Italian Police Uncover Dead Mob Boss’s $230 Million Business Empire

© Carabinieri, via Associated Press

© Carabinieri, via Associated Press

© Yuki Iwamura/Associated Press

和去年一样,在正式的 Google I/O 开发者大会之前,谷歌为 Android 单独开了一次小型发布会。
本次 Android Show 上,谷歌几乎是一股脑将未来一年有关 Android 和整个 AI 产品生态的「宏愿」抖了出来。

除了作为基底的 Android 17 之外,我们这次还迎来了一些意料之外的平台更新和几款硬件产品的发布。
重点在于:虽然产品不多、距离 Android 17 广泛推送还有一段时间,但我们已经足以看到谷歌未来五年甚至十年,对于人工智能生态的计划了。
更要命的是,今晚的谷歌,刚好是苹果梦想里进入 AI 时代之后的自己。
活动刚开场,谷歌就宣布了一项意义重大的举措:
Android 将不再是一个单纯的操作系统(Operating System),而是一个智能系统(Intelligence System)。

图|Google
换句话说,曾经以「用户手动操作」为绝对主流的软硬件使用方式,在谷歌看来已经是上一个时代的符号了。
从今天起,Android 作为一个智能功能的集群,会更加主动地介入到用户操作流程的更前端,想你所想、做你所做。

图|Google
这也是为什么在本次活动上,Android 17 这个具体的系统版本出现的次数屈指可数,Gemini Intelligence 却成为了系统的代名词。
首先,Gemini Intelligence 作为谷歌 Gemini 在手机硬件上的最新形态,极大强化了它多模态、跨环境、高度整合的运行模式。
比如作为系统输入的第一入口,默认键盘 Gboard 就得到了一次功能强化。
基于 Gemini Intelligence 的多模态能力,原本在键盘功能中存在感不高的「自动填入」功能极大地拓展了它的信息来源:

图|Google
除了手动保存的各种密码之外,Gboard 还将会支持自动填入图库里面的证照信息、聊天提到的地址信息、邮件撰写的日程信息等等。
更直白地说:Gemini Intelligence 已经远超「帮忙记住密码」的水平,而是真正像个助理一样帮你记住和建议各种来源、各种类型的信息。
另一种有效利用这些多模态信息的方式,则是 Android 的桌面小组件(widget)。
在 Android 17 中,Gemini Intelligence 将会支持一项名为「Create my widget」的功能,但不是第一时间上线、而是目标今年晚些时候。

图|Google
这个新功能主要做的,就是用类似 vibe coding 的模式,根据你的指令在桌面上创建新的小组件,打破了小组件只能是 app 预置的那些。
举例来说,相比功能单一的记录卡路里的 app,我可以和 Gemini 说:做一个每周工作日向我推荐两次高蛋白餐的小组件。

图|Google
这样一来,桌面小组件就真正变成了一项复合任务的入口,本质上和人 vibe coding 一个 app 的性质是完全相同的。
此外,喜欢语音输入的用户也有福了——新版 Gboard 将会支持类似 Typeless 的高智能化语音输入功能,名字叫做 Rambler。
相比以前要亲口说「逗号…句号…」,Rambler 可以将一整段充满了「嗯嗯啊啊」的口述转译、清洗、整理成一段整洁的文字:

图|Google
另一方面,Gemini Intelligence 的自动执行功能也得到了进一步加强。
去年的 Google I/O 和发布会上,谷歌演示过给 Gemini 下命令,让它自动帮你点外卖、叫车、订票之类的操作,正式上线之后反响不错。
而在 Android 17 中,Gemini 升级成 Gemini Intelligence,这种「代操作」也支持多步骤任务了。
比如以前只支持简单的「帮我订一张票」,你现在可以在 Gemini 对话框里直接拍下旅游宣传册,和 Gemini 说「在携程上帮我找一个类似的双人团行程」:

图|Google
重点不在于 Gemini 能够执行什么任务,而是它拥有了更强大的「多做一步」的能力,有时候就是多的这一步,让 AI 从「能用」变成了「有用」。
当然 Android 17 的更新也不是 Gemini 的独角戏,谷歌同样对很多「Android 核心体验」进行了优化。
在 Android 17 中,谷歌和 Meta 达成了合作,在 Facebook、Instagram 等等 app 里支持了调用原生相机功能,比如 Ultra HDR、超级防抖、夜景视频等等。
而谷歌使用了好多年的平面风格 emoji 也迎来了一次更新——从原本的纯 2D 变成了 2.5D,在风格上更接近 iOS 使用的 emoji 了:

图|Google
而我们此前介绍过的 QuickShare 兼容 AirDrop 的功能,也将在 Android 17 上支持更多厂商的设备。
除了三星和 Pixel 之外,(国际版)OPPO、Vivo、一加和荣耀的较新机型也将在今年下半年陆续更新兼容 AirDrop 的固件:

图|Google
在 Gemini Intelligence 之外,谷歌也没有忘记给这些更复杂、更强大的 AI 功能打造一套量身定制的硬件。
这个新硬件的形态,既不是吊坠,也不是耳机,更不是手表手环——而是曾经的 Chromebook。
没错,在 AI 时代,谷歌又双叒给自己的笔记本改名了。
从 Pixelbook,到 Chromebook,再到最新的 Googlebook:

图|Google
和 Chromebook 一样,Googlebook 并不是某一款具体的笔记本电脑,而是同样和第三方厂商合作、只要符合标准的都可以叫这个名字。
而 Googlebook,就是「第一款为 Gemini Intelligence 量身打造」的硬件产品。
除了上面的全新 Gemini Intelligence 功能之外,Googlebook 在日常使用最频繁的基础人机交互层面,做出了堪称革命性的创新——
在 Googlebook 上呼出 Gemini 功能,既不需要说话、也不需要按键、更不是右键菜单,只需要「摇一摇光标」就行。

图|Google
基于 Gemini Intelligence 的多模态能力,推荐的 AI 指令甚至可以根据光标下面的内容、选中的内容、屏幕上可以进行的操作等等因素自动调整。
在如今电脑端 AI 功能越来越密集、笔记本键盘空间不够充裕的情况下,Googlebook 的「魔法指针」无疑是最直观且优雅的解决方案之一。
此外,Googlebook 还解决了 ChromeOS 历史上的老大难问题:它是谷歌的产品,却跑不了 Android app。
换句话说,所有 Android 手机里面的 app,在 Googlebook 上都可以直接运行,基本看齐了如今 macOS 跑 iOS 软件的水平。
这一切的基础,就是爱范儿之前文章中提到的谷歌大力推行的 GKI(通用内核镜像)计划,正在让 Android 脱离手机的桎梏、无缝衔接到更多形态的设备上。
虽然谷歌目前没有提到这个功能的兼容情况,但我们猜测,依据处理器规格和网络状态,Googlebook 应该同时支持本地运行和画面投屏手机 app 两种方式。

