run out of money, or we would have been unable to source the necessary electrical power, or China has surplus advanced AI chips I was not previously aware of and no way to deploy them.

4. Entering into these partnerships is more diplomatically impactful, and these friendships are more valuable, than they appear to me based on public info.

Check out some ChinaTalk coverage from least year about China’s Middle East AI ambitions.

Mood Music