行业热点 腾讯、百度、阿里争发数亿红包，角逐国民级 AI 应用 整理 | 傅宇琪、褚杏娟 导语：腾讯、百度、阿里争发数亿红包，角逐国民级 AI 应用；机器人扎堆抢上春晚，出场要花 1 亿；DeepSeek 正招兵买马，布局 AI 搜索与智能体领域；95 后清华博士加盟腾讯混元；英伟达 CEO 黄仁勋否认对 OpenAI 不满，计划进行巨额投资；Clawdbot 更名 OpenClaw，15 万个 Agent 自主发 帖、协作、吐槽人类；字节禁止员工利用公司资源做号谋利；贵州茅台出资参与 SpaceX 上市 A 轮融 资？不实；阿里明确云 +AI+ 芯片战略，PPU 芯片出货已数十万片…… 2026 年春节期间，字节、阿里、腾讯、百度等大厂围绕 AI 超级入口（Agent 时代）展开激烈争夺 战，以现金红包为核心抓手，结合产品迭代、生态布局、投流推广等策略抢占用户注意力，角逐首款 国民级 AI 应用。 1 月 25 日，腾讯官方发布关于春节分 10 亿现金的通知：将在 2 月 1 日上线春节活动，用户上元宝 App 分 10 亿现金红包，单个红包金额可达万元。马化腾表示希望此次活动能够再次迎来微信红包的 盛况。 ...

作者 | Tina Linus Torvalds 常开玩笑说自己会"活到永远"。但以防万一，Linux 内核社区现在也准备好了一套交接方案——只是这份方案并没有点名具体的接班 人。 如果 Torvalds 发生意外，或者哪天决定退休，Linux 不再把一切寄托在"到时候再说"。核心内核社区已经正式起草了一份项目连续性计划：一旦顶层维 护者出现空缺，应该如何在最坏情况或有序过渡中，选出新的顶层维护者（可能是一人，也可能是多人），确保项目长期稳定。 Torvalds 本人则明确表示自己暂无退休打算。被问到未来是否会交棒时，他依旧以一贯的幽默回应，暗示自己更倾向于"继续干下去"。随后他又补充了 一个更现实的理由：家里人同样不希望他突然闲下来，尤其是太太，大概更不想每天被一个无所事事、没事找事的丈夫缠着。 因此也有人提出，与其再找一位新的"终身仁慈独裁者"（BDFL），不如把顶层维护者的职责拆分给多位值得信赖的开发者共同承担。 | ... | @@ -0,0 +1,41 @(a | | --- | --- | | 1 | .. SPDX-License-Identifier: GPL-2.0 | | 2 | + | ...

编译 | Tina 过去几十年，软件工程有一个稳定不变的前提：系统的行为写在代码里。工程师读代码，就能推断系 统在大多数场景下会怎么运行；测试、调试、上线，也都围绕"确定性"展开。但 Agent 的出现正在动 摇这个前提：在 Agent 应用里，决定行为的不再只是代码，还有模型本身——一个在代码之外运 行、带着非确定性的黑箱。你无法只靠读代码理解它，只能让它跑起来、看它在真实输入下做了什 么，才知道系统"到底在干什么"。 在播客中，LangChain 创始人 Harrison Chase 还把最近一波"能连续跑起来"的编程 Agent、Deep Research 等现象视为拐点，并判断这类"长任务 Agent"的落地会在 2025 年末到 2026 年进一步加 速。 这也把问题推到了台前：2026 被很多人视为"长任务 Agent 元年"，现有的软件公司还能不能熬过 去？就像当年从 on-prem 走向云，并不是所有软件公司都成功转型一样，工程范式一旦变化，就会 重新筛选参与者。长任务 Agent 更像"数字员工"——它不是多回合聊天那么简单，而是能在更长时间 里持续执行、反复试错、不断自我修正。 在这期与红 ...

Core Insights - The article discusses the transition from passive large models to proactive agents in 2025, marking a significant shift in AI capabilities and applications [1] - It emphasizes the importance of standardized protocols like MCP and A2A in facilitating the integration and collaboration of AI agents across different platforms and systems [2][4] Group 1: Protocols Driving Agent Applications - The MCP (Model Context Protocol) was introduced by Anthropic to standardize how AI models access external tools and services, akin to a "USB-C interface" for AI agents [2] - The A2A (Agent-to-Agent) protocol by Google aims to establish a common language for collaboration among agents from different backgrounds, enabling them to communicate and coordinate tasks effectively [4][5] - Both protocols reduce integration costs, enhance reliability, and accelerate automation capabilities by providing a unified interaction framework [3][5] Group 2: Engineering Challenges in Agent Collaboration - Despite the growth in applications, challenges such as inefficiency and miscommunication among agents arise in enterprise environments [6][7] - The need for quantifying agent collaboration and identifying effective communication paths is highlighted as a significant hurdle for developers [7] - Current agents lack the self-regulation seen in traditional business process management (BPM) systems, necessitating a clear definition of their roles and boundaries within existing workflows [7][8] Group 3: Real-World Applications and Value Creation - The most successful applications of agents are found in programming and operations, with significant efficiency improvements reported [8] - Agents are evolving to mimic engineer experiences in automated operations, enhancing their ability to troubleshoot and respond to system errors [8] - The article suggests that agents will increasingly integrate into business processes, acting as "digital employees" rather than fully autonomous entities [9][10] Group 4: Future Perspectives on Agent Evolution - Experts express differing views on the ultimate form of agents, with one suggesting they will become highly autonomous entities, while another sees them as collaborative digital employees [9][10] - There is a consensus that agents will transition from niche applications to becoming foundational infrastructure in various business contexts [10][11]

