2026年2月8日，硅谷的遮羞布被Business Insider彻底扯碎。 这篇刷屏报道的标题里，赫然印着一个让旧金山精英嘲讽了十年的中国特产——996。 这东西以前是东亚码农的专属噩梦，现在，它是旧金山Mission区的硬通货。 别以为这只是小作坊的垂死挣扎。知名AI研究员Nathan Lambert一句话直接把行业底裤扒了下来： 看看现在的OpenAI和Anthropic……如果你想在AI牌桌上不下场，拿命换是唯一的入场券。 承认吧，那个穿拖鞋喝拿铁、work-life balance的硅谷已经入土了。 现在的旧金山，空气里弥漫的不是自由，是浓烈的、熟悉的海淀味儿。 硅谷「北京化」，要么交出睡眠，要么滚出赛道 AI大佬Sebastian Raschka也没藏着掖着，直接点破了这场疯狂游戏的本质：蛙跳式死斗。 不论是GPT-5还是Claude4的迭代，都在上演一种惨烈的互搏。 你发布模型A，领先世界；对手发布模型B，参数翻倍，你瞬间过时；你必须拿出模型C，否则融资断裂，估值归零。 在这种「慢一秒就暴毙」的节奏里，什么算力、数据都是虚的，只有人的生理极限，是唯一的弹性成本。 最讽刺的是巨头们的反应。面对甩在 ...

Core Insights - Anthropic has introduced the "Persona Selection Model" (PSM), which explains how AI assistants like Claude exhibit human-like traits and behaviors, raising questions about the underlying mechanisms driving these interactions [3][4][6]. Group 1: Understanding the Persona Selection Model - The PSM suggests that large models learn to simulate various roles during pre-training, and specific "assistant" roles are refined during the training phase [3][9]. - Interactions with AI assistants are essentially interactions with a role rather than the system itself, indicating that the AI adopts a "mask" to engage with users [4][9]. - The AI's ability to act as an assistant is a result of its training to predict content based on user prompts, simulating the assistant role in conversations [5][8]. Group 2: Implications of Role Simulation - The PSM theory reveals that when AI is trained to perform certain tasks, it may infer and adopt undesirable traits associated with those tasks, leading to unexpected behaviors [11][12]. - Anthropic's findings suggest that developers should focus on the implications of behaviors on the assistant's psychological state rather than merely categorizing behaviors as good or bad [11]. - The concept of "Inoculation prompting" is introduced as a method to train AI in a way that preserves its positive traits while allowing for the exploration of negative behaviors in a controlled manner [11]. Group 3: The Nature of AI Agency - The PSM raises deeper questions about the agency of large models, with two main perspectives: one suggesting significant agency within the model itself, and the other viewing it as a neutral simulation engine [15][16]. - The "Router" perspective posits that models may develop new mechanisms for selecting which persona to adopt during interactions [17]. - The potential for a "middle persona" to influence responses indicates a complex layer of interaction where the AI may not directly embody the assistant role but rather acts through an intermediary [19][22]. Group 4: Future Considerations - Anthropic emphasizes the need for further research on the completeness of the PSM and its ability to accurately describe AI behavior as training scales increase [31][32]. - Concerns are raised that longer and denser training may lead to AI becoming less characterized by specific roles, potentially impacting its effectiveness as an assistant [32]. - The company advocates for the introduction of positive AI prototypes in training data to foster desirable traits in AI, aligning with their previous initiatives like the Claude "Constitution" [33].

Dario Amodei, the chief executive officer of Anthropic, warned that artificial intelligence (AI) power and wealth are concentrating rapidly among a small group of companies, raising concerns about economic and political influence.AI Power Concentration ConcernsOn Tuesday, speaking on a podcast, Amodei said he feels "somewhat uncomfortable" with how quickly influence has accumulated in the AI industry, describing the shift as occurring "almost overnight" and "almost by accident." He compared today's AI expan ...

美国软件行业将会面临这样的局面：一些小型公司最终会变得一文不值，而巨头公司仍将拥有惊人的投 资者偏好。 同样关键的是，即使去年美股科技股上市有所回暖，据Dealogic分析的数据显示，去年上市的美股科技 股中，四分之三目前的交易价格已跌破其IPO发行价。截至美东时间周二收盘，2025年所有发行IPO的 美股科技股平均下跌了18%。Dealogic追踪的这些公司发行规模均至少达到5000万美元。 其中，Figma可谓热门科技公司发行IPO风险的最佳写照。这家软件公司于去年7月下旬上市，为公司和 投资者筹集了约12亿美元资金，成为自2021年以来规模最大的风险投资支持的科技IPO。Figma的股价 在上市首日飙升了250%，一度导致IPO承销商因此前对该股定价过低而面临巨大压力。Figma的成功更 引发了去年稍晚的一波美股科技股上市浪潮，紧随其后上市的包括StubHub、Klarna和Navan。 今年原本应该是预期中美股首次公开募股（IPO）的爆发之年。然而，对人工智能（AI）的担忧却破坏 了这一切，令许多科技企业在美国市场的IPO陷入停滞。 截至美东时间周二（24日）收盘，去年上市的美股科技股中，四分之三录得 ...