图|Google
根据活动消息,首批 Googlebook 的生产厂商还是那几个熟悉的身影:宏碁、华硕、戴尔、惠普、联想等等,首批产品预计在今年内上市。

图|Google
除了笔记本电脑,Android 17 同样更新了一部分 Android Auto 车机系统的功能。
比如更像 OpenClaw 能力的 Gemini Intelligence 代操作、优化的 3D 道路画面、更加智能化的流媒体播放功能等等。
当然也支持把手机上自创的自定义 widget 显示在车机上。

图|Google
同时,原生支持 Android Auto 的品牌范围也在增加,部分型号甚至支持记忆当前车辆信息,类似后备箱尺寸、仪表盘规格等等。
这样一来,用户在使用 Gemini 问答的时候,车机就能给出具体回答,比如「能不能同时放俩 27 寸旅行箱?」或者「那个像是刺客的警示灯是什么意思」之类的。
可惜的是,这项功能目前也不会立即上线,同样预计为「今年晚些时候」才会有产品搭载。
总的来说,本次活动只是今年 Google I/O 的开胃菜,但它涉及到的理念变革却是非常根本性的——
其实在活动的开头,谷歌就指出了:好用的人工智能技术,就应该是让人感受不到的,它会融入进每一层软件和硬件的体验。

图|Google
而这正是 Gemini Intelligence 在做的。
无论是 Pixel 手机、Android Auto 车机还是 Googlebook,这些硬件最终都只是 Gemini 智能的一种体现方式而已。
值得玩味的是——谷歌今天晚上所做的,刚好就是苹果削尖脑袋想要实现的那套 AI 生态。
让 iPhone、手表和 Mac 共用一套智能体系,用户无论在哪里使用,功能和体验都是高度相似的,硬件只区分交互方式、不影响智能水平。

图|Apple
可惜的是,苹果挣扎了这么久,也没有搞定「模型」的部分,反而让自己的硬件成了别家模型的嫁衣。
将来的智能系统(Intelligence System),形式比现在更多样、但核心却比现在更加统一。
#欢迎关注爱范儿官方微信公众号:爱范儿(微信号:ifanr),更多精彩内容第一时间为您奉上。
Many of the great buildings, their friezes and statues of classical times, and in the Middle Ages, were painted, a practice known as polychrome. This article looks at some more modern examples made by those who also painted flat surfaces.

Lawrence Alma-Tadema’s beautiful painting of Phidias Showing the Frieze of the Parthenon to his Friends from 1868 shows a group including Pericles (at the right), Aspasia, Alcibiades and Socrates, admiring this famous frieze in the glory of its original colours.

This magnificent polychrome limewood sculpture shows The Last Judgement (c 1520) as an outbreak of plague, and may be based on descriptions of the Black Death.
Much of the output from the workshops of Spanish painters during the seventeenth century wasn’t two-dimensional painting, but polychrome wood carvings destined for religious use.

Juan de Mesa (1583–1627) was a Cordoban sculptor who trained and worked in Seville from 1606 to his death there in 1627. He was responsible for many of the processional effigies which were – and some still are – featured in the celebrations of Holy Week in Seville. The Immaculate Conception is a fine example of his work from about 1610-15.
Many painters have also sculpted, but one specialised in polychrome figures, Jean-Léon Gérôme. Not only that, but he made paintings of him working on his sculptures, and of polychrome sculptures more generally.

Gérôme’s The End of the Pose (1886) is the first of a series of unusual compound paintings, which are at once self-portraits of Gérôme as a sculptor, studies in the relationship between a model and their sculpted double, and forays into issues of what is seen, visual revelation, and truth.
Here, while Gérôme cleans up, his model is seen covering up her sculpted double with sheets, as she remains completely naked. Apart from various diversionary entertainments, including a couple of stuffed birds and a model boat, there is a single red rose on the wooden platform on which the model and statue stand.
It’s not far from here to the myth of the perfect sculptor, Pygmalion, whose creation was given life, turning statue into lover, which was Gérôme’s next step in 1890.

Gérôme’s finished Pygmalion and Galatea (c 1890) just managed to stay on the right side of what in the day was deemed decent. His attention to detail is, as always, delightful, with two masks against the wall at the right, Cupid ready with his bow and arrow, an Aegis bearing the head of Medusa, and a couple of statues about looking and seeing.

In 1874, archaeologists discovered large numbers of polychrome cast terracotta figurines in the Boeotian town of Tanagra. These dated from Greek civilisations active there during the late fourth century BCE, and their quantity and quality impressed many, including Gérôme.
Having painted the story of Pygmalion and Galatea in 1890, Gérôme moved on to two works in which he built imaginary worlds around the Tanagra figurines. Sculpturae Vitam Insufflat Pictura (Painting Breathes Life into Sculpture) (1893) combines his own love of polychrome sculpture with a celebration of those archaeological discoveries.
In doing so, Gérôme progresses his long-running theme of visual revelation and truth, with these painted miniature humans, mimicking reality, and the wooden box of masks in the foreground. Several of the figurines refer to his own painting and sculpture.

As a painting, I don’t think that his Atelier of Tanagra (1893) is as effective and convincing as the previous work, but its wider range of figurines gives deeper insight into what Gérôme was thinking about. Several of the figurines shown here are his own polychrome sculptures, most obviously that of the naked woman sitting on the top of a blue pillar, in the centreline of the painting.

In The Artist’s Model (1895), Gérôme paints himself at work on his marble figure Tanagra (1890), currently in the Musée d’Orsay, which he had already included among the figurines in his Sculpturae Vitam Insufflat Pictura. He thus painted himself making a sculpture which he had previously painted in a painting as a sculpture. Scattered in the image are reminders of gladiatoral armour and other props used for his paintings, one of his paintings of Pygmalion and Galatea, together with one of his polychrome sculptures of a woman with a hoop, at the right edge.

