Reading view

There are new articles available, click to refresh the page.

广州越秀发生伤人事件 两人受伤嫌犯被抓

中国广州市发生伤人事件,警方通报称两人受伤,已抓获嫌犯。

广州市公安局越秀分局星期五(12月19日)深夜在官方微博发布警情通报称,广州110当天下午5时37分接群众报警,称越秀南路有两人受伤。

通报称,警方立即到场处置,迅速抓获34岁的犯罪嫌疑人赵姓男子,两名伤者及时送医救治,均无生命危险。

警方没有透露嫌犯作案动机与受害者身份等细节,称案件正在进一步侦办中。

同一天,广西来宾市忻城县也发生持刀伤人案,导致三人死亡,一人受伤,35岁郭姓犯罪嫌疑人被控制。

网传高雄车站为下个袭击目标 台行政院长:全力追查

针对有网民预告高雄车站将成为下一个袭击目标,台湾行政院长卓荣泰受访时表明,将通过警方系统全力追查此人,“让他成为社会公惩对象”。

综合ETtoday新闻云和TVBS新闻网等台媒报道,卓荣泰星期六(12月20日)到台北内湖三军总医院探视捷运袭击事件伤者。

在被问及台北捷运发生袭击事件后,有网民在社群网络Threads上预告高雄车站将成为下一个袭击目标,卓荣泰回应时说,这次袭击事件给社会造成很大震撼,政府要重新检讨很多工作。此时,他拜托社会要减少对这种不实在、不正常的信息的互相传递。

他强调,政府从星期五已透过警方系统,开始针对网络出现的恶意模仿言论开展全力追查,“一定要找出IP(网络地址)、这个行为人,让他变成社会公惩的对象”。

卓荣泰直言,他非常痛恨这样的行为,这对社会完全负面、败坏,无论真假,都不应该这么做。希望大家以此为戒,不要跟风模仿这样的行为。

高雄市长陈其迈星期六受访时说,已掌握上述贴文账号的网络地址位置,共计五个境外可疑账号,与警政署、国安单位密集联系,在全力追查中。

台北捷运星期五傍晚下班尖峰时段发生袭击事件,迄今为止,已造成包括27岁嫌犯张文在内的四人身亡、11人受伤。

港府申请解散宏福苑业主立案法团管委会

香港大埔宏福苑11月26日发生致命火灾,截至当晚8时半,这场五级大火已导致13死15伤,包括一名37岁的消防员殉职。 (中新社)

香港特区政府星期五申请解散宏福苑业主立案法团管理委员会。

据香港政府公报,港府星期五(12月19日)引用了《建筑物管理条例》(第344章)第31条,向土地审裁处提交申请,解散现有宏福苑业主立案法团的管理委员会,并建议委任华懋集团旗下的合安管理有限公司为委任管理人,以协助宏福苑业主处理火灾后的各项管理及后续事宜。

港府称,如土地审裁处批准有关申请,民政事务总署将成立专班,联同合安管理有限公司,密切跟进屋苑管理的各项工作,包括工程合约处理、跟进保险索赔、财务管理及居民支援等环节。

港府说,合安管理有限公司已明确表示,在这次委任中将不会收取任何行政费用。与此同时,龚如心慈善管理有限公司也表示如土地审裁处批出有关申请,将捐出500万港元(83万新元)予法团支付因须聘用律师、会计师或其他专业服务的费用。

香港民政及青年事务局局长麦美娟当天就宏福苑管理事宜会见媒体时说,这是政府主动启动上述法律程序的首例。

她指出,当前宏福苑的管理正面临前所未有的挑战,尤其在大型维修工程后续跟进、相关工程合约及法团维修账目的处理、保险赔偿、现有管理公司约满后屋苑的管理及保安等复杂问题上,涉及多重技术、法律与财政考量,非一般法团管委会有能力处理。

她说,政府深切体会受灾居民的巨大困难和不安,明白他们急需获得适切支援,也关注法团本届管委会是否有能力就各复杂事宜保障居民权益。政府近日也收到不少居民意见,表示希望政府能协助接管法团的管理。

麦美娟称,政府经审慎评估后,认为有必要回应居民诉求,因此决定主动介入,依法向土地审裁处提出申请,委任具丰富经验的华懋集团旗下的合安管理有限公司作为临时管理人,透过引入专业管理资源,做好屋苑在灾后的管理跟进工作,保障所有业主的权益。临时管理人的任命能减轻业主在这段期间的财政负担。

她强调,管理人只是取代了法团管委会,上述申请不会影响宏福苑业主对重要屋苑管理事务的最终决定权。管理人在处理大型维修工程的后续跟进、相关工程合约、法团账目管理等复杂问题时,会先寻求专业意见,进行详细分析,并向业主做好解说,所有最终决定仍依照《建筑物管理条例》的规定交由业主作投票通过。

香港大埔宏福苑11月26日发生致命火灾,截至目前已造成161人死亡。

中国公安侦破由美通报涉洗钱案

中国公安侦破了一起美国通报的涉嫌洗钱案,两人以非法经营罪被判刑。

据中国《人民日报》报道,2024年4月一条美国通报的佟姓中国籍人员涉嫌为毒贩洗钱线索,引起中国警方高度重视。公安部立即部署辽宁省公安厅开展核实工作。经前期侦查,辽宁省公安厅以涉嫌非法经营(外汇)犯罪立案,并将案件移交给佟姓被告等人户籍所在地沈阳市公安机关继续侦办。

报道称,围绕佟姓被告等人犯罪线索,沈阳警方成立工作专班,克服线索碎片化、境外调查取证不便、中美法律体系差异等困难,开展多维度缜密侦查,迅速固定了重要证据。沈阳警方2024年5月20日在湖北武汉将企图潜逃出境的佟姓和陈姓两名被告当场抓获。

佟姓被告等人被指在美国开设车行销售汽车并为购车客户提供人民币和美元兑换服务,后期转为专门从事非法买卖外汇等犯罪活动。

佟姓被告被指2020年1月至2021年6月期间在美国主要采取“对敲”方式,指使居住在中国境内的陈姓被告等人,利用在国内办理的多张银行卡,接收换汇转账。资金到账后,佟姓被告在美国向换汇人或佟姓被告亲属支付美元现金,收取1%—2%手续费。佟姓被告通过境内银行卡实施非法买卖外汇金额合计1600余万元人民币(294万新元)。

佟姓被告承认,在美国经营车行期间,结识了墨西哥商人JC。JC获知佟姓被告因给他人换汇需要美元现金,提出以美元现金向其购买虚拟货币。于是,佟姓被告指使陈姓被告通过多种渠道购买比特币、泰达币等虚拟货币后,卖给JC换取美元现金。

据佟姓被告供述,为JC换汇只占其业务一部分。佟姓被告在美期间,主要为有大额换汇需求的留学生、商人,以及美国当地的一些赌客提供换汇服务。

警方介绍,据佟姓被告交代,佟姓被告虽在美国从事换汇活动,但并不知晓JC等人提供的资金具体来源,各方掌握的现有证据不能确认佟姓被告明确知道上游资金性质。侦查工作中,警方同时发现一些不法分子在大额现金存取、虚拟货币交易等方面规避美国监管,目前已通过中美反洗钱合作沟通渠道反馈美国,并就在反洗钱领域加强执法合作达成共识。

沈阳市皇姑区人民法院2025年9月经审理认为,佟姓和陈姓两名被告违反国家有关法律规定,非法买卖外汇,扰乱金融市场秩序,情节严重,以非法经营罪分别判处佟姓、陈姓被告有期徒刑一年七个月、一年四个月,并处罚金。

新华社早前报道,中美两国禁毒部门开展执法合作,侦破一起走私可卡因案。

赖清德要求追查台北袭击案有无共犯

台湾总统赖清德到医院探视台北捷运袭击事件伤者后对媒体说,他要求检警调单位彻查此案,并追查是否有共犯或背后有人指使。

综合台湾《自由时报》和中时新闻网报道,赖清德星期六(12月20日)上午到台大等医院探视台北捷运袭击事件伤者,之后对记者说,他要对在星期五(12月19日)晚暴力攻击事件中不幸往生的民众表达哀悼,并慰问遇难者家属。他同时感谢台大医院、三军总医院等尽全力抢救伤患,也对在事件过程中奋勇阻止嫌犯进一步攻击的民众致谢。