模力工场 新鲜事 想用一个下午快速摸清一个领域，并产出一份条理清晰、信息量丰富的深度内容？本周模力工场带你体验 "AI 增效 流水线：从信息到作品的智能生产工作流"。从智能阅读提炼（语鲸）、一键生成研报（AI 快研侠），到跨平台记 忆管理（MemOS-MindDock），再到自动视觉设计，这条流水线覆盖"读、写、研、记、设计"全流程，助你将碎 片信息快速整合为结构化的知识作品。例如，若你对近期热议的 Clawdbot 等 AI 助手产品感兴趣，不妨以此为主 题，用这套工作流实践一番。点击进入模力工场首页，查看顶部专题横幅，扫码添加模力小 A，获取完整工具链 与实操指引。 030 周上榜应用精选（附用户热评） 模力工场 第 030 周 AI 应用榜来袭！本周共有 32 款应用上架，榜单完全由用户真实使用、测评与讨论热度驱动。我们 从社区声量最高的应用中精选出十款，并透过用户真实评论，为你解读榜单背后的产品逻辑与行业风向。 创作平民化：人人都能成为内容创作者 随变: 潮人必备 AI 创作神器，让灵感瞬间变潮流短片！ "玩了几天随变，感觉有点像简洁版抖音…但 AI 创作出来的视频，如'创作一条刀马刀马的舞蹈片段'它 ...

Core Viewpoint - Google has launched "Project Genie," a groundbreaking world model prototype that allows users to create interactive virtual worlds with just a sentence or an image, marking a significant advancement in the field of artificial general intelligence (AGI) [2][12]. Group 1: Project Genie Overview - Project Genie is built on the latest world model, Genie 3, and utilizes a self-regressive generation mechanism to create environments based on user descriptions and actions, rather than pre-recorded content [10][11]. - The quality of the generated virtual worlds is significantly higher than previous research demos, approaching that of mature gaming products, with a resolution of approximately 720p and a frame rate of 20-24 frames per second [7][16]. - The application potential of world models is vast, including areas such as autonomous driving simulations, environmental understanding for embodied intelligence, game development, film production, and interactive education [13][14]. Group 2: User Interaction and Experience - Users can select from predefined templates or fully customize their environments and characters, allowing for a unique virtual world creation experience [20][23]. - The system allows for real-time interaction, with a maximum exploration time of 60 seconds per generated world, and can remember key changes made by users for up to one minute [17][19]. - Despite its innovative features, early user experiences have highlighted limitations, such as low-quality generated worlds, simple structures, and occasional input delays affecting the overall experience [15][32]. Group 3: Future Implications and Concerns - The launch of Project Genie has sparked discussions about its potential impact on the gaming industry, with concerns that it may lead to job losses among game developers [30]. - Critics have pointed out that the generated worlds can lack depth and complexity, with limited interactive elements and occasional inconsistencies in the virtual environment [32][34]. - Google emphasizes that Genie is not a game engine but rather a tool for enhancing creativity and accelerating prototyping, with ongoing improvements expected as user feedback is collected [35][40]. Group 4: Development and Collaboration - The development of Project Genie involved extensive collaboration across various Google teams, highlighting the company's ability to integrate advanced technologies into user-friendly applications [48][51]. - The team acknowledges that while the current model has limitations, it represents a significant step towards creating interactive and immersive virtual experiences [41][46]. - Future iterations of the model aim to expand its capabilities and applications, particularly in entertainment and education, with a focus on personalized learning experiences [55][57].