In 1894, Jean-Léon Gérôme started work on this polychrome marble head of the famous actor Sarah Bernhardt (1894-1901), which he completed in 1901. He bequeathed it to the French nation.

In about 1902 Gérôme returned to his series of paintings of himself as a sculptor. His Self-Portrait Painting The Ball Player is a fascinating variation of the traditional form of self-portrait, in that he is here applying the colour to one of his polychrome sculptures, a figure of a ball player, who closely resembles those seen earlier in his paintings of sculptures, including Pygmalion and Galatea.
Georges Lacombe was the sculptor among the Nabis, who painted Breton motifs and sculptures he created in wood.

Some of Lacombe’s sculptures are stunning: this wood carving in mahogany with added colour shows the goddess Isis, and was made in about 1895. It’s now in the Musée d’Orsay. Lacombe’s symbolism here may refer to the role of this Egyptian goddess in producing and protecting the heir to Osiris, Horus.
(Abstract)有大神开发出了可以在 macOS 上原生运行的版本。 Few accounts of painting in the last decades of the nineteenth century consider the importance of photography at the time. Yet photography was enjoying remarkable technical advances: in 1884, George Eastman started replacing glass plates with light-sensitive film, and four years later he launched the first Kodak camera, the predecessor of the Kodak Brownie that was to follow in 1901. As Naturalist painting was taking the Salon by storm, its rival was becoming more widely available, and no longer required a private chemistry lab.
First responses to photography by painters were often hostile.

Philipp Sporrer’s The Photo (1870) is probably the most pointed painted propaganda. The young photographer is not the sort of man you would leave your wife or daughter with. He’s down at heel, unkempt, and his straw hat is abominably tatty. His studio is poorly-lit, probably an old shed, its floor littered with rubbish, and its window broken. His subject is manifestly poor and uncouth, sitting in ill-fitting clothes and picking his nose as he waits for the photographer to fiddle with his equipment.

Pascal Dagnan-Bouveret’s A Wedding at the Photographer’s (1879) seems more calculated. Hugely successful at the Salon, this artist saw no threat from wedding photography, a market in which there was no competition between painting and photography. But he still takes the opportunity to show the photographer and his studio as being tatty and tawdry.
Gradually, painting started to become influenced by the nascent art of photography, most obviously in the use of views through the lens of a camera.

Gustave Caillebotte’s major painting of 1875 shows three workmen preparing a wooden floor in the artist’s studio at 77 rue de Miromesnil. It’s thoroughly detailed, Realist, and despite its innovative view and unusual subject, it conformed to the highest standards of the Salon at the time.
Caillebotte was hurt and angry when he was informed that this painting had been rejected by the Salon jury. The grounds given seem extraordinary now: apparently the jury was shocked at this depiction of the working class at work, and not even fully-clothed. It was deemed to have a ‘vulgar subject matter’ unsuitable for the public to view. Or was it really because of his wide-angle photographic effect?

Caillebotte was one of the first established painters to experiment with photography, as demonstrated in another wide-angle view of Paris Street, Rainy Day from 1877.

During his development of Naturalism, Jules Bastien-Lepage arrived at a compositional formula that achieved similar effects, as seen in his Haymakers or Hay making in the same year, with its high horizon and fine detail in the foreground. Together these also give the visual impression that the whole canvas is meticulously realist, although in fact much of its surface consists of visible brushstrokes and other more painterly forms.

Eugène Burnand’s magnificent painting of Bull in the Alps from 1884 is fascinating for his use of both optical effects and extreme aerial perspective. Not only are there marked contrasts between the foreground and background in terms of chroma, hue and lightness, but Burnand has used defocussing in a photographic manner. The crisp edges of the bull stand proud of the softer edges and forms in the mountains behind. It’s worth noting that Burnand had been a pupil of Jean-Léon Gérôme.

Paul Louis Martin des Amoignes’ painted his wonderful In the Classroom two years later, in 1886. It bears unmistakeable evidence that it was either painted from photographs or strongly influenced by them. One boy, staring intently at the teacher in front of the class, is caught crisply, pencil poised in his hand. Beyond him the crowd of heads becomes more blurred.
By the 1890s, more painters were experimenting with photography.

Among them was Edgar Degas. This is an albumen print of patron and amateur painter Henry Lerolle with two of his daughters, Yvonne and Christine, taken by Degas in 1895-96.
The realist painter Jean-Léon Gérôme not only experimented with photography for many years, but was an enthusiastic advocate for its recognition as an art in its own right.

Gérôme’s Truth Coming out of her Well to Shame Mankind (1896) is based on a quotation attributed to Democritus, “Of a truth we know nothing, for truth is in a well” (or, more literally, ‘in an abyss’). Gérôme used the same allusion in his preface to Émile Bayard’s posthumous collection of collotype plates of photographs of nudes, Le Nu esthétique. L’Homme, la Femme, L’Enfant. Album de documents artistiques inédits d’après Nature, published in 1902, where he wrote:
Photography is an art. It forces artists to discard their old routine and forget their old formulas. It has opened our eyes and forced us to see that which previously we have not seen; a great and inexpressible service for Art. It is thanks to photography that Truth has finally come out of her well. She will never go back.
The Naturalist painter Jules-Alexis Muenier became a photographer by the time he travelled to North Africa with Pascal Dagnan-Bouveret in 1888, armed with cameras.

Although I have been unable to find a suitable image of the painting, this photograph shows Muenier with his painting of The Harpsichord Lesson in about 1911, which became his most famous work during his lifetime. Muenier, Gérôme and Dagnan-Bouveret weren’t just happy snapper photographers, but believed in photography as fine art. All three were early members of local photographic clubs, and Muenier and Dagnan-Bouveret exhibited their photographs as seriously as their paintings.