赖清德强调,他已经要求检警调相关单位,务必针对嫌犯个人背景、犯案动机、是否有共犯、背后是否有人指使进行全面深入彻底的调查,并向社会交代真相。

根据台湾卫生福利部公布的最新伤病患收治情况,27岁嫌犯张文星期五在捷运台北车站及中山站犯下连续无差别袭击事件。截至星期六上午6时25分,事件共造成15人送医,包括嫌犯在内已有四人不治,其余11人受伤,其中仍有六人留院治疗,包含两人在加护病房观察中。  

AI 大神 Karpathy 2025 年度总结刷屏:AI 既是天才也是智障,这 6 个转折最关键

最近各种年度回顾陆续上线, OpenAI 的前联合创始人 Andrej Karpathy 也交出了自己对大模型的年度总结

就在今年早些时候,他在 YC 的一场演讲刷爆了全网,提出了不少新的观点:

  • 软件 3.0 已来: 从最初的人写代码(1.0),到喂数据训练模型(2.0),现在我们进入了直接对模型「念咒语」(Prompt)的 3.0 时代。
  • LLM 是新操作系统: 它不是像自来水一样的普通商品,而是一个负责调度内存(Context Window)和 CPU(推理算力)的复杂 OS。
  • Agent 的十年: 别指望 AI Agent 一年就成熟,从 99% 到 99.999% 的可靠性,我们需要走上十年。

Karpathy 今天这篇《2025 年度总结》,他再次向我我们剖析了这一年 AI 究竟长出了什么样的「脑子」。

以下是对 Karpathy 年度总结的精译,APPSO 在不改变原意的基础上做了更多通俗解读。

如果想看原文可以点击🔗 https://karpathy.bearblog.dev/year-in-review-2025/

2025 年是 LLM(大语言模型)发展强劲且充满变数的一年。以下是我列出的几点个人认为值得注意且稍感意外的「范式转变」——这些变化不仅重塑了行业版图,更在概念层面上狠狠地冲击了我的认知。

🛑 太长不看版:

2025 年既让人兴奋,又有点让人措手不及。

LLM 正在作为一种新型智能涌现,它同时比我预期的要聪明得多,也比我预期的要笨得多。

无论如何,它们极其有用。我认为即使以目前的能力,行业甚至还没挖掘出其潜力的 10%。同时,还有太多的想法可以尝试,从概念上讲,这个领域感觉依然广阔。正如我今年早些时候提到的,我同时(表面上矛盾地)相信:我们将看到持续快速的进步,但前方仍有大量艰苦的工作要做。

系好安全带,我们要发车了。

1. RLVR:教 AI 像做奥数题一样「思考」

在解释这个复杂的基础概念之前,先看看以前是大模型训练是怎么做的?

在 2025 年初,各大实验室训练 LLM 的「老三样」配方非常稳定:

1. 预训练(Pretraining):像 GPT-3 那样,让 AI 读遍全网文章,学会说话。

2. 监督微调(SFT):找人写好标准答案,教 AI 怎么回答问题。

3. 人类反馈强化学习(RLHF):让 AI 生成几个答案,人来打分,教它讨人喜欢。

现在发生了什么变化?

2025 年,我们在这个配方里加了一味猛药:RLVR(从可验证奖励中进行强化学习)。

这是什么意思?

简单来说,就是不再让人来打分(人太慢且主观),而是让 AI 去做那些「有标准答案」的任务,比如数学题或写代码。对就是对,错就是错,机器能自动验证。

在数百万次的自我博弈和试错中,模型自发地演化出了看似「推理」的策略。它们学会了先把大问题拆解成小步骤,甚至学会了「回过头来检查」这种高级技巧(参考 DeepSeek R1 论文)。

核心对比:

  • 旧范式(RLHF): 像是教小孩写作文。因为没有标准答案,AI 很难知道自己哪一步想错了,只能模仿人类的语气。
  • 新范式(RLVR): 像是把 AI 关进奥数训练营。不用教它具体怎么想,只要给它足够多的题和对错反馈,它自己就能摸索出解题套路。

这一招太好用了,以至于 2025 年大部分算力都被这只「吞金兽」吃掉了。结果就是:模型并没有变大,但训练时间变长了。 我们还获得了一个新旋钮:让 AI 思考得久一点。OpenAI 的 o1 是开端,而 o3 则是真正的拐点。

2. 幽灵 vs 动物:AI 不是「电子宠物」

2025 年,我和整个行业终于从直觉上理解了 LLM 智能的「形状」。

一个惊悚的比喻:我们不是在像养宠物一样「进化/养育动物」,我们是在「召唤幽灵」

为什么这么说?

因为 AI 的一切都和生物不同。人类的大脑是为了在丛林里活下来、为了繁衍后代而优化的;而 LLM 的大脑是为了模仿人类文字、在数学题里拿分、在竞技场里骗赞而优化的。

参差不齐的智能(Jagged Intelligence):

正是因为 RLVR(可验证奖励)的存在,AI 的能力在某些领域(如数学、编程)会突然飙升成刺状。这就导致了一种极其滑稽的现象:

  • 它同时是一个绝世天才(秒解高数题);
  • 又是一个智障小学生(会被简单的逻辑陷阱骗得团团转)。

▲这里 Karpathy 引用了一张梗图:人类智能是圆润的蓝色圆圈,AI 智能是像海胆一样满是尖刺的红色图形。这很形象.

这也解释了为什么我对现在的「跑分榜单」(Benchmarks)失去了信任。

什么是「刷榜」的本质?

既然榜单是可验证的,那就可以用 RLVR 针对性训练。现在的实验室都在搞「应试教育」,把 AI 的能力尖刺专门往考题上长。「在测试集上训练」已经不仅仅是作弊,而成了一门新的艺术形式。

3. Cursor:不仅是编辑器,更是「包工头」

Cursor 今年的爆火,揭示了一个新真相:LLM 应用层比我们想象的要厚。

大家开始谈论「医疗界的 Cursor」、「法律界的 Cursor」。这些应用到底做了什么?

  • 「上下文工程师」: 帮你整理好所有背景资料喂给 AI。
  • 「工头」: 在后台偷偷指挥多个 LLM 干活,把复杂任务拆解,还要帮你省钱。
  • 「遥控器」: 给你一个调节「自主性」的滑块,决定放手让 AI 干多少。

预测:大模型实验室(如 OpenAI)会负责培养「全科大学生」;而应用开发商(如 Cursor)则负责给这些学生提供私有数据和工具,把他们组建成「专业施工队」

4. Claude Code:住在你电脑里的「赛博幽灵」

Claude Code (CC) 的出现让我眼前一亮。它不仅仅是一个能写代码的 Agent(智能体),更重要的是:它活在你的电脑里

对比来看,我认为OpenAI 搞错了方向。

OpenAI 早期的 Agent 都在云端跑(ChatGPT),离你的真实环境太远。虽然云端智能体听起来像是 AGI 的终局,但在当前这个「参差不齐」的过渡阶段,本地才是王道。

为什么本地很重要?

因为你的代码、你的配置、你的密钥、你的混乱环境,都在本地。Anthropic(Claude 的母公司)搞对了优先级,他们把 AI 塞进了一个小小的命令行界面(CLI)里。

它不再是你浏览器里的一个网页(像 Google 那样),它变成了一个寄宿在你电脑里的「赛博幽灵」,随时准备帮你干活。这才是未来 AI 交互的样子。

5. Vibe Coding

什么是 Vibe Coding?

这是我在推特上随口造的一个词(居然火了):意思是写代码不再需要你真的懂语法,你只需要用英语描述你的「意图」和「感觉」,剩下的交给 AI。

这带来了什么改变?

  • 对于普通人: 编程的门槛彻底消失了。
  • 对于专家: 代码变得像纸巾一样「廉价、一次性、用完即弃」。

举个例子,我为了找一个 Bug,可能会让 AI 现场写一个专门的 App 来测试,测完就删。放在以前,为了找个 Bug 专门写个 App?疯了吧!但在 2025 年,代码是免费的。

Vibe Coding 将会彻底重塑软件行业,也会改写程序员的招聘 JD。

6. Nano Banana:AI 终于有了自己的「脸」

为什么现在的 AI 交互很反人类?

