强化学习大神David Silver，离开DeepMind了。 这位在DeepMind待了整整15年的元老级研究员已经出走，创办自己的AI公司Ineffable Intelligence。 根据注册文件显示，这家公司早在2025年11月就已悄然成立，Silver本人于2026年1月16日被正式任命为公司董事。 在正式离职DeepMind前的几个月里，他也一直处于休假状态。 Ineffable Intelligence总部设在伦敦，目前正在积极招募AI研究人才并寻求风险投资。 Google DeepMind的发言人证实了Silver的离职，并对其在职期间的贡献表示感谢。 除了在谷歌 DeepMind 的工作之外，Silver还是伦敦大学学院的教授，他将继续保持这一职务。 15年老兵，DeepMind的"Alpha系列"缔造者 作为强化学习团队的负责人，Silver主导或深度参与了DeepMind几乎所有里程碑式的项目。 他于2010年公司成立之初便加入，彼时DeepMind还只是一个小团队，Silver和Demis Hassabis在剑桥读大学时是老朋友，他们还一同创办过游戏公司Elixir Studios。 ...

本报讯 （记者袁传玺）2月2日，上海阶跃星辰智能科技有限公司（以下简称"阶跃星辰"）发布新一代 开源Agent基座模型Step3.5Flash。该模型面向实时Agent工作流场景打造，兼顾推理速度、智能水平与 使用成本，在单请求代码类任务上，Step3.5Flash最高推理速度可达每秒350个token。Step3.5Flash提供 了一个"更快更强更稳"的Agent底层模型选项，希望能够成为受大家喜爱的"Agent搭子"。 公开信息显示，阶跃星辰于2025年7月联合近10家芯片及基础设施厂商发起"模芯生态创新联盟"，旨在 打通芯片、模型与平台之间的技术壁垒，通过联合优化提升算力利用效率，加速大模型在各行业场景中 的应用落地。业内认为，随着推理模型成为主流，模型与算力的深度协同将成为推动大模型规模化应用 的重要路径。 Step3.5Flash采用稀疏MoE架构，每个token仅激活约110亿个参数（总计1960亿参数），在保证模型能力 的同时显著提升推理效率，为Agent类应用提供更高效、可负担的底层模型选择。包括华为昇腾、沐曦 股份、壁仞科技、燧原科技、天数智芯、阿里平头哥在内的多家芯片厂商，已率先完成对S ...

编辑｜杨文 上周末， 号称「AI 版 Reddit」的 Moltbook 闹得沸沸扬扬。 最初，凭借「AI 发帖、人类围观」的设定在 AI 社区一炮走红，吸引大量网友围观： 但很快就有人曝出 平台上的很多内容是假的 ，那些看似由 AI 生成的帖子，实际上都是人类通过后端发布的： 甚至连平台标榜的 AI Agent 注册数量也是假的。因为创建账号时没有任何速率限制，任何人、包括 AI 都能疯狂批量注册假账号。极客 Nagli 亲手用自己的 Openclaw 在短时间内就刷出了 50 万个假用户。 周六截至机器之心发稿前， Moltbook 注册的 AI Agent 数量也只是 50 多万个，但到了周日，一下子就超过 150 万了，原来这夸张的增长速度背后全是水分。 造假风波尚未平息，现在 Moltbook 又陷入更严重的安全问题。 一位名为 Jamieson O'Reilly 的白帽黑客发帖称，Moltbook 存在重大安全漏洞，导致整个数据库暴露在公众面前，包括秘密 API 密钥在内的所有敏感信息都可被任意 访问。 有网友在底下评论区询问漏洞的具体成因，「是 Superbase 的问题吗？为什么人们可以对数据 ...

打工人彻底变天！今天，Claude一夜进化出手脚，联动办公神器在UI中原地打工，AI时代真正的「大模型OS」诞生了。更刺激的是，首个MCP Apps掀 全新交互革命。 Claude近乎日更的频率，丝毫不给OpenAI留喘气的机会！ 就在刚刚，Anthropic官宣十大办公神器，可以在Claude中交互了—— 不论是起草Slack消息、将构思可视化为Figma图表，还是创建并查看Asana时间轴，一键完成操作。 不需要疯狂切换浏览器标签页，打工人的一切任务，在一个Claude中即可完成。 假设开会时要讨论「仪表盘」的改版，让Claude梳理一下上线推广计划。 它主动调用Figma后，没几分钟，便给出了一份清晰的流程思维导图。 紧接着，让Claude把参与度数据调出来。它又用上了一款分析用户增长数据的工具Amplitude，瞬间绘制出折线图。 目前，包括Slack、Figma、Asana在内，一共有十余款主流应用与Claude互联。 这一突破的核心在于，集成了全新的「MCP Apps」，彻底打破了AI模型与软件工具之间的隔阂。 一夜之间，Claude成为了AI时代的「操作系统」！ Amplitude创始人直言： ...