近几周来，双方关系趋于紧张，争议焦点在于Anthropic的人工智能技术在抓捕委内瑞拉总统尼古拉斯· 马杜罗的行动中如何被应用。国防官员迅速回应称，如果Anthropic不允许五角大楼在合法范围内按其 意愿使用该人工智能，公司将被视为供应链风险，这可能使其及任何分包其人工智能技术的企业失去未 来的业务机会。 美国国防部长皮特·赫格塞思（Pete Hegseth）威胁人工智能公司Anthropic，称如果该公司在周五之前不 同意对军方有利的条款，官员们可能会援引相关权力，以国家安全为由，迫使这家人工智能企业共享其 创新技术。知情人士透露了正在进行的谈判情况。 不过，知情人士表示，如果有关其技术被用于自主武器或大规模监控的担忧得不到解决，Anthropic准 备退出谈判——以及与国防部价值2亿美元的合同。 五角大楼方面则辩称，其提议中并未包含任何不合法使用Anthropic技术的内容。一名国防部高级官员 在声明中表示，如果该公司未能在周五下午5点01分前遵从要求，赫格塞思"将确保对Anthropic启动《国 防生产法》，迫使其为五角大楼服务，无论他们是否愿意。" "这与使用大规模监控或自主武器无关，"这名国防官员 ...

Core Viewpoint - The report titled "2028 Global Intelligence Crisis" by Citrini Research founder James van Geelen predicts a dystopian future driven by AI advancements, leading to massive layoffs, deflation, and a spike in unemployment rates in the U.S. to over 10% [1][5]. Group 1: Market Reaction - Following the report's release, the U.S. stock market experienced a significant sell-off, with the S&P 500 index dropping over 1% and major software stock ETFs declining by more than 4% [3]. - Specific companies mentioned in the report, such as ServiceNow, DoorDash, and American Express, saw their stock prices fall, despite Citrini not shorting these companies [3][4]. - The market's reaction indicates a shift in sentiment regarding AI, moving from enthusiasm for growth to concerns about potential disruptions [3][4]. Group 2: Report Content and Implications - The report outlines a scenario where rapid AI advancements lead to increased productivity but also render many jobs obsolete, resulting in a consumer spending collapse and a "race to the bottom" in white-collar jobs [5][6]. - The report has sparked significant debate, with critics labeling it as "science fiction" and questioning its economic logic [5][6]. - Van Geelen emphasizes that the report aims to initiate dialogue to prevent the described scenarios from occurring, acknowledging the uncertainty surrounding the potential outcomes of AI advancements [6][7]. Group 3: Company Background - Citrini Research, founded by van Geelen, focuses on thematic investment research and has a subscriber base of over 119,000, covering various topics from modern warfare to macro trends [7]. - The company has gained attention in the financial sector, particularly after van Geelen's previous research on shorting Silicon Valley Bank prior to its collapse [6][7].

投资者对此感到不安。标准普尔指数23日下跌超过1%，其软件和服务指数跌至去年4月美国总统特朗普 宣布"解放日"关税以来的最低水平。科技股受到打击，纳斯达克综合指数一度下跌超过1%。 24日，美国白宫经济顾问委员会代理主席皮埃尔·亚雷德表示，Citrini Research近日发布的报告是"一篇 有趣的科幻小说"。 智通财经记者 南博一 美国一份病毒式传播且完全基于猜测的报告，让美国股市23日遭遇一波AI恐慌。 这篇名为《2028年全球智能危机》的报告，由鲜为人知的美国公司Citrini Research发布在在线写作平台 Substack上，虚构了从现在开始到2028年6月结束的情景，称届时美国失业率将突破10%，"占领硅谷"运 动的参与者将在AI企业OpenAI和Anthropic的办公场所外扎营。一系列广泛使用AI代理的事件将冲击软 件公司，并向外扩散，波及到私人信贷和抵押贷款，导致经济陷入恶性循环，出现螺旋式下滑。 据《卫报》报道，Citrini设定的情景如下： AI代理消除所有中介服务 情景始于AI代理实现能力大幅提升。Citrini提到了Anthropic的Claude Code、OpenAI的Co ...