不管是 ChatGPT 还是 Claude,我们还在用「打字」跟它们聊天。这就像 80 年代还在用 DOS 命令行的黑底白字。

事实是: 计算机喜欢文本,但人类讨厌读文本。人类是视觉动物,我们喜欢看图、看表、看视频。

Google Gemini Nano banana(这是一个虚构的模型代号,指代某种多模态交互模型)是 2025 年的另一个范式转变。它暗示了未来的 LLM GUI(图形界面) 是什么样子的

未来的 AI 不应该给你吐一堆字,它应该直接给你画一张图、生成一个网页、弹出一个交互面板。 这不仅仅是「画图」,而是将文本生成、逻辑推理和视觉表达纠缠在一起的混合能力。

连接热爱,发现创新价值的科技媒体,传播有价值的发声文本。

#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。

爱范儿 | 原文链接 · 查看评论 · 新浪微博


XREAL 1S 体验:巨幕随身带,不到三千块

这两天大家都在为《阿凡达 3》摩拳擦掌,而如何挑选影厅,到底该看 CINITY、IMAX 还是杜比,是个巨大的纠结——没关系,我们有一篇文章专门帮大家答疑解惑。

但如果你想要一块随身巨幕的话,我想在 2025 年底,没有比 XREAL 1S 更合适的(多么丝滑的转场!)

消费级增强现实 (AR) 眼镜的进步,基本上被 XREAL 这一家公司撑着。作为唯一同时具备自研空间计算芯片和自研光学系统能力的 AR 技术公司,XREAL 在「眼镜作为随身巨幕」的这个细分市场上,可以说让竞争对手望尘莫及。

这个优势,被 XREAL 最新发布的 1S 眼镜再次保持住了。

虽然实话来说,各方面和之前该公司发布的同为 One 系列的 AR 眼镜大差不差(都以随身巨幕为主要功能点),但这次 1S 的销售价格更加亲民,明摆着冲着年轻消费人群而来。

更重要的是,XREAL 独家自研的 X1 空间计算芯片,使得无设备歧视的纯原生云台增稳、极低 M2P 延迟的画面悬停等 3DoF 能力成为可能;以及 52° 的 FoV、自动电致变色——诸多过去只在旗舰机上才有的体验和功能,这台价格更亲民的 1S 更是一个没落下。

诚恳地讲,如果你只是需要一个随身巨幕,无论是飞机高铁看电影,还是日常当个隐私感更强的扩展屏,而且对体验颇为吹毛求疵的话,那么国补后价格打下来到了 2799 的新款 XREAL 1S,比价格更高的竞品,以及同品牌的前序产品,还要更物超所值。

爱范儿最近使用了 XREAL 1S 一个星期左右,主要场景是办公屏幕、看电影,以及外接手机在长途通勤时刷剧/打游戏,来说说这台随身巨幕眼镜的主要卖点和对应的体验。

很多 AR 眼镜都宣称自己有巨幕,但除了「巨」之外,一切其它的体验细节,诸如画面稳定性、动态延迟等等,都好像没有被顾及到。

XREAL 1S 则并非如此,它的屏幕不仅大,而且观感更「真实」:戴上眼镜,选定屏幕位置之后,屏幕能够一直稳定悬浮在选定的位置,无论怎样晃动、旋转头部,甚至摘下眼镜再戴上,那块屏幕都会一直在那里。

我觉得悬停效果的优劣,对于巨幕眼镜的体验好坏起到决定性作用。稳定的悬浮,让画面感到真实,能够显著改善 AR 眼镜的眩晕问题。AR 眼镜的可视角度逐渐增广,显示面积越来越大,但既然我们把它当做一块巨幕,那么画面的稳定性才更重要。

这就好比你沉浸在一部电影那引人入胜的剧情当中,旁边有人喊你,你转了个头,再转回来的时候,画面的位置突然变了,沉浸感没了,那不会是好的观影体验。悬停效果的好坏,其实和人类视觉的「直觉性」关系很大。而越符合直觉,体验就会越好。

现在我们建立了「画面空间悬浮固定」很重要这个前提。在此基础上,更进一步,我认为画面能够跟随用户的需要,及时移动位置,且过程中保持稳定和顺滑,同样很重要。

如果需要移动画面位置,XREAL 做了一个很简单的设计:长按镜腿下方的 X 按钮,也就是眼镜上唯一的红色按钮/主要的交互按钮,即可重新定位画面。你不需要进入多级菜单,绕来绕去,即可完成操作。

这个功能的背后有多重场景考虑:比方说你脖子酸了,后仰靠在椅背上,可以把画面重新定位到仰角更高的位置,来放松一下脖子;再比如,当你临时需要在多个屏幕上进行多任务操作,可以将眼镜当做副屏,把画面重新定位到桌面主显示屏的旁边(当你转头看回主屏幕的时候,镜片还会自动变色,让你能够清晰看到主屏幕上的内容)——无论场景是什么,这个功能都很方便。

而与悬停模式平行的,是另一个模式:云台跟随。同样无需任何外挂传感、适配器、第三方软件,在跟随模式下,1S 的画面能够实现类似「云台」一样的稳定效果,无论是下意识的头部/脖子轻微晃动,还是大幅度的活动,画面都会在一定范围内保持稳定,只在动作幅度足够大的时候,才会顺滑地跟随。

上述的画面悬浮锁定、云台防抖等功能的背后,是 XREAL 自研 X1 空间计算芯片的功劳。这颗芯片不仅能够实现优秀的原生 3DoF 能力,另一大作用是显著降低了 M2P(motion to photon)延迟,也即从「用户头部运动开始」到「渲染图像发生相应变化」中间的延迟。

XREAL 的官方数据显示,没有 X1 芯片时,数据处理的链路较长,甚至需要视频输入源设备(电脑、手机等)的 CPU 和 GPU 运算,M2P 延迟可能超过 20 毫秒,延迟是肉眼可察觉,是反直觉的,会加重眩晕效果。

而基于 X1 芯片的数据处理链路,眼镜本地获取运动数据,本地芯片完成计算并矫正视频图像,直接反应到屏幕上,M2P 延迟最低可以达到 2 毫秒,已经进入肉眼不可察觉的区间。

X1 芯片还有其他重要的作用,比如从过去的逐帧画面补偿,进化为逐行补偿。在画面悬停模式下,向左右方向大幅度歪头的时候,你就会发现逐行补偿的强大之处,即便大角度倾斜,画面本身的显示,和画面的位置,仍然稳如泰山(只有在角度超过了计算极限的时候才会熄屏)。

至于在内容消费体验上最有意思的,还得是 3D 模式——就像当年第一次用上任天堂 3DS 一样奇妙。纯靠 X1 芯片自身的算力,将任何传统 2D 内容转化为 3D 显示效果,哪怕只是追个剧,看个下饭视频,都能获得不一样的体验。

说到下饭,戴着 1S 吃饭这件事我还真试了一下……它有三档可调的电致变色,所以在任何环境下都可以调整成我需要的通透效果。以及在悬停模式下视线移动到屏幕以外的位置,1S 会自动切换到无色的通透模式。

这样就可以做到:低头扒口饭,抬头接着看……

52° 视场角或许还不够宽,但基本达到了「不转动眼球即可清晰感知」的视野上限

52° 视场角或许还不够宽,但基本达到了「不转动眼球即可清晰感知」的视野上限

除此之外, XREAL One 系列标配的 Sound by Bose 定制音效,1S 也没有落下;700 尼特的入眼亮度,结合电致变色,在显示效果和护眼之间取得平衡。索尼 0.68 英寸 Micro-OLED 配合双眼 1200p 分辨率,并且支持常规屏幕、超宽屏模式,以及搭配电脑的多种分辨率/屏幕比例可调,让 1S 的使用方式更加多样。

比如工作时候就开 1.6:1 的全尺寸显示,画幅更高、显示内容更多。观看影视作品,换成超宽屏则沉浸感更强。

最重要的可能还是价格:1S 的官方定价 3299 元,国补/各类补贴后的参考价格在 2799 元浮动,成功打进 2500-3000 价位段。要知道这个区间里,在此之前是没有 1S 性能和体验水准的产品的。

这也是为什么我在标题里说:不到三千块,巨幕随身带。

在我看来,这句话就是 XREAL 1S 的精髓。1S 是一块各方面素质顶级,且产品定义设计充分考虑了 AR 眼镜关键痛点的「随身巨幕」。而如果你对这个细分市场有兴趣,经常长途通勤需要一块随身巨幕,那么在当前这个时间点上,似乎真的没有比 1S 更合适的替代品……

——直到 XREAL 这家进化速度极快的公司,再次用更强大的新产品「背刺」老用户之时。

#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。

爱范儿 | 原文链接 · 查看评论 · 新浪微博


I've worked just as hard as the other Strictly finalists, says Amber Davies

Wireimage/Getty Images Amber Davies attends the Glamour Women Of The Year Awards 2025 at 180 Studios on October 30, 2025 in London, EnglandWireimage/Getty Images

After weeks of intense competition, the Strictly Come Dancing finalists are primed to take to the dance floor one last time.