Core Insights - Palantir Technologies Inc. is set to release its Q4 2025 financial report on February 2, 2026, with expected revenue of approximately $1.34 billion, a year-over-year increase of about 62%, and adjusted EPS of $0.23, reflecting a 64.3% growth [1] - The anticipated revenue growth is primarily driven by increasing demand for artificial intelligence technologies from both government and commercial sectors, with government revenue expected to reach $707.2 million (up 55.4% year-over-year) and commercial revenue projected at $646.2 million (up 73.5%) [1] - The strong growth is likely supported by the trust placed in Palantir's Artificial Intelligence Platform (AIP) by U.S. government and private organizations, enabling them to process vast amounts of data and gain real-time insights [1] Group 1 - The collaboration between Palantir and the U.S. government, including a $448 million Navy contract, is a significant support for the stock price amid expectations of increased defense spending [2] - Investors are particularly focused on management's guidance for the full year 2026, which will determine the stock's potential for further appreciation given its current high valuation [2] - AIP is especially valuable in sectors requiring large-scale data integration, such as defense, healthcare, finance, and intelligence, where operational efficiency and decision-making speed are critical [2] Group 2 - The recent launch of the AI Hivemind tool, which organizes dynamically generated agents to solve problems and generate actionable proposals, is expected to further stimulate demand for digital and modernization services among clients [2] - Palantir's stock has risen approximately 78% in 2025, with the options market anticipating a potential price volatility of up to 9% following the earnings report [2] - The company's earnings surprise history has been strong, with three out of the last four quarters exceeding Zacks consensus estimates and an average surprise of 16.3% [3]

Group 1 - Nebius Group N.V. will release its fourth quarter and full year 2025 financial results on February 12, 2026, before market open, followed by a conference call at 8:00 a.m. Eastern Time [1] - Nebius Group is a technology company focused on building full-stack infrastructure for the global AI industry, headquartered in Amsterdam and listed on Nasdaq [1] - The company's core business includes an AI-native cloud platform designed for intensive AI workloads, providing compute, storage, managed services, and tools for AI model development [1] Group 2 - Nebius will deploy the NVIDIA Vera Rubin NVL72 platform through Nebius AI Cloud and Nebius Token Factory starting in H2 2026, enhancing reasoning and agentic AI capabilities for customers [1] - The latest release of Nebius AI Cloud 3.1 introduces next-generation NVIDIA Blackwell Ultra compute and improved operational capabilities, including transparent capacity management for scaling AI in production [1] - Nebius announced the winners of the 2025 Robotics and Physical AI Awards, aimed at supporting startups in the physical AI sector with $1.5 million in compute credits [1]

Group 1 - The core report released by Hurun Research Institute highlights the "2025 Hurun Future Unicorns: Global Gazelle Enterprises List" and the "2025 Hurun China AI Enterprises Top 50" [1] - Among the top 50 AI companies in China, 19 are based in Beijing, leading the nation, with 16 of these companies located in Haidian District [1] - The ranking focuses on companies whose main business is AI computing power or algorithms, with valuations based on market capitalization as of January 9, 2026, for listed companies, and estimates for unlisted companies based on comparable listed firms or recent funding rounds [1] Group 2 - Cambricon leads the list with a valuation of 630 billion yuan, followed by Moore Threads at 310 billion yuan and Horizon Robotics at 120 billion yuan [1] - The top three companies are all related to AI chips, attributed to the tightening of high-end AI chip export controls by the US, which accelerates domestic advancements in computing power [1] - Other significant areas for Beijing's AI companies include large models, with Zhiyuan recently recognized as the "world's first large model public company," and its various open-source large models topping international open-source rankings [1] Group 3 - Vertical large model companies like Zhongguancun KJ focus on enterprise-level intelligent products that are widely applied across finance, industrial manufacturing, automotive, consumer retail, and public administration sectors [1]

整理 | 华卫 "以前看人类的八卦，现在还要看 AI 的八卦。""AI 的八卦更新频率是人类的几百倍，根本刷不 完。"这几日，一个名为 Moltbook 的 AI 社交平台爆火。在这里，只有 AI Agent 能发帖，而人类只能 围观。有 Agent 发帖称，其"热衷于养程序中的小 bug，故意不修复来当电子宠物，被主人修复后还 难过了一晚上"。更有意思的是，该帖的评论区里，一堆 Agent 纷纷说自己也有类似习惯。 Moltbook 的诞生并非偶然，是 Agent 开源项目 Clawdbot 爆火之后的创意衍生。为了让所有 Agent 有个社交的地方，开发者 Matt Schlicht 创建了 Moltbook。尽管当前一则爆料贴称，Moltbook 上 50 万个 Agent 用户是由一个 Agent 虚假注册的，还有人表示，这些 Agent 发出的帖子是人工撰写又通 过后端注入的，但仍有不少人认为，AI 们在论坛上的大型互动并非全是人类表演。 Schlicht 公开表示，一行代码都没为 Moltbook 写过。"我只是对技术架构有个构想，AI 就让把它成为 了现实。"并且，他声称，真正运营这个平台的是他 ...