一台 14 或 16 英寸的笔记本电脑,将几十上百 GB 内存直接封装进 SoC,实现超过 200 GB/s 的高性能内存带宽,还有轻薄的机身和安静又狂暴的性能……
你可能以为这是 MacBook Pro——但如果我告诉你,这是一台 ARM 架构的轻薄型 Windows 本呢?
4 月 27 日,华硕发布了灵耀 16 Air 的骁龙版,搭载的是高通骁龙 X2 Elite Extreme 平台,也即高通去年推出的第二代 Windows on ARM 处理器。
这是第一颗将 LPDDR5X 内存做进 SoC 封装的骁龙旗舰 PC 平台,是与苹果「统一内存架构」理念一致、执行接近的平行方案。尽管没能做到 M 芯片的百分百效果,仍然是高通在这条新路上,最关键的一次尝试。

这台华硕灵耀 16 Air 骁龙版,整机 1.2kg、厚度 13.9mm,48GB 内存(频率 9523 MT/s),可提供 20-30 小时续航。机器于 4 月 28 日京东首发,售价 13999 元。华硕同时也有 14 寸版本提供。
同期亮相的还有面向创作者的 ProArt 创 X 2026 二合一笔记本,重 0.82kg、提供 22 小时续航与 2.8K 144Hz OLED 屏。这些机型共同组成了华硕在 ARM Windows 阵营的 2026 全新产品矩阵。
回到顶配 X2 Elite Extreme 的共享内存架构:将内存放进芯片封装内,放到 CPU、GPU 和 NPU 的身边,并不只是改了改电路板布局。实际上,整个计算资源调度的方式,都发生了很大的改变。
苹果在 2020 年的 M1 芯片开始,不仅将内存封装进 PC 级芯片,更让调度变得更加灵活,内存反复读写的次数要求有所降低,结果就是让内存带宽暴增——称为统一内存架构。今年 3 月发布的 M5 Pro 和 M5 Max,则更是将内存带宽推到了 307 GB/s 和 614 GB/s。
骁龙 X2 Elite Extreme 是 Windows on ARM 笔记本第一次通过内存内封装的思路,让 1.2 公斤左右的轻薄本也可以享受类似于统一内存架构带来的快乐。
这背后,是高通和华硕等各大 OEM 一起,想让 Windows 笔记本追上 MacBook Pro 的企图。
需要注明的是,「统一内存架构」是苹果使用的说法,高通官方称自己的方案为 SiP(System-in-Package)。
两者所指不完全相同:UMA 描述的是内存访问架构,SiP 则指的是具体的封装技术。但它们的实现效果和追求目标高度一致——共享物理内存池、跨 IP 块缓存一致。
可用于算力密集型任务(比如 AI 推理)的「显存」上限,直接等于整机的内存上限。哪怕是一台 48GB 的轻薄本,理论上也可以本地运行数百亿参数级别的大模型,这在传统架构上需要工作站级独显,采用集显的轻薄本很难做到。(X2 Elite Extreme 最高 SKU 为 128GB 共享内存。)
系统级缓存(SLC)可以在 CPU、Adreno X2 GPU、Hexagon NPU 之间动态分配,比上一代带宽高 70%;192-bit 内存总线搭配 LPDDR5X-9523,能够实现高达 228 GB/s 的C/G/NPU 共享内存带宽。
而传统的混合计算负载(同时依赖 C/G/NPU),被内存搬运所掣肘的情况,也得到了极大缓解。并且,整机功耗也能维持在轻薄本可以接受的水平。
更值得一提的是,这一代 Hexagon NPU 还专门把 DMA 单元升级到 64 位虚拟寻址,让 NPU 终于可以访问超过 4GB 的内存,一定程度上突破了 NPU 坐端侧大模型推理任务的瓶颈。
这的确不是 Windows 阵营第一次试水类似统一内存架构的方案,在此之前,英特尔、AMD 都做过尝试(稍后会详述)。
不过在今天,华硕灵耀 16 Air 骁龙版的高配机型,是 Windows 阵营里首个最大限度接近统一内存架构效果,并且还做到 1.2 公斤左右 ARM 轻薄本上的方案。

在共享/统一内存架构的道路上,每家芯片巨头对的判断都不一样,首先是工程问题,更深一层是商业问题。
一名在某芯片巨头供职的专家告诉爱范儿,行业里无人质疑统一内存架构的优秀,但做与不做,能否持续做,分歧在于厂商对性能目标和成本之间的平衡。
在 X2 顶配 SKU 上,高通目前的看法是:将统一内存架构所解锁的强大性能,交给给到真正需要它的硬核用户,特别是那些工作流里重度依赖 AI 模型/AI 功能的专业用户和创作者,这件事值得花成本去做。
再看英特尔,在上一代 Lunar Lake 架构上做过类似尝试,然而成本炸裂难以控制,不得不终止。英特尔前 CEO Pat Gelsinger 在财报会上明确将该次尝试定义为「one-off」,理由是封装内存把毛利压得太低。
今年 1 月发布的 Panther Lake 机型则回归了传统外置内存路线,据信后续的 Nova Lake 架构也将延续老的策略。英特尔仍然在高端 AI 笔记本市场上占有一席之地,但可以说短期内不会再走统一内存架构这条路了。
AMD 那边,Ryzen AI Max+ 395(Strix Halo)同样采用类似的共享内存架构架构,最高 128GB 板载 LPDDR5X,能够实现高达 256 GB/s 内存带宽,比 X2 Elite Extreme 还激进。
正因为此,在 AMD 的定义下 Strix Halo 属于移动工作站芯片,搭载的笔记本价格都更高,形态也更厚重,抑或是搭载于迷你工作站,不在个人笔记本电脑消费者的选购范围内。
三家芯片厂商,三种不同答案。骁龙 X2 Elite Extreme 消费级笔记本在这个时间点正式面市,虽然很难说撞上了换机窗口(毕竟今年的内存实在太贵),但至少填补了消费级市场的真空。
老实说,骁龙 X2 Elite Extreme 目前也只是跟苹果那边的 M5 基础款能打个有来有回,跟 M5 Pro/Max 这样的工作站级「顶级牛马」距离还比较远。
最直接的差距在于内存带宽的极限值:X2 Elite Extreme 的带宽宣传值能够达到 228GB/s,是 M5 Max 的 ⅓ 左右,比 M5 Pro 的 ⅔ 多一点。
当然还是要给 X2 挽尊一下,这一代仍然是单 die(晶粒),内存带宽存在物理上限。
而苹果在 M5 Pro/Max 这一代用上了新的「融合封装」,也即将两块 die 拼到一起,把内存总线扩展到更高。
在最直接的大模型推理任务上,内存带宽差距直接意味着 token 吞吐速度的差距;在 4K/8K 等极高清的视频剪辑和 AI 处理任务上,或者在其他工程软件的算力密集型任务上,也会有明显体现。
不过至少,Windows 平台在这些专业/工业软件的兼容性上是要比 macOS 好的……
我想,骁龙把共享内存架构带进消费级 Windows 笔记本市场,这件事的意义讨论或许不应该局限于性能数字上谁暴打谁,
而在于 Windows 平台用户不应该一直享受「二等公民」的体验。
即便是一台不超过 1.5 公斤的大屏轻薄本,仍然可以提供远比其它 Windows 性能本更好的 AI 算力,而且仍能保住轻薄本应该有的功耗优势——这,才是更重要的。