Just three couples remain of the 15 at the start of the series, with celebrities Amber Davies, George Clarke and Karen Carney vying to get their hands on the glitterball trophy.

It's also the last time Tess Daly and Claudia Winkleman will present a Strictly final. The pair announced they would be stepping down at the end of the current series.

The grand final will be broadcast on Saturday from 19:00 GMT on BBC One and iPlayer.

Amber Davies says she's earned her place in final

Amber Davies and Nikita Kuzmin in white outfits performing on Strictly

One of the main talking points leading into the final has been whether Amber Davies' performing background has given her an unfair advantage.

The Love Island star has faced a backlash online with people pointing out she's had roles in West End musicals Pretty Woman, 9 to 5: The Musical, and The Great Gatsby, as well as competing in the 16th series of Dancing on Ice.

This week, she told BBC News and other reporters at a Strictly press conference that she could "appreciate" where the comments are coming from.

"I have experience compared to these two," she said, gesturing at the other finalists. "I'm not shying away from that. However, I would say that the love actually does outshine the hate."

When asked whether she deserves to win, the actress, who is paired with Nikita Kuzmin, said: "I feel like I have worked just as hard as these guys going into the final, because I had to fight to get into the final."

The backlash is perhaps a symptom of a wider theme on Strictly, where over the years some viewers have appeared to form their own judgements on contestants, regardless of what the judges think or what the overall audience vote says.

Earlier this series Lewis Cope received criticism for having been a child actor in Billy Elliot, for example, and last year JB Gill came under fire for having danced before in the boyband JLS.

'I can't believe a footballer is in the final'

Karen Carney and Carlos Gu performing against a glittery background on Strictly

And it seems many viewers value an element of surprise or beating the odds, as with Strictly's first blind contestant Chris McCausland winning last year.

For Karen Carney, becoming the first footballer to reach the Strictly Come Dancing final feels "weird" but she's also "really, really proud".

"I can't believe a footballer is in the final," she said in the press conference.

The former Lioness, who is partnered with Carlos Gu, said she had struggled with her confidence in recent years, but added: "I always knew in my heart Strictly would be the thing that would fix me... It's by far the happiest I've ever been."

Crown for the social stars?

George Clarke and Alexis Warr in red outfits performing on Strictly

For social media star George Clarke, reaching this point in the contest is "mental".

"It's a completely different world to what I'm used to," he said.

Another online star, AngryGinge, was recently crowned king of the jungle in this year's I'm A Celebrity... Get Me Out of Here! It opens up the possibility of two social media stars winning two of the biggest shows on television.

"I feel like it's a whole demographic that hasn't really been touched massively in mainstream TV," said Clarke, who is partnered with Alexis Warr.

Who is dancing to what this week?

The finalists will perform three routines: they will revisit a previous routine, perform a show dance and then take to the floor with their favourite dance from the series.

This is what they'll be dancing to:

Amber and Nikita

  • Showdance to Rain On Me by Lady Gaga and Ariana Grande
  • Judges' Pick - Paso Doble to Dream On by Aerosmith
  • Favourite Dance - Jive to Proud Mary

George and Alexis

  • Showdance to Human by The Killers
  • Judges' Choice - Viennese Waltz to Somebody to Love by Queen
  • Favourite Dance - Paso Doble to Game of Survival by Ruelle

Karen and Carlos

  • Showdance to Inner Smile by Texas
  • Judges' Pick - Argentine Tango to Red Right Hand by Nick Cave & The Bad Seeds
  • Favourite Dance - Jive to One Way Or Another by Blondie

Other highlights from the show will include a group routine from the professional dancers and this year's finalists.

The cast of 2025 will return to the dance floor one last time for a group number, although BBC News understands that Apprentice contestant Thomas Skinner will not join.

What time does the Strictly final start?

Strictly Come Dancing 2025's grand final is at 19:00 GMT this Saturday on BBC One and iPlayer.

It will also be available to watch on the live page on the BBC News website.

The live page will also have plenty of build up ahead of the show - plus extra insight on the dances, the outfits, the judges' comments, and much more.

How can I vote for my favourites?

This year, Strictly has moved to online-only voting.

When the vote is open, shimmy over to this website to make your decision!

As with every final, all of us at home have the full power to decide the winner.

The judges will give their scores, but these will just be for guidance and don't get combined with the viewers' votes.

What happened last weekend?

Warning: The section below includes spoilers from last week's Strictly

The four remaining couples line up on the ballroom floor. Karen Carney is wearing a blue dress. George Clarke and his partner Alexis Warr wear red. Balvinder Sopal is in an orange dress and Amber Davies wears a dark ballgown.
The semi-finalists had their fate revealed last week

Standards were incredibly high last week, as you would expect for a semi-final.

Amber Davies and Nikita Kuzmin kept up with their perfect score for a third consecutive week, while Karen Carney's Waltz had judge Anton Du Beke "a bit emotional" and George Clarke captivated the judges with his hip movement.

In the end, Balvinder Sopal and Caillon left the competition, following a dance off on Sunday night against Davies and Kuzmin.

Tess and Claudia's last Strictly final

Claudia Winkleman and Tess Daly with glitterball trophies
During last Sunday's results show, Claudia Winkleman and Tess Daly were each awarded their own glitterball trophy.

In October, we had the shock announcement by Daly and Winkleman that they would be stepping down at the end of this year's series.

The celebrity contestants later told me that the news took them by surprise too. "No one saw it coming," one said.

Daly has presented since the first series 21 years ago. Winkleman joined in 2014, having hosted Strictly's sister show It Takes Two since 2004.

This Saturday will be their last time presenting a Strictly final. Speaking on Strictly's sister show It Takes Two last week, the pair admitted they might feel "emotional" after the show.

Their last ever appearance on the show will be the Christmas special, which airs on 25 December.

The Strictly glitterball trophy

There have been plenty of twists and turns in the show's 21st year.

But the show has also been embroiled in controversy.

In November, it emerged that an unnamed star from Strictly has been arrested on suspicion of rape. It comes after another man involved with the hit BBC dance show was also arrested on suspicion of rape in August.

It is understood that neither developments were related to the current series of Strictly.

It came after a difficult few years for the show, with a number of former celebrity contestants speaking up about their negative experiences on Strictly.

Last year, the corporation introduced new duty of care measures, including putting chaperones in rehearsals.

Are these AI prompts damaging your thinking skills?

yacobchuk/Getty Woman lying on a medical bed wearing a cap with multiple electrodes and wires attached to the head. In the background, a healthcare professional in a white coat is examining brain scan images displayed on a monitor.yacobchuk/Getty
MIT researchers used electroencephalography to record brain activity while people used generative AI

What was the last thing you asked an AI chatbot to do for you?

Maybe you asked it for an essay structure to help answer a tricky question, provide an insightful analysis of a chunky data set, or to check if your cover letter matches the job description.

Some experts worry that outsourcing these kinds of tasks means your brain is working less - and could even be harming your critical thinking and problem-solving skills.

Earlier this year, the Massachusetts Institute of Technology (MIT) published a study showing that people who used ChatGPT to write essays showed less activity in brain networks associated with cognitive processing while undertaking the exercise.

These people also couldn't quote from their essays as easily as those in the study who didn't use an AI chatbot.

The researchers said their study demonstrated "the pressing matter of exploring a possible decrease in learning skills".

All 54 participants were recruited from MIT and nearby universities. Their brain activity was recorded using electroencephalography (EEG), which involves electrodes being placed on the scalp.

Some of the prompts used by the participants included asking AI to summarise essay questions, track down sources as well as refine grammar and style.

It was also used to generate and articulate ideas - but some users felt AI wasn't very good at this.

'AI makes it too easy to find answers'

Separately, Carnegie Mellon University and Microsoft, which operates Copilot, found people's problem-solving skills could diminish if they became too reliant on AI.

They surveyed 319 white-collar workers who used AI tools for their jobs at least once per week about how they apply critical thinking when using them.

They looked at 900 examples of tasks given to AI, ranging from analysing data for new insights to checking whether a piece of work satisfies particular rules.

The study found that higher confidence in the tool's ability to perform a task was related to "less critical thinking effort".

"While GenAI can improve worker efficiency, it can inhibit critical engagement with work and can potentially lead to long-term overreliance on the tool and diminished skill for independent problem-solving."

Schoolchildren in the UK were similarly surveyed for a study published in October by Oxford University Press (OUP).