Core Viewpoint - OpenAI has launched Prism, a new AI research tool powered by GPT-5.2, aimed at enhancing scientific research collaboration and efficiency, now available for free to all ChatGPT personal account users [2][3]. Group 1: OpenAI's Strategic Move - OpenAI's entry into the scientific research field is seen as a response to the growing importance of AI in academia, with the goal of empowering scientists to conduct advanced research by 2030 [2][3]. - The establishment of the OpenAI for Science team indicates a focused effort to explore how large language models (LLMs) can assist researchers and optimize tools for scientific support [2][3]. Group 2: Model Capabilities and Limitations - Kevin Weil, OpenAI's VP, acknowledges that while current models can accelerate research by preventing time wastage on solved problems, they are not yet capable of making groundbreaking discoveries [4][5]. - The latest version, GPT-5.2, has shown significant improvement, achieving a 92% accuracy rate in the GPQA benchmark, surpassing the performance of 90% of graduate students [7][8]. Group 3: Research Applications and Feedback - Researchers have reported that GPT-5 can assist in brainstorming, summarizing papers, and planning experiments, significantly reducing the time needed for data analysis [13][14]. - Feedback from various scientists indicates that while GPT-5 can provide valuable insights, it still makes basic errors, and its role is more about integrating existing knowledge rather than generating entirely new ideas [14][15]. Group 4: Future Directions and Enhancements - OpenAI is working on two main optimizations for GPT-5: reducing confidence in its answers to promote humility and enabling the model to fact-check its outputs [4][19]. - The goal is to create a collaborative workflow where the model can serve as its own verifier, enhancing the reliability of its contributions to scientific research [19][20].

作者 | 姚戈 世界模型领域迎来了一个重要开源模型。 今天，蚂蚁集团旗下的具身智能公司"蚂蚁灵波"，正式发布并开源其通用世界模型 LingBot-World。 与许多闭源方案不同，蚂蚁灵波选择 全面开源代码和模型权重，而且不绑定任何特定硬件或平台 。 去年 DeepMind 发布的 Genie 3，让人们看到了世界模型能够根据文本或图像提示，实时生成一个可 探索的动态虚拟世界。LingBot-World 沿袭了这条路线，并在交互能力、高动态稳定性、长时序连贯 性以及物理一致性等维度取得了突破。 更令人惊喜的是， LingBot-World 呈现出从"生成"到"模拟"的跨越 。随着模型规模的扩大，灵波团 队观察到，LingBot-World 开始表现出远超普通视频生成的复杂行为，涌现出对空间关系、时间连续 性和物理规律的理解。 可以看到，鸭子腿部蹬水的动作、水面对扰动的响应、以及鸭子身体与水之间的相互作用都比较符合 物理规律。 这显示出模型不仅记住了视觉表象，还在某种程度上理解了流体力学等基础物理机制。同时，水面对 扰动的反应，显示出模型对因果关系的理解。 用户切换视角后再回来时，环境中的智能体（比如这只猫）仍 ...

整理 | 褚杏娟 Clawdbot（现名：Moltbot）火了到国内，社交平台上到处都是部署教学、使用教学和使用展示。国 内的腾讯云、阿里云等也相继宣布上线 Clawdbot 云端极简部署及全套云服务，钉钉也在 Github 上 开源了 Moltbot 接入方式。 项目背后的创始人 Peter Steinberger 也红极一时，他的构建方式成为很多人的学习对象。Peter 之 前就是一位非常出色的开发者，打造了一个被用在超过十亿台设备上的 PDF 框架。后来他经历了严 重的职业倦怠，卖掉股份，整整三年从科技圈消失。今年，他回来了，而他现在的构建方式、正在做 的事情，已经和传统软件开发完全不同。 Peter 近期在"The Pragmatic Engineer"节目中，用近两个小时的时间分享了他的开发经历。他解释 了，为什么他现在发布的代码，大部分自己都不再逐行阅读，而这其实并没什么大不了；他具体是如 何打造了 ClawdBot 这个看起来就像 Siri 未来版本的个人助手的；他如何利用"闭环原则"，高效进行 AI 编程；为什么代码评审已经过时，PR 应该改名叫 Prompt Request 等，他还分享了很 ...

作者 | 冬梅 全球半导体光刻机巨头 ASML（阿斯麦） 今日正式宣布，由于全球成熟制程设备需求放缓以及出口 监管政策的持续收紧，公司计划在全球范围内削减约 1700 个工作岗位。 这是 ASML 自 2023 年 AI 浪潮爆发以来首次进行大规模裁员，标志着即便是在 AI 高速发展的背景 下，半导体设备行业也未能完全抵御周期性调整的冲击。 ASML CEO Christophe Fouquet 发布的全员信如下： 尊敬的 ASML 同事们： 今天，我们发布了 2025 年全年财务业绩以及对未来一年的展望。半导体生态系统有望在未来几年迎 来显著增长，而 ASML 已做好充分准备，把握这一积极发展机遇。我谨代表管理委员会，感谢各位 同事为取得这一成功所做出的贡献。 我们取得的成功归功于我们对客户的专注、卓越的工程技术以及与生态系统的协作。我们的创新和执 行能力为客户、供应商、同事和投资者带来了显著的效益。我们将根据客户需求，继续扩大员工队伍 和业务规模，包括计划在埃因霍温建设的第二个园区。 在技术部门，我们计划将项目 / 矩阵式组织架构转变为以特定产品和模块为核心的模式。这将有助于 简化流程和决策。我们从公司各 ...