当然,围绕在 Windows on ARM 周围的种种问题,比如软件生态、x86 模拟层稳定性、专业软件适配等等,仍然无法被共享内存一劳永逸地解决。
从芯片厂,到微软,再到 ISV,大家都在加紧马力。比如 Photoshop、Lightroom 已经能够稳定运行 ARM 原生版本;达芬奇也早在两年前就完成了 Windows on ARM 的原生支持,甚至比 Adobe 还早。
但软件生态兼容仍有不完美之处,比如 Adobe AE 的部分渲染器和工作流仍然只能在 x86 平台上使用;Blender 的一些渲染功能在 ARM 架构上也会性能打折。
这是一个软件追硬件的时代。只有 X2 这一代能够让足够多用户,特别是创作者和专业用户,真正将骁龙本纳入主力机考虑——ARM 生态才会进入「用户越多适配越多,适配越多用户越多」的正反馈。
苹果也走过同样的路,所以这绝非不可能完成的任务。
#欢迎关注爱范儿官方微信公众号:爱范儿(微信号:ifanr),更多精彩内容第一时间为您奉上。
A mask is a cover worn over part or all of the face. Covering the face more generally has been considered in this previous article. Here I’ll concentrate on masks that represent the face, and those used in the theatre, carnivals and masked balls.
Masks have been associated with acting and the theatre since ancient times, and can be seen as symbolic of drama and the stage.

Georges Rochegrosse painted this Portrait of Sarah Bernhardt in about 1900, including three masks taken from Japanese Noh theatre at the left edge. Bernhardt was one of the most famous actresses of this period.
Some of the most remarkable archaeological discoveries of the late nineteenth century were unearthed at Tanagra in Greece around 1874, and included numerous painted terracotta figurines that Jean-Léon Gérôme featured in several of his paintings.

Sculpturae Vitam Insufflat Pictura (Painting Breathes Life into Sculpture) (1893) is a combination of a manifesto for Gérôme’s own polychrome sculpture, and a celebration of those discoveries at Tanagra. In addition to painted figurines, there’s a wooden box of masks in the foreground, and some on the shelves behind.

Gérôme’s Atelier of Tanagra from the same year includes a wider range of figurines and masks in a similar setting. These are typical of classical masks, where the face with mouth agape stands for tragedy, and the smiling face for comedy.

If his painting of The Artist’s Model from 1895 is to be believed, Gérôme kept several masks as props in his own studio, where he’s working on full-size sculpture of his model Emma.

Emil Orlík painted this unusual Model in 1904. She appears to have become shy or distressed. Her gown, screen, floor and hanging masks are evidence of the artist’s Japonisme.
Masks worn for balls seldom covered the whole face, and their primary purpose has been to disguise identity, thus to encourage extra-marital relationships, even promiscuity.

Marià Fortuny’s Masquerade (1868) is a marvellously loose watercolour showing an open-air masked ball, presumably held in Italy in the autumn, which is arousing the interest or bemusement of two swans. Dress is liberal to say the least, with the masked woman in the centre baring her breasts and holding a parasol.

Despite its title, Lovis Corinth’s Witches (1897) isn’t a depiction of sensuous rites taking place in a coven. Instead the women are preparing a younger woman to attend a masked ball. Their subject has just got out of the wooden tub in the foreground, has been dried off, and is about to don the fine clothes laid over the chair at the left, including the black mask.

Fernand Pelez’s La Vachalcade (The Cow-valcade) from 1896 shows young revellers taking part in a carnival procession, perhaps one of those in Montmartre at the time. Several are wearing full masks.
Once Gustave Moreau had recovered from the trauma of the Franco-Prussian War, he started on his next major works. He had been invited to join the select group of artists who were engaged to paint murals in major public buildings, but declined. Despite that, in 1875 he was made a member of the Legion of Honour, which pleased him and his mother deeply. All he needed now was another success at the Salon.
One of his four paintings exhibited at the Salon of 1876 told the second of Hercules’ twelve labours, his battle with the Lernean Hydra.

The Hydra was a poisonous monster with the body of a dog and multiple serpent heads, whose breath alone could kill. According to surviving written accounts, Hercules covered his mouth and nose with a cloth for protection from the deadly fumes, fired flaming arrows into the Hydra’s lair to awaken it, then set about trying to kill it.
When he discovered that cutting off its heads with a sickle or sword only resulted in two more growing back, Hercules enlisted the help of his nephew Iolaus, who cauterised the wounds with a firebrand to prevent regrowth. Hercules then cut off the one immortal head using a golden sword given to him by Athena. He also took some of the Hydra’s blood, which was the poison used on the arrow with which he later killed Nessus.
Moreau puts his canvas into its portrait orientation to emphasise the Hydra towering over Hercules, who is fully armed, with club, bow and arrows, and more. The moment chosen is the initial confrontation, with Hercules staring steely-faced at the Hydra. This is consistent with Moreau’s aversion to a more theatrical treatment.
This was well-received and extensively debated, even generating a long-standing controversy over its possible political connotations. It was suggested at the time that the Hydra represented the forces of anarchy behind the insurgency of the Commune in 1871. Others preferred instead that the Hydra represented Bismarck and the German princes behind the Franco-Prussian War of 1870.
Two of his other paintings shown in 1876 were based on the story of Salome and the execution of Saint John the Baptist. Their underlying narrative is biblical, and straightforward. The unnamed daughter of Herodias (subsequently named as Salome) performed a dance at a birthday feast thrown by King Herod. The dance so pleased Herod that he offered her anything that she wanted, up to half his kingdom. She asked not for riches, but for the head of Saint John the Baptist, the earthly messenger sent to announce the birth and ministry of Jesus Christ. Reluctantly, Herod agreed, John was beheaded in prison, and his head brought to her on a plate; the dancer gave the head to her mother.
This has been a popular story for religious paintings, and by far the most common scene involves John’s head being brought on a plate, or variations around that. Moreau was clearly interested in other parts of the story, and in Salome herself. Moreau’s apparently sudden interest in Salome was sparked by the story, probably mythical, of a woman Communard known as the pétroleuse, who seemingly took delight in setting buildings alight. That suggests it wasn’t until the summer of 1871 that he started work on his paintings of Salome.