It found six in 10 felt AI had negatively impacted their skills in relation to schoolwork.

So, with the massive explosion of AI use, are our cognitive skills at risk of decline?

Klaus Vedfelt/Getty Students seated at desks in a classroom, with several raising their hands holding pens. Open notebooks and water bottles are visible on desks. At the front, a teacher stands near a whiteboard displaying a projected presentation with circular diagrams.Klaus Vedfelt/Getty
A survey by OUP found six in 10 schoolchildren felt AI had negatively impacted their skills

Not necessarily, says Dr Alexandra Tomescu, a generative AI specialist at OUP who worked on the school survey.

"Our research tells us that nine in 10 students say AI has helped them develop at least one skill related to schoolwork - be it problem-solving, creativity or revision.

"But at the same time, about a quarter state that AI use made it too easy to do work for them… So [it's] quite a nuanced picture."

She adds that many pupils want more guidance on how to use AI.

ChatGPT, which has more than 800 million weekly active users according to boss Sam Altman, has published a set of 100 prompts for students designed to help them get the most out of the technology.

But Prof Wayne Holmes, who researches critical studies of artificial intelligence and education at University College London (UCL), says this isn't enough.

He wants much more academic research to be done about the effects of AI tools on learning before pupils and students are encouraged to use them.

He tells us: "Today there is no independent evidence at scale for the effectiveness of these tools in education, or for their safety, or even for the idea they have a positive impact."

Better results but worse learning?

Prof Holmes points to research about cognitive atrophy, where someone's abilities and skills become worse after using AI.

He says this has been a problem for radiologists who use AI tools to help them interpret X-rays before they diagnose patients.

A study by Harvard Medical School published last year found AI assistance did improve the performance of some clinicians but damaged others for reasons researchers don't fully understand.

The authors called for more work to be done on how humans interact with AI so we can figure out ways of using AI tools that "boost human performance rather than hurt it".

Prof Holmes fears that students, whether in school or university, could become too reliant on AI to do their work for them and not develop the fundamental skills an education provides.

A student's essay might receive better marks thanks to help from AI but the issue is whether they end up understanding less.

As Prof Holmes puts it: "Their outputs are better but actually their learning is worse."

Jayna Devani, who leads international education at OpenAI - the company that owns ChatGPT - and helped secure a deal with the University of Oxford, says the firm are "very aware of this debate right now".

The Bodleian Library surrounded by a black metal fence and green lawn. Several bicycles are parked along the fence. Adjacent Gothic-style buildings with spires and arched windows are visible under a clear blue sky.
The University of Oxford began providing students and staff with ChatGPT for free in September

She tells the BBC: "We definitely don't think students should be using ChatGPT to outsource work".

In her view, it's best used as a tutor rather than just a provider of answers.

The example she gives is of a student having a back and forth with ChatGPT using the study mode setting.

You enter the question you're having difficulty answering and the chatbot can break down its components and help you understand it.

The example she gives is of a student doing an assignment late at night about a topic they don't quite understand.

"[If] you have an upcoming presentation to give and… it's midnight, you're not going to email your [university] tutor and ask for help," she says.

"I think the potential is truly there for ChatGPT to accelerate learning when it's used in a targeted way."

But Prof Holmes insists that any student who uses AI tools should be aware of how its reasoning works and how the companies providing them handle data. He stresses that results should always be checked.

"It is not just the latest iteration of the calculator," he says, describing AI's far-reaching capabilities and implications.

"I never say to my students, you shouldn't use AI… But what I do try to say is look, we need to understand all these different things about it so that you can make informed decisions."

The Papers: 'Epstein files and photos released' and 'Walliams is axed'

"Thousands of Epstein files and photos released" reads the headline on the front page of the Times.
The Times leads with a new tranche of documents released by the US Department of Justice, including "a previously unseen photograph of Andrew Mountbatten-Windsor lying across the laps of five women". The photo appears to show the former prince, Ghislaine Maxwell, and redacted faces of six people. Andrew has repeatedly denied all wrongdoing, and says he did not "see, witness or suspect any behaviour of the sort that subsequently led to his arrest and conviction". Being named or pictured in the files is not an indication of wrongdoing.
"Wills joins the SAS: Proud Prince of Wales to be patron of elite unit's welfare charity" reads the headline on the front page of the Daily Mirror, featured below a photo appearing to show Andrew Mountbatten-Windsor lying across the laps of five women.
The Prince of Wales "joins the SAS" as a "patron of [the] elite unit's charity", which "supports soldiers, veterans and families", reports the Daily Mirror citing an unnamed royal source. Also featured on the front page is the black-and-white photo of Andrew "draped across women's laps... as Maxwell looks on", according to the paper.
"Walliams 'harrassed female staff'" reads the headline on the front page of the Daily Telegraph.
The Daily Telegraph leads with publisher HarperCollins UK dropping best-selling author and comedian David Walliams. The paper says it investigated allegations of inappropriate behaviour towards young women. The publisher does not specify the reason for dismissing Walliams. A spokesperson for Walliams says he has "never been informed of any allegations raised against him" by his publisher. "He was not party to any investigation or given any opportunity to answer questions. David strongly denies that he has behaved inappropriately and is taking legal advice," the spokesperson adds.
"Walliams is axed: Star 'harassed' young women" reads the headline on the front page of the Daily Star.
The Walliams story also leads the Daily Star's front page. A spokesperson for HarperCollins says it has decided not to release any new titles by Walliams "after careful consideration, and under the leadership of its new CEO".
"Reeves ruins Christmas: Shops suffer disastrous December after £30bn Budget tax raid", reads the headline on the front page of the Daily Mail.
Chancellor Rachel Reeves is accused of ruining Christmas for retailers, reports the Daily Mail. It says customers "surprisingly shunned" Black Friday sales, as retailers say "December has been disastrous, with the 'bleak' outlook stretching into January". Critics cited by the paper blame the government's "£30bn Budget tax raid".
"Reeves: I can't rule out tax rises in 2026" reads the headline on the front page of the i Paper.
The i Paper leads with its interview of Reeves, who does not "rule out tax rises in 2026". The paper quotes the chancellor saying she hopes "further changes to tax are less necessary", while noting "the world is incredibly volatile at the moment". Her remarks follow the government's "£66bn in tax hikes" laid out in the November Budget, which the paper reports led to accusations of "breaking the Labour manifesto pledge not to increase income tax, NI (national insurance), or VAT (value-added tax)".
"EU agrees €90bn loan to Ukraine" reads the headline on the front page of the FT Weekend.
The Financial Times leads with the EU's €90bn (£79bn) loan to Ukraine, calling it a "lifeline for hard-up Kyiv". It notes the bloc's "plan to use frozen Russian assets collapsed" following objections from Belgium, warning it would be legally and financial liable.
"Ukraine gets €90bn boost – and blows up Russian ship in Med" reads the headline on the front page of the Independent.
The Independent leads with the latest in Ukraine, combining the EU's €90bn (£79bn) loan and strike on a Russian oil tanker in the Mediterranean Sea in its headline. Russian President Vladimir Putin's response is also included, "as he accuses Western allies of 'dirty tricks'" over the European loan and "vows revenge for (Ukrainian President Volodymyr) Zelensky's drone attack".
"Robot surgery revolution to save NHS millions" reads the headline on the front page of the Daily Express.
A "robot surgery revolution" is splashed on the Daily Express's front page. It says the new technology could "cut patients' recovery time and free up beds".
"No 10 told: fix youth jobs or risk a lost generation'" reads the headline on the front page of the Guardian.
Warnings to the government that it could "risk a 'lost generation'" due to "rising inactivity among young people" leads the Guardian. Former health secretary Alan Milburn tells the paper the government must tackle some "uncomfortable truths" about the labour market, as he leads an inquiry that will include a review into the "future of the youth minimum wage".
"James gets down on one knead" reads the headline on the front page of the Sun.
And finally, the Sun reports TV chef James Martin has announced he is engaged to his personal trainer partner Kim Johnson. The paper writes he got "down on one knead".
News Daily banner

Sign up for our morning newsletter and get BBC News in your inbox.

News Daily banner

Blank canvas or tone-deaf? Pantone's white Colour of the Year sparks backlash

Pantone Person wearing all white in front of a backdrop of sky and cloudsPantone
Cloud Dancer has been named Pantone's Colour of the Year

For anyone who has spent years renting, staring at the same landlord-approved magnolia walls and dreaming of one day adding some actual personality to their home, Pantone's 2026 Colour of the Year might feel like a personal attack.