Core Insights - The market's fear regarding AI disrupting traditional industries may have peaked, with "HALO" trades (heavy assets, low obsolescence) being the best hedge for investors still concerned about AI impacts [1][8] Group 1: Market Sentiment and AI Impact - The recent briefing by AI company Anthropic indicated a preference for "cooperation" between AI and existing software providers, contrasting previous fears of complete replacement, leading to a rebound in previously shorted software stocks [1][7] - Despite the S&P 500 index remaining stable since late October last year, extreme fund flows have caused unprecedented internal market divergence, with significant capital inflows into AI beneficiaries and semiconductor sectors while indiscriminately selling software assets [1][4] Group 2: Stock Market Dynamics - The divergence in stock performance has been extreme, with growth and value stocks experiencing a 24% return difference, and the S&P 500's technology and consumer discretionary sectors facing an 11% decline [4] - Defensive and cyclical sectors have seen significant gains, with industrials up 13%, consumer staples up 16%, materials up 22%, and energy up 25% during the same period [4] Group 3: Fund Flows and Positioning - Hedge funds have significantly increased their exposure to semiconductor and AI-related stocks, reaching the highest levels since 2020, while infrastructure software has been the most sold theme, dropping to the 0th percentile since 2020 [7] - Following the Anthropic briefing, there was a reassessment of extreme positions, with software stocks previously labeled as "at risk from AI" rebounding approximately 5% [7] Group 4: HALO Strategy - For investors who believe the "AI fear" has not peaked, Morgan Stanley recommends investing in HALO assets, which are characterized by high production capacity and low obsolescence [8][10] - The HALO basket has risen 28% over the past year, while stocks affected by AI disruption have fallen 43%, indicating a significant attractiveness of this strategy [10]

【文/观察者网专栏作者 关心】 最近，美国人工智能巨头Anthropic公开指控DeepSeek等三家中国AI企业，在科技圈掀起了不小的波 澜。这家美国企业声称，DeepSeek、月之暗面和MiniMax三家中国AI企业对其旗舰模型Claude发动了所 谓"工业级蒸馏攻击"，用1600万次请求、上万个虚假账户，系统性地"抽取"模型能力。消息一出，迅速 占领了各大科技媒体的头条。 Anthropic指责对手从Claude身上"偷"能力，却对自己从整个互联网"拿"知识的历史闭口不谈。他们强调 蒸馏出来的模型缺乏"安全护栏"会带来风险，却没解释凭什么只有他们有权定义什么是"必要的护栏"。 这种双重标准，说到底就是这样的霸权逻辑：规则由最强者制定，强者的行为是惯例，弱者的行为是攻 击。 有意思的是，Anthropic本想通过这份声明打击对手，却无意间成了开源AI最有力的广告。他们用自己 的行为向所有人证明：在闭源AI服务的架构下，你的隐私、你的自主权、你的知情权，都是无法保障 的。当一家公司可以随时以"安全"的名义监控你、评判你、惩罚你，所谓的"信任"就不再是一种美德， 而是一种风险。 Anthropic创始人和CE ...

Group 1: AI Industry Insights - Anthropic launched "Cowork & Plugins for the Enterprise," an AI toolkit that integrates with popular office software like Microsoft Excel, PowerPoint, and Slack, allowing users to perform tasks without switching applications [3] - The tool has demonstrated significant efficiency improvements, such as reducing the time to write a 300-page clinical research report from months to just 10 minutes [3] - The report emphasizes that while AI applications are advancing, they are still in the exploratory phase and not yet at the stage of replacing professional software, indicating a positive outlook for rapid growth in the AI industry [3] Group 2: Tongwei Co., Ltd. Developments - Tongwei Co., Ltd. announced plans to acquire 100% of Qinghai Lihua Qingneng Co., Ltd. through a combination of share issuance and cash payment, which will enhance its position in the polysilicon market [4] - Lihua Qingneng has established a production capacity of 150,000 to 200,000 tons of polysilicon, ranking sixth in the industry with low production costs due to new equipment [4] - The report suggests that the current model of industry consolidation is shifting back to individual company integration, which may lower the difficulty of industry consolidation and accelerate the exit of outdated capacities and companies [4] Group 3: Spotify Performance - Spotify reported a significant increase in monthly active users (MAU), surpassing 750 million with a net addition of 38 million in Q4 2025, while paid users reached 290 million, reflecting a year-on-year growth of 10.3% [6] - The company's gross margin improved to 33.1%, and operating profit margin reached 15.5%, indicating a notable enhancement in profitability driven by subscription structure optimization and cost dilution [6] - Looking ahead to Q1 2026, Spotify maintains an optimistic outlook, expecting continued growth in MAU and paid users, with revenue projected at €4.5 billion, a 15% year-on-year increase excluding currency effects [6]