The culmination of Moreau’s quest for the right scene to show the story of Salome the dancer is this extraordinary oil painting shown at the 1876 Salon.
The cadaveric King Herod sits on this throne while Salome is almost static on her points, and pointing towards the right. The executioner stands at the foot of the throne, and a couple of other women (including, perhaps, Salome’s mother) are at the left. Salome holds a lotus flower in her right hand, and other flowers are strewn on the floor. John’s head is nowhere to be seen, so we must presume that the moment selected by Moreau is when Salome chooses to receive that as her reward.
The rest of the painting consists of an unprecedented fusion of images, icons, and objects drawn from a diverse range of cultures. Detailed examination has shown these to be associated with the Hagia Sophia in Istanbul, the Alhambra in Granada, the Great Mosque of Cordoba, and several mediaeval cathedrals. Motifs have been identified from Etruscan, Roman, Egyptian, Indian, and Chinese art and culture.
But Moreau wasn’t content to show only that scene from the story. The other painting was to consider Salome with the head of John the Baptist as an apparition, and is now represented in three different versions.

The Apparition (1875) in the Musée National Gustave-Moreau is one of Moreau’s earliest attempts to express this. It takes the central part of Salome and adds the floating, severed head of John. Salome has now been transformed into the provocative, under-dressed femme fatale shown by subsequent artists. King Herod’s throne has been moved to the left of the painting, and he now looks in the direction of the apparition.

This watercolour painting of The Apparition (c 1876), now in the Musée d’Orsay, was that shown at the Salon, although its colours are far weaker than when it was first exhibited. The cadaveric King Herod sit on his throne, overseeing the scene from the left edge. Herodias, presumably, sits by his feet, and a musician for Salome’s dance is shown further back. At the right edge is the executioner, John’s blood still on his sword.
Salome is now nearly nude, her body decorated with an abundance of strategically-placed jewellery and adornments. She points at the apparition with her left hand, trying to stare it out, her face as blank as everyone else’s. She stands on her points, but there is no sign of movement. The floor isn’t just strewn with flowers, but is now stained with the dripping blood from the severed head.

Facial expressions are not theatrical as might have been expected in the work of a more conventional history painter of the day.

This slightly later oil version of The Apparition (1876-77), now in the Fogg Museum, gives a better idea of the original effect of Moreau’s watercolour, although the panther has moved across to replace the musician, and the background is quite different.
Moreau hadn’t painted Salome and The Apparition as a pair. Their compositions are individual, and mutually conflicting in details of the palace, the position of Herod’s throne, and more. Salome is one of the most iconographically rich paintings ever made, and it’s not surprising that some critics found it phantasmagoric. The Apparition is dominated by the same eye-to-eye contact that made Moreau’s Oedipus and the Sphinx so compelling, but here it’s between a notorious dancer and the severed head of the holiest man after Christ himself.
In 1877, the year after that Salon, Gustave Flaubert published three short stories, including an extended account of the traditional biblical narrative with Herodias at its centre. The British writer Oscar Wilde was introduced to that by Walter Pater (philosophical leader of Aestheticism), and in 1884 Joris-Karl Huysmans’ À rebours was published, a novel including a description of Moreau’s Salome paintings.
Wilde’s one-act play Salome was first published in French in 1891, and was soon translated into English and German. Banned from public performance in Britain, it received its premier in Paris in 1896, but wasn’t performed in public in England until 1931. At the centre of Wilde’s play is the perversion of lust and desire in Salome, best summarised in her words at the end of the play (he calls John the Baptist Jokanaan):
But, wherefore dost thou not look at me Jokanaan? Thine eyes that were so terrible, so full of rage and scorn, are shut now. Wherefore are they shut? Open thine eyes! Lift up thine eyelids, Jokanaan! Wherefore dost thou not look at me? Art thou afraid of me, Jokanaan, that thou wilt not look at me?
If thou hadst looked at me thou hadst loved me. Well I know that thou wouldst have loved me, and the mystery of love is greater that the mystery of death.
After seeing Wilde’s play performed in Berlin in 1902, Richard Strauss resolved to turn it into an opera. He started work on that in the summer of the following year, and Salome was completed and premiered in 1905. A year later, the dancer and choreographer Maud Allan produced a show called Vision of Salomé in Vienna, featuring a notorious version of the Dance of the Seven Veils, Wilde’s title for the dance of Salome before Herod, included in Strauss’s opera. The name quickly became a euphemism for a striptease, and the growing popularity of Salome as an erotic figurehead was named Salomania.
In around fifty years, from the appearance of Moreau’s The Apparition at the Salon in Paris, the traditional story of Herodias obtaining her vengeance by exploiting her daughter’s dance before Herod has been all but forgotten. The martyrdom of the second holiest figure in the gospels has been transformed into a perverse confusion of sex and death. The anonymous daughter of a woman who married her divorced husband’s brother has become the ultimate femme fatale: beautiful, sexy, and dangerous to know. Most unusually this change in story was largely triggered and driven by a painting: Moreau’s The Apparition.
Moreau was then concerned with the preparation of other paintings for the Exposition Universelle of 1878 in Paris.

Infancy and dawn are themes in Moïse Exposé sur le Nil (The Infant Moses) (c 1876-78), a radiantly beautiful depiction of the infant Moses asleep, prior to his discovery in the bullrushes. Moses is new life, new Judaeo-Christian beliefs, new law, and the new regime. Set against a background derived from photographs of Egyptian ruins symbolising the ancient, pre-Jewish, and decaying, it laid out Moreau’s hope for the French nation.
The baby Moses is marked out as being holy by the rays emanating from his temples, and surrounded by exotic flowers and birds. Most unusually, Moreau doesn’t show the traditional and popular moment of discovery of the infant in the bullrushes, but a static scene beforehand.