This year, Pantone has chosen white as its colour of the year.

More specifically, Cloud Dancer - a white described as "lofty" and "billowy" which "serves as a symbol of calming influence in a society rediscovering the value of quiet reflection".

It is the first time white has been chosen as the colour of the year, and it has raised more than a few eyebrows.

According to Pantone, Cloud Dancer "is not just a colour, it's a mindset" and reflects a collective desire to slow down, reset and find calm after years of visual overload.

Pantone White room with white curtains in the background and fluffy white cushions in the foregroundPantone
An all-white take on Pantone's Cloud Dancer – soothing in theory, stressful if you drink red wine or coffee

But some critics have labelled the decision as "Pantonedeaf", arguing that elevating minimalist and pristine white spaces can feel far removed from the smaller, messier homes most people inhabit.

Others argue that positioning white as aspirational risks racial undertones, landing uncomfortably amid ongoing political and cultural conversations about race and representation.

Clinical or calm?

Designer Chris Beaumont says the backlash goes beyond aesthetics, arguing that white carries cultural undertones and is rarely a neutral choice.

"White is a signal," he says, shaped by a decade of minimalist celebrity interiors – most notably Kim Kardashian's stripped-back home – that came to represent "wealth, order and distance from chaos".

He explains that white is "not about inspiration but being careful not to offend", with Cloud Dancer embodying "austerity, moral minimalism and the idea that neutrality signifies virtue".

Chris Beaumont A man, Chris Beaumont, wearing a salmon pink coloured blazer and blue and red flowery shirt, poses in his home. His home is similarly colourful featuring a mixture of patterns on furniture including a sofa and bar area Chris Beaumont
Chris believes Pantone's choice for 2026 is "tone-deaf"

Beaumont points to the pandemic as a turning point in how people relate to their homes.

"Overnight they became our offices, sanctuaries and emotional anchors," he says, adding that against this backdrop, pushing the Colour of the Year towards "further visual emptiness feels rather tone-deaf".

Rather than offering calm, he believes white now risks amplifying "a sense of bleakness", particularly when paired with cool, clinical lighting.

'Hospital vibes'

Lara Clark, a Surrey based interior designer, agrees and says Cloud Dancer "doesn't really read as a colour" and is "firmly in the bin for me".

While it may suit highly minimal or architectural spaces, she says bright whites rarely create calm in real homes.

"What looks serene in a styled shoot can easily feel clinical at home," she explains.

"White can feel stark and unforgiving and you don't want your home to give hospital vibes."

Lara Clark Blonde woman looking at a colour chartLara Clark
Lara believes using white in real homes can create a clinical vibe

She adds that "homes should spark joy and feel warm and lived-in, and this shade feels quite removed from that."

Beaumont hopes that homeowners will reject the Pantone Colour of the Year in favour of "full-scale dopamine décor", using colour to express personality.

Laurie Pressman, VP of the Pantone Colour Institute, told the BBC that people "bring different feelings" to the meaning of the colour, but it was chosen as it "works seamlessly with everything around it, offering a refined neutrality that feels intentional and adaptable".

Since its launch in 2000, Pantone's Colour of the Year has become a powerful industry signal.

Past choices have included the optimism of Living Coral, the calm confidence of Classic Blue, the vibrancy of Viva Magenta, and most recently softer, emotionally driven tones like Peach Fuzz.

These colours don't just live on trend forecasts - they show up in fashion collections, beauty launches, interiors, packaging and even technology, shaping how products are marketed and how consumers imagine the year ahead.

Carlos Avila Gonzalez/The San Francisco Chronicle via Getty Images Sephora cosmetics including a blush palate, mascara, lipstick and a blush brush is shown. They are in a pink-ish / nude colour called Marsala which was picked as Pantone's colour of the year in 2015Carlos Avila Gonzalez/The San Francisco Chronicle via Getty Images
Sephora cosmetics previously released a collection based on Pantone's 2015 Colour of the Year - Marsala - described as a "full-bodied red-brown"

Victoria Robinson, a style and trend expert, says Cloud Dancer is a "beautiful choice" and while it may seem simple, "this particular shade feels soft and elegant rather than stark".

In contrast to those who say the colour is boring, Robinson sees it as "adaptable" and says it's best used in "bedrooms and living areas where you want a serene, restful atmosphere".

"Even if you don't want to repaint, you can introduce the colour into a room with cushions, throws and curtains."

Pantone Woman wearing brown sunglasses a dark brown turtleneck and light brown fluffy coat with only the collar visible. Pantone
Pantone's Colour of the Year for 2025 was Mocha Mousse - a soft brown which "promotes a pure and organic approach to luxury"

Interior designer James Mellan-Matulewicz says although he was surprised that Pantone's pick this year was "essentially the absence of colour", it is a bit like vanilla ice cream, in that "everyone likes it but it's nobody's favourite".

He can see its merits, explaining that white can work particularly well as a backdrop for architectural details like panelling and arched doorways which is a "growing trend in modern homes".

In fashion, white has long been a staple rather than a statement and as a Colour of the Year it presents a different challenge to designers more accustomed to bold shades.

Luxury stylist Oriona Robb says elevating white feels "less about novelty and more about intention", reflecting a shift towards refinement after years of trend overloads.

Oriona Robb Image of a blond woman sitting at a table with magazine and colour wheels on the table and a mood board behind her of models wearing different outfitsOriona Robb
Oriona says white is a hard colour to pull off in fashion

"White forces designers and wearers alike to focus on form, proportion and quality, there's really nowhere to hide," she explains.

But she adds that it also carries assumptions around body confidence, lifestyle and privilege, saying: "When white is treated as something only a narrow group of people can pull off, it becomes exclusionary."

She says industry is already aware of the uncomfortable undertones tied to celebrating white as an ideal, particularly amid ongoing conversations about representation and accessibility, and the real test will be whether "brands engage with that nuance honestly, or simply aestheticise the colour".

A cultural mood, not a trend

Stylist Katie Malik admits the choice initially surprised her, given Pantone's history of bold colours, but says it reflects a genuine shift and fits within a wider mood of "quiet luxury", burnout and a rejection of excess.

She says feedback from her clients has been largely positive, with many craving calmer, more restorative spaces.

"Many people are actively seeking tranquillity and serenity in their homes and aren't always ready to commit to more daring colours," she explains.

Whether white feels calming or sterile, Malik argues, depends on how it is used and far from being "Pantonedeaf", she sees it as one of Pantone's most usable picks.

Cloud Dancer is described as a blank canvas that allows "all colours to shine", a view Malik shares.

"A blank canvas isn't an empty space, it's a space filled with potential," she says.

Adding that its success "won't be in its universal adoption, but in how it anchors a larger conversation about what we want from our homes".

For those tempted to embrace Cloud Dancer in all its pristine glory, one thing may be essential: a very good stain remover, kept firmly within reach.

巴黎法院驳回政府要求希音暂停在法平台运营

法国巴黎一家法院星期五驳回政府对中国跨境快时尚巨头希音(Shein)提出暂停在法平台运营三个月的诉求。

综合法新社和彭博社报道,法国政府以希音在平台售卖性玩偶和违禁药物等,提出上述诉求,并要求希音制定严格措施确保不会再犯,才能重启平台(marketplace)运营。

巴黎法院星期五(12月19日)作出裁决时确认,相关行为确实“严重危害公共秩序”,但发现被指不得售卖的物品数量“零星”,希音在接获指令后也迅速下架相关物品。对希音实施禁令是“过分”的要求,也是对自由企业的无理侵犯。

法院称,希音不能仅靠简单声明,而是要列明年龄验证措施,才能重新出售性玩偶。

法国政府原本打算通过无需法院批准的行政程序对希音实施禁令,但由于希音下架原本销售的性玩偶和违禁药物等,并同时暂停平台运营,该行政程序未启动。

在法国官员发出警告后,希音11月全面禁止在平台上销售性玩偶,并承诺将“全力配合”法国司法机关的调查。

中国禁平台经营者为独占市场以低于成本价卖商品

中国出台《互联网平台价格行为规则》,规定平台经营者不得以排挤竞争对手或者独占市场为目的,以低于成本的价格销售商品或者提供服务。

《互联网平台价格行为规则》自2026年4月10日起施行,有效期为五年。

根据中国网信网星期六(12月20日)发布的规则内容,平台经营者、平台内经营者不得违反《中华人民共和国价格法》第十四条第二项规定,以排挤竞争对手或者独占市场为目的,以低于成本的价格销售商品或者提供服务,扰乱正常的生产经营秩序,损害国家利益或者其他经营者合法权益。