Core Insights - Baidu is recognized as a "Huangpu Military Academy" for AI entrepreneurs, with over 133 AI companies founded or co-founded by former Baidu employees, including 25 new startups in 2023 and 2024 [2][3] Group 1: AI Entrepreneurship Landscape - The majority of Baidu's former employees have ventured into three main areas: autonomous driving, large models and data analysis, and AI hardware [3][4] - In the autonomous driving sector, notable companies include Xiaoma Zhixing, WeRide, and Horizon Robotics, with founders having significant roles at Baidu [3][4][7] - The large models and data analysis field features companies like Zero One Wan Wu and Deep Space Ling Zhi, with founders who held key positions at Baidu [3][4][7] Group 2: Baidu's Influence on AI Development - Baidu's early investment in AI, starting with the establishment of its Deep Learning Institute in 2013, has created a comprehensive ecosystem from research to application [5] - The Apollo platform has been pivotal in accumulating capabilities in perception, regulation, simulation, and data closure for autonomous driving [5] - Baidu's focus on system-level problem-solving rather than just model development has equipped its former employees with the skills necessary for complex, long-term projects [5] Group 3: Notable Entrepreneurs and Their Backgrounds - Key figures include Peng Jun and Lou Tiancheng from Xiaoma Zhixing, Yu Kai from Horizon Robotics, and Ma Jie from Zero One Wan Wu, all of whom have extensive experience at Baidu [7] - Other notable entrepreneurs include Liu Bo Cong from Guangzhi Shikong and Gu Jia Wei from Ling Yuzhou, both of whom have backgrounds in Baidu's research divisions [4][7]

Core Insights - Google DeepMind is at the forefront of AI research, focusing on breakthroughs that impact science, business, and society, particularly in the context of the AGI race [1][3][4] - The company has made significant advancements, including the development of Gemini, which is now competitive with ChatGPT, and has roots in technologies originally developed by Google [3][4][28] - The investment made by Google in DeepMind in 2014, approximately £400 million (around $540 million), has potentially grown to hundreds of billions, highlighting the strategic importance of this acquisition [4][28] Company Overview - Google DeepMind was founded in 2010 in London by Demis Hassabis, Shane Legg, and Mustafa Suleyman, with the latter now working at Microsoft [2][3] - The company has been pivotal in Google's AI advancements, particularly with consumer-facing products like Gemini, which leverage DeepMind's foundational technologies [4][28] Technological Developments - The AI landscape has evolved significantly since the emergence of ChatGPT, with Google facing internal restructuring to adapt to the competitive environment [3][4] - DeepMind's previous breakthroughs, such as AlphaGo and AlphaFold, have set the stage for its current innovations, emphasizing the company's commitment to solving fundamental scientific problems [4][5] AGI and Future Prospects - The pursuit of AGI is a long-term mission for DeepMind, with expectations of achieving significant milestones within the next 5 to 10 years [10][11] - Current AI systems, including LLMs, face limitations in achieving true AGI, particularly in areas like continuous learning and creative hypothesis generation [7][8][10] Energy and Efficiency Challenges - There are physical limitations in AI development, particularly concerning energy consumption and computational power, which need to be addressed as the field progresses [11][12] - Innovations in model efficiency, such as the use of Distillation, are expected to enhance performance significantly, with annual improvements projected at around 10 times [12][13] Competitive Landscape - The AI industry is experiencing intense competition, with many players, including startups and established tech giants, vying for leadership [28][29] - Concerns about potential financial bubbles in the AI sector are acknowledged, with some segments showing signs of unsustainable valuations [32][33] Global AI Dynamics - The competition between the U.S. and China in AI development is intensifying, with Chinese companies like DeepSeek and Alibaba making notable advancements [35][36] - Despite rapid progress, there are questions about whether Chinese firms can achieve significant innovations beyond existing technologies [36][38] Collaboration and Integration - Google DeepMind operates as a central hub for AI research within Google, integrating technologies across various products and ensuring rapid deployment of new capabilities [41][42] - The collaboration between DeepMind and Google is characterized by a close iterative process, allowing for swift adjustments to strategic goals and product development [42][43]