Moreau revisited his new myth of Salome and John the Baptist, in his strange watercolour of Salome in the Garden (1878). A beautiful and decorated figure of Salome is walking in an overgrown garden, carrying the severed head of John the Baptist on a large platter. Her eyes are closed, or perhaps looking down at the head, and John’s eyes are closed. Beside her is a headless statue of a man crawling, which could perhaps be the body of John, and outside is a man, possibly the executioner waving his sword.
References
Cooke P (2014) Gustave Moreau, History Painting, Spirituality and Symbolism, Yale UP. ISBN 978 0 300 20433 9.
Mathieu P-L (1998, 2010) Gustave Moreau, the Assembler of Dreams, PocheCouleur. ISBN 978 2 867 70194 8.
This weekend we’re off to visit Lake Geneva, also known by its French name of Lac Léman, the largest in Switzerland. It’s located in the far south-west of the country, where it forms much of its border with France. It makes a broad arc running north-east from the capital city of Geneva, with some of the highest peaks of the Alps to its south.

Today I start with a selection of paintings almost exclusively from the nineteenth century, when Switzerland was on the itinerary of the Grand Tour undertaken by aspiring young men of the upper class in both Europe and the Americas.

The city of Geneva has long attracted artists, and it was here the eccentric pastellist Jean-Étienne Liotard was born and later kept his studio, and where he eventually retired. His View of the Mont Blanc Massif from the Artist’s Studio from 1765-70 reveals only a little of the southern extreme of the lake, with a cameo self-portrait.

JMW Turner was by no means the first to paint the lake, but his watercolour of Lake Geneva and Mount Blanc from 1802-05 is one of its earliest depictions by a major artist. This view looks south-east over the city of Geneva towards the Mont Blanc massif in the far distance.

Alexandre Calame’s View of Bouveret from 1833 shows a grey heron fishing on the shore at the southern end of the lake, close to the border with France.

While Turner had toured the Alps once travel from England had become possible again in the early nineteenth century, Calame pioneered the painting of views like this of the lake, completed in his studio in 1849. It includes some of the distinctive sailing boats of the Swiss lakes, and a small bird in the shallows, but not a heron here.

In the middle of the nineteenth century, several major American artists visited Switzerland to develop their skills painting mountain views. Despite its finish, John Ferguson Weir’s Lake Leman (Lake Geneva), Switzerland may have been painted in front of the motif, on 11 June 1869.
Following Gustave Courbet’s release from prison for his involvement in the Paris Commune and destruction of the Vendôme Column in 1871, he was forced to flee to the safety of Switzerland, where he lived his remaining years there, unable to return to France.

Courbet painted some of the finest landscapes of his career during his exile in Switzerland, like this Sunset over Lac Léman from 1874, the year of the First Impressionist Exhibition in Paris.

He became particularly obsessed with the island château at the extreme eastern end of Lake Geneva, Chillon Castle, here in 1875. This picturesque château dates back to a Roman outpost, and for much of its recorded history from about 1050 has controlled the road from Burgundy to the Great Saint Bernard Pass, a point of strategic significance. It has since been extensively restored, and is now one of the most visited mediaeval castles in Europe.

Chillon Castle from 1874-77 is another of the views he painted of the castle on the lake.