不过,降价处理鲜活商品、季节性商品、积压商品、临期商品等商品,或者有正当理由降价提供服务的除外。

平台经营者也不得强制或者变相强制平台内经营者按照其定价规则,以低于成本的价格销售商品或者提供服务,扰乱市场竞争秩序。

平台经营者的商业模式系对用户长期免费的,且有利于推动创新进步、有利于提升经营者和消费者长远福利的,可以不认定为违反《中华人民共和国价格法》第十四条第二项规定。

此外,平台经营者也应当公平公正开展补贴促销,不得虚假、夸大宣传补贴金额和力度。平台经营者开展补贴促销,应当在网站或者应用程序(APP)相应活动页面显著位置标示补贴及相关促销活动规则,明确补贴对象、补贴方式、参与条件、起止时间等信息。

北京同仁堂为磷虾油产品零磷脂事件致歉

中国北京同仁堂一款南极磷虾油产品被发现没有磷脂含量后,集团誓言一查到底,并说责令涉事经销主体下架涉事产品。

上海市消费者权益保护委员会本月11日在微信公众号公布,一款标称“北京同仁堂99%高纯南极磷虾油”的产品,在网页上写着“高含量”、“智利进口原料”、“高度提纯”、“1粒≈100只磷虾”,磷脂标识值为43%,看似是品质出众的优质产品,但检测结果显示该款产品磷脂含量为0。

消保委通报,针对相关企业涉嫌造假行为,为了保护消费者合法权益,上海市消保委“将于近日约谈该产品的生产企业安徽哈博药业有限公司和经销商北京同仁堂(四川)健康药业有限公司,要求其对产品的真实情况做出解释”。

北京同仁堂四天后(12月15日)通过“同仁堂健康会员”微信公众号指这款产品未经授权擅自突出使用“北京同仁堂”字样,涉嫌违法,并已启动司法程序,对涉事企业进行起诉。

澎湃新闻之后报道,被上海市消保委点名的产品具体名称为“南极磷虾油凝胶糖果”,经销商为“北京同仁堂(四川)健康药业有限公司”,生产企业为“安徽哈博药业有限公司”。生产商和经销商12月15日被约谈。

约谈中,生产商安徽哈博药业有限公司陈述,涉事产品由经销商北京同仁堂(四川)健康药业有限公司定制采购,采购价格远低于正常水平,产品包装按照北京同仁堂(四川)健康药业有限公司要求制作,所有涉事产品均供给北京同仁堂(四川)健康药业有限公司,由其对外销售。

安徽哈博药业有限公司承认,在生产过程中未添加其对外宣称的 “南极磷虾油”。澎湃新闻报道,经销商北京同仁堂(四川)健康药业有限公司则在约谈中回避问题,推脱责任,声称对产品的涉嫌造假行为毫不知情,与己无关。

报道称,涉事企业用鱼油提取物冒充南极磷虾油进行生产销售。

北京同仁堂星期六(20日)在“中国北京同仁堂官微”微信公众号发声明称,针对有关南极磷虾油的舆论关注,“北京同仁堂集团高度重视、深感痛心,对因此给消费者造成的权益损害,向广大消费者和社会各界致以最诚恳的歉意”。

集团公布,已派专项工作组到四川,对涉事产品开展全面深入核查,坚决一查到底,并责令并监督涉事经销主体四川健康立即下架涉事产品,严格按照监管部门要求依法依规对消费者进行赔偿。

集团也宣布,依法追究四川健康、哈博药业及相关电商平台的商标侵权等责任;对其违法使用“北京同仁堂”字样,误导消费者判断,损害消费者合法权益的行为追究法律责任;对四川健康及其控股股东健康药业等单位在经营、管理、监督环节存在的问题开展问责,严肃追究相关人员责任。

北京同仁堂也宣布,全面排查品牌使用中的违法违规行为,对集团各级企业的品牌使用情况进行拉网式核查;全面开展品牌清理行动,坚决清理“擦边”使用“同仁堂”字号、商标使用不规范等行为,严格规范品牌管理体系。

南京导航出现异象 官方称卫星信号受干扰

针对南京网民反映称,所使用的导航软件出现异常情况,南京卫星协会回应称,包括北斗和GPS在内的卫星信号受到干扰。

综合每日经济新闻和快科技报道,有南京网民星期三(12月17日)反映,所使用的导航软件系统出现突发异常状况。导航界面持续显示“行驶在无数据道路上”,同时外卖、共享单车等日常服务也受到影响,有外卖骑手反映外卖软件等内置导航出现问题,显示“看不见导航”或导致外卖配送延时。

南京卫星应用行业协会星期五(12月19日)通过“南京卫星协会”微信公众号发布《关于12月17日南京区域导航信号异常情况的技术性探讨分析》指出,这次导航失灵的核心原因是全球导航卫星系统(GNSS)卫星信号(含北斗、GPS)受到临时干扰压制,而非网络信号中断,因此下载离线地图并无实际意义。

协会称, 离线地图的核心功能是存储地理数据,仅解决“地图显示”问题,无法替代定位信号来源;手机导航需依赖GNSS卫星信号(至少四颗卫星)解算位置坐标,再将位置信息匹配至地图,定位信号与地图数据是相互独立的两大系统。

协会说,这次干扰精准针对北斗、GPS民用频段,导致定位信号无法被接收机识别,即便搭载离线地图,也因缺乏位置数据源出现定位漂移、无数据反馈等问题,无法实现有效导航。这次信号异常仅涉及民用GNSS频段,北斗军用频率完全不受干扰。

港官员:政府部门成员欺瞒火灾委员会 将有严肃法律后果

香港特区官员星期五(12月19日)主持会议,讲解政府部门如何配合针对宏福苑火灾而成立的独立委员会调查工作,并强调任何欺骗、隐瞒或误导行为将面临严肃的法律后果。

香港律政司司长林定国在脸书发文写道,他向向十多个政策局及部门首长讲解政府部门应如何配合独立委员会的调查工作,包括准备事项及应遵循的原则。

林定国说,律政司已就此成立专责小组,由他亲自监督领导。小组负责协助政府部门向独立委员会提供证据和资料,并解答相关法律问题。

林定国续称:“我向所有同事强调,政府必须坚守两大原则配合独立委员会的调查工作。一是,主动性。所有政府部门均有责任主动协助调查,向独立委员会提交职权范围相关的材料,而非被动等待查询;二是,全面及坦诚的资料提供(full and frank disclosure)。只要与调查范围相关,即使资料可能对部门或个人构成负面影响,也必须如实提供。任何欺骗、隐瞒或误导行为将面临严肃的法律后果。”

他指出,独立委员会须在九个月内提交报告,并说:“时间上相当紧迫,我们在会上也详细讨论了具体工作方式及时间表,以确保高效运作。”

林定国也说:“我们会立即全面检视现行制度,针对存在的问题进行修正与改革,透过相应改善措施,确保同类不幸事件不会重演。此外,我们更要推动大厦大型维修工程在充分保障业主整体合法权益的前提下,以公平、安全且高效的方式进行。”

这场火灾是香港近80年来最严重火灾,造成160人死亡。特首李家超上星期成立独立委员会,审视大埔宏福苑火灾的事故原因及相关问题,提出建议,防止同类事故再次发生。

河南小学火灾案 三人获刑

中国河南省一所学校发生致命火灾,官方经过两年调查后对25人进行处罚。此案星期五一审宣判,包括创始人在内的三人获判监禁。

综合中国央视新闻和澎湃新闻报道,河南省方城县法院星期五(12月19日)上午9时开庭审理河南方城县独树镇砚山铺村英才学校火灾案,经过10小时的庭审后,与当晚宣判,以犯教育设施重大安全事故罪,对学校创办人、实际控制人李宇、校长徐向阳和学校起火宿舍宿管员贾霞,分别判处七年、六年半和六年有期徒刑。

法院同时禁止李宇、徐向阳、贾霞自刑罚执行完毕之日起,五年内从事与教育教学相关的职业,并驳回附带民事诉讼原告人范永归等人的诉讼请求。

英才学校宿舍去年1月19日深夜发生火灾,造成13名学生遇难、四名学生受伤。

事发后,英才学校与13名死亡小学生的近亲属签订赔偿协议,赔偿每名遇难学生近亲属各项费用130万元人民币(24万新元)。上述赔偿款已于2024年1月支付完毕。  

河南省政府事发生后牵头成立事故调查组,对起火原因进行了深入调查,认定事故性质和责任,并在星期三(12月17日)发布调查报告,对包括南阳市副市长在内的25人,因监督管理不到位等原因作出处理处分。

公安机关对英才学校五名相关人员采取刑事强制措施,其中李宇、徐向阳和贾霞等三人被检察机关以犯教育设施重大安全事故罪提起公诉,其他两人因未参与学校实际管理,不构成犯罪,不予追究刑事责任。

法院认为,李宇、徐向阳、贾霞明知学校校舍及教育教学设施有危险,既不采取措施整改,也不及时报告,致使发生重大伤亡事故,其行为均已构成教育设施重大安全事故罪。这次火灾事故后果特别严重,社会影响极其恶劣,对三名被告应依法从严惩处。

法院称,李宇、徐向阳、贾霞违背职业要求,怠于履行学校安全管理职责,在判处刑罚的同时,也禁止其在刑罚执行完毕之日起,五年内从事教育教学相关的职业。附带民事诉讼原告人范永归等人案发后已依法获得赔偿,对其诉讼请求依法不予支持。

26 元/小时,人人喊贵,广州终于要对天价停车费出手了

有时候经常被朋友问,「你不是做汽车相关内容的嘛,为什么不见你开车上班?」

其实很多时候不是不想开,而是真的停不起。

公司所在的广州塔附近大概是全广州停车收费最高的地区之一,露天停车场白天工作时间段 16 元一小时,月保在千元左右,晚上回到家的小区内稍微低一点,临停 5 元一小时,月保 600 元。

也就是说如果每天开车通勤加上周末偶尔去商场,光停车费一项,每月额外支出就要接近 2000 元,足够在广州再租个一居室。

正因如此,尽管公司有位已提车 1000 天的理想车主,我却始终未能蹭到他的车,十分遗憾。

▲ 广州塔小车收费 16 元/小时

长期以来,广州的停车费一直居高不下。几年前,界面新闻曾做过一项统计,广州重点区域路边停车场的平均成本高达 26 元/小时,这一数据在全国遥遥领先。作为对比,同为一线城市的上海、深圳约为 20 元/小时,北京仅为 10 元/小时,而新一线城市武汉的 4 元/小时甚至排到了全国第十一名。

理论上来讲,深圳本应是停车更难、更贵的城市。

深圳的土地面积仅为广州的三分之一,却承载了更多的汽车。截至 2025 年 8 月,深圳机动车保有量已达 460 万辆,远超广州的 404 万辆。在人均 GDP 和居民收入均高于广州的背景下,深圳的支付能力理应更强,土地资源理应更稀缺。

然而,现实却给出了相反的答案。

在深圳的一类区域(如福田 CBD ),停车场收费大多为第一小时 15 元,此后每小时 5 元。而在广州同类区域,往往起步就是每小时 16 元甚至更高,且阶梯涨价更猛烈。

深圳这种「地更少、人更富、车更多,停车反而更便宜」的现象,恰恰说明了广州的高价并非完全由供需决定,更多是由于定价机制的传导失效与存量管理的粗放。

以广州天河 CBD 珠江新城为例,这里的停车费被车主们戏称为「天价刺客」。临时停车费最高可达 26 元/小时,单日封顶 328 元。而在一些更高端的商业场所,价格直逼 30 元/小时且不设上限。

▲ 广州太古汇停车场,周末 20 元一小时,不设最高限价

小区的月保车位也不便宜。

有网友分享,其在越秀某小区的停车场月租为 1800 元/月,周边小区多数也都在 1500 元往上,而且还要靠抢。

▲2000 元的非固定月租车位

高昂的停车成本,迫使大量市民在出行方式上做出妥协。

大部分的广州家庭往往采用「一辆四轮+一辆两轮」的出行配置。毕竟骑着电鸡出门,不仅可以避开拥堵,还可以避开无法预估的停车费账单。毕竟,任何时间、任何地点,电鸡随便停且免费。

价格高不过只是表象。停车难、停车贵的归根结底,是城市空间资源的极度稀缺与分配不均。

官方数据显示,截至 2025 年 6 月,广州全市停车位约 437.8 万个,小汽车保有量约 404.4 万辆。乍看之下,1.08 的车位比似乎意味着供需基本平衡。

但汽车是流动的,一个高效运转的城市交通网络,要求一辆车不仅要有一个「家」(居住停车位),还需要有「办公室」(工作停车位)和「游乐场」(商业/公共停车位)。

通常认为,理想的城市停车位供需比应达到 1.2 甚至更高,才能应对车辆流动带来的潮汐式需求。

但在广州,停车资源往往是失衡的。越是核心、越是老旧的区域,停车矛盾越是尖锐。

在越秀、荔湾等老城区,以及医院、学校、商圈周边,停车难是常态化的「重灾区」。对于居住在老旧小区的居民来说,每天下班回家就是一场现实版的「抢车位」游戏。由于车位极度匮乏,车辆不得不层层叠叠地停放。

这里衍生出了一种独特的「熟人社会治理模式」,管停车场的大爷,手里往往握着全小区几十把车钥匙。哪辆车要出,哪辆车要进,全凭大爷运筹帷幄,「开完你的,开他的」。

因为车位属于稀缺资源,很多地方甚至出现了「寻租」空间,有的业主甚至需要私下「打点」物业,才能勉强在小区里拥有一个车位。

即便是在硬件条件较好的新建住宅区,矛盾也同样存在。部分住宅停车收费随意定价、强行涨价的现象屡见不鲜,物业方利用信息不对称和管理优势,单方面提高停车费,让业主们苦不堪言。前段时间,某豪宅物业甚至做出了业主不同意涨停车费就不放车辆通行的离奇操作。

高得无法理解,涨得毫无道理,呈现出一种反正就是贵,反正我要涨的神奇局面。

▲ 老小区停车现状

面对各种问题交织的困境,即将于 2026 年 2 月实施的《广州市停车场条例》试图通过一套组合拳来打破僵局。

通读新规,可以清晰地看到立法者试图在「存量挖掘」和「规则重塑」两个维度上寻找突破口。

首先是「存量挖掘」。

广州市人大常委会法工委有关负责人在解读中提到,资源利用效率不高是导致车位紧缺与闲置并存的重要原因。

为此,新条例要求「要求本市国家机关、事业单位的自用停车场与办公场所分开管理的,应当在非工作时间有序向社会开放」。并且这些车位需要置明显标识,并将泊位信息接入全市停车信息管理系统,所得收益上缴财政。

这一举措若能落地,将极大缓解老城区夜间「一位难求」的现状。

其次是支持支持和鼓励社会力量投资建设公共停车场。

对于社会力量投资建设公共停车场的,不仅可以给予适当的资金支持,还可以在在不改变用地性质、不减少停车位的前提下,配建不超过公共停车场建筑面积百分之二十的附属商业设施。

第三,在我们最为关注的定价权博弈方面,新条例引入了「协商议价」机制。如果停车场经营者想制定或提高收费标准,必须与业委会或业主大会协商。

为了保证这一条款不沦为一纸空文,新规设定了高额的违法成本,未依法协商议价的,最高可处 50 万元罚款。

最后,是对路面秩序的「收权」。

针对路内泊位「乱划线」的乱象,新条例明确了区交通运输行政主管部门的唯一管理权,严禁任何单位和个人擅自划设或撤除。对于私自圈地划线的行为,个人最高罚款 5000 元,单位最高罚款 2 万元,并没收违法所得。

这意味着,过去那种随便拿桶油漆就在路边圈地收钱的「混乱时代」,将彻底终结。

即使在 2026 年 2 月之后,广州车主可能还是会为找车位发愁,老城区的巷子依然窄,医院门口的车龙也不会一夜消失。

但至少,规则更清楚了,住宅停车费不能全由物业说了算,得和业主商量;路边划线谁有权、谁违规,也有了明确罚则;机关单位的停车场在晚上能不能用,也不再是「内部事务」。

这些变化未必立刻缓解一位难求的窘境,但至少让市民在面对涨价、占位或混乱管理时,知道该找谁、依据是什么。

在一座常住人口超过 1800 万、机动车保有量已突破 400 万辆的城市里,「停好一辆车」从来不是小事。它牵涉空间规划、公共资源分配、社区治理,甚至日常生活的尊严感。

新规无法凭空变出车位,但若能推动闲置资源流动起来、遏制随意定价、减少管理盲区,就已经是在复杂现实里往前走了一步。

#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。

爱范儿 | 原文链接 · 查看评论 · 新浪微博


❌