Sunset on Lake Geneva from about 1876 is reminiscent of Courbet’s earlier seascapes with breaking waves, but now the water is calm once more.
In May 1877, the French government informed Courbet that the cost of rebuilding the Vendôme Column would be over 300,000 Francs, which he could pay in instalments of 10,000 Francs each year, starting on 1 January 1878. Courbet died in Switzerland the day before, on 31 December 1877, at the age of only 58.
这篇文章将这 31 个技巧汇编成一份详尽的指南,按从“入门基础”到“高级模式”的逻辑重新组织,并补充了 280 个字符无法容纳的深度背景信息。
无论你是刚刚起步,还是希望利用 Claude Code 提升段位,这里都有适合你的内容。
在深入研究具体功能之前,首先要配置 Claude Code,让它真正理解你的项目。
每个新成员都需要入职文档。使用 /init,Claude 会为自己写一份。
Claude 会读取你的代码库并生成一个 CLAUDE.md 文件,包含:
这是我在任何新项目中运行的第一条命令。
对于大型项目,你还可以创建一个 .claude/rules/ 目录,用于存放模块化、特定主题的指令。该目录下的每个 .md 文件都会作为“项目记忆”与 CLAUDE.md 一起自动加载。你甚至可以使用 YAML frontmatter 基于文件路径有条件地应用规则:
可以把 CLAUDE.md 想象成你的项目总指南,而 .claude/rules/ 则是针对测试、安全性、API 设计等特定领域的专项补充。
想把某些东西存入 Claude 的记忆,又不想手动编辑 CLAUDE.md?
在过去,你需要用 # 开头来让 Claude 将内容追加到文件中。但从 Claude Code 2.0.70 版本开始,流程变得更简单了——你只需要直接告诉它去更新。
直接告诉 Claude 记住它:
“Update Claude.md: always use bun instead of npm in this project”
(更新 Claude.md:在这个项目中始终使用 bun 而不是 npm)
无需打断你的心流,继续编码即可。
@ 提及是将上下文传递给 Claude 的最快方式:
在 Git 仓库中,文件建议的速度提高了约 3 倍,并且支持模糊匹配。@ 是从“我需要上下文”到“Claude 已获取上下文”的最短路径。
这些是你会频繁使用的命令。请将它们刻入肌肉记忆。
不要浪费 token 去问“你能运行 git status 吗?”
只需输入 ! 加上你的 bash 命令:
! 前缀会立即执行 bash 命令并将输出注入到上下文中。没有模型处理延迟,不浪费 token,无需切换多个终端窗口。
这一看似微小的功能,当你每天使用五十次后,就会意识到它的巨大价值。
想尝试一种“如果我们这样做……”的方法,但又不想承担后果?
尽管去试。如果情况变得奇怪,按两次 Esc 键即可跳回到干净的检查点。
你可以回退对话、代码更改,或者两者都回退。需要注意的是:已运行的 Bash 命令无法撤销。
你过去的提示词(Prompts)都是可搜索的:
不要重打,要去回忆。 这对斜杠命令(slash commands)同样适用,体验无缝衔接。
这就好比 git stash,但是用于你的提示词。
Ctrl+S 保存你的草稿。先发送其他内容。当你准备好时,你的草稿会自动恢复。
再也不用复制到记事本,再也不用担心在对话中途打断思路。
Claude 可以预测你接下来要问什么。
完成一项任务后,有时你会看到一个灰色的后续建议出现:
Tab 键曾经用于自动补全代码。现在,它自动补全你的工作流。可以通过 /config 切换此功能。
Claude Code 是一个持久化的开发环境,根据你的工作流对其进行优化,将极大地提升效率。
不小心关掉了终端?电脑在任务中途没电了?没问题。
上下文得以保留,势头得以恢复。你的工作永远不会丢失。你还可以通过 cleanupPeriodDays 设置会话保留的时间。默认是 30 天,但你可以将其设置得更长,或者如果你不想保留会话,可以设为 0。
你的 Git 分支有名字,你的 Claude 会话也应该有。
/resume 界面会对分叉(forked)的会话进行分组,并支持快捷键:P 预览,R 重命名。
在网页上开始任务,在终端里完成它:
这会将云端会话拉取并恢复到本地。无论在家还是在路上,Claude 都在。这也适用于 iOS 和 Android 的 Claude 移动应用,以及 Claude 桌面应用。
有时你需要一份关于发生了什么的记录。
/export 将你的整个对话转储为 Markdown 格式:
非常适合用于文档编写、培训,或者向过去的自己证明:是的,你确实已经尝试过那种方法了。
这些功能旨在消除摩擦,帮助你更快地行动。
厌倦了伸手去拿鼠标来编辑提示词?
输入 /vim,解锁全功能的 Vim 风格编辑体验:
以思维的速度编辑提示词。你几十年的 Vim 肌肉记忆终于在 AI 工具中得到了回报。退出 Vim 模式也前所未有地简单,只需再次输入 /vim。
Claude Code 在终端底部有一个可自定义的状态栏。
/statusline 让你配置显示的内容:
一目了然的信息意味着更少的手动检查和中断。
想知道是什么吃掉了你的上下文窗口?
输入 /context 查看究竟是什么在消耗你的 token:
当你的上下文开始变满时,这就是你找出问题所在的方法。
输入 /stats 查看你的使用模式、最爱用的模型、连续使用天数 (Streaks) 等。
橙色是新的绿色 (Orange is the new green)。
“我快达到限额了吗?”
了解你的极限,然后超越它们。
控制 Claude 如何处理问题。
通过一个关键词按需触发扩展思考:
当你在提示词中包含 ultrathink 时,Claude 会在回答之前分配最多 32k token 用于内部推理。对于复杂的架构决策或棘手的调试会话,这往往决定了你得到的是肤浅的答案还是真正的洞察。
注:以前你可以指定 think, think harder, ultrathink 来分配不同数量的 token,但现在我们已将其简化为单一的思考预算。当配置了 MAX_THINKING_TOKENS 时,ultrathink 关键字将失效,配置项将优先控制所有请求的思考预算。
先驱散战争迷雾。
按两次 Shift+Tab 进入计划模式 (Plan Mode)。Claude 可以:
但在你批准计划之前,它不会编辑任何内容。三思而后行 (Think twice. Execute once.)。
我有 90% 的时间都默认处于计划模式。最新版本允许你在拒绝计划时提供反馈,使迭代更快。
直接使用 Claude API 时,你可以启用扩展思考来查看 Claude 的逐步推理:
Claude 在回答之前会在思考块 (thinking blocks) 中展示其推理过程。这对调试复杂逻辑或理解 Claude 的决策非常有用。
没有控制的力量只是混乱。这些功能让你设定边界。
/sandbox 让你一次性定义边界。Claude 在边界内自由工作。
你获得了速度,同时拥有真正的安全性。最新版本支持通配符语法,如 mcp__server__*,用于允许整个 MCP 服务器。
厌倦了 Claude Code 做什么都要请求许可?
这个标志对一切说 Yes。它的名字里带有“dangerously”(危险地)是有原因的——请明智地使用它,最好是在隔离环境或受信任的操作中。
Hooks 是在预定生命周期事件发生的 shell 命令:
通过 /hooks 或 .claude/settings.json 进行配置。
使用 Hooks 来阻止危险命令、发送通知、记录操作或与外部系统集成。这是对概率性 AI 的确定性控制。
Claude Code 的作用不止于交互式会话。
你可以将 Claude Code 用作脚本和自动化的强大 CLI 工具:
流水线中的 AI。-p 标志以非交互方式运行 Claude 并直接输出到标准输出 (stdout)。
将任何提示词保存为可复用的命令:
创建一个 Markdown 文件,它就变成了一个斜杠命令,并且可以接受参数:
不要重复自己。你最好的提示词值得被复用。
Claude Code 可以看到并与你的浏览器交互。
Claude 现在可以直接与 Chrome 交互:
“修复 Bug 并验证它能工作”现在只需一个提示词。从 claude.ai/chrome 安装 Chrome 扩展程序。
这是 Claude Code 真正强大的地方。
圣诞老人不会自己包装每一份礼物。他有精灵。
子代理 (Subagents) 就是 Claude 的精灵。每一个子代理:
像圣诞老人一样放权。子代理可以在后台运行,而你继续工作,它们拥有访问 MCP 工具的完全权限。
技能 (Skills) 是指导 Claude 完成特定任务的指令、脚本和资源的文件夹。
它们一次打包,随处可用。而且由于 Agent Skills 现在是一个开放标准,它们可以在任何支持该标准的工具中工作。
把技能看作是按需赋予 Claude 专业知识。无论是你公司特定的部署流程、测试方法论,还是文档标准。
还记得以前分享 Claude Code 设置意味着要跨 12 个目录发送 47 个文件吗?
那个时代结束了。
插件将命令、代理、技能、Hooks 和 MCP 服务器打包在一起。通过市场发现新的工作流,市场包含搜索过滤功能,便于发现。
LSP 支持赋予了 Claude IDE 级别的代码智能:
LSP 集成提供:
Claude Code 现在像你的 IDE 一样理解你的代码。
驱动 Claude Code 的代理循环、工具和上下文管理现在作为 SDK 提供。只需不到 10 行代码即可构建像 Claude Code 一样工作的代理:
这仅仅是个开始。
当我开始这个“倒数日历”时,我以为我只是在分享技巧。但回顾这 31 天,我看到了更多的东西:一种人机协作的哲学。
Claude Code 中最好的功能都是为了给你控制权。计划模式、代理技能、Hooks、沙盒边界、会话管理。这些是与 AI 协作的工具,而不是向它投降。
能从 Claude Code 中获得最大收益的开发者,不是那些输入“帮我做所有事”的人。而是那些学会了何时使用计划模式、如何构建提示词、何时调用深度思考 (Ultrathink),以及如何设置 Hooks 在错误发生前捕获它们的人。
AI 是一个杠杆。这些功能帮助你找到正确的抓手。
致 2026 年。