OpenAI 模型接连"泄露"？Altman：未来几个月将发布大量东西；Anthropic 对 OpenAI 封锁 API；曝特斯拉拖欠 1.1 亿美元，逼垮小企业；宇树科技 CEO 王兴兴获聘杭州具身智能应用中试基地首席科学家；字节跳动辟谣人均在职 7 个月，并称高绩效员工离职率约 5%；海康威视披露海康机器人 IPO 进程：目 前还在深交所创业板排队；英伟达回应被约谈；裁员 9000 人后：微软第四财季净利润暴增 24%；10 亿天价挖角惨遭 12 人连环拒，扎克伯格钞能力或失 灵；苹果 1 个月内 4 位 AI 研究员跳槽，库克安抚员工…… 行业热点 OpenAI 模型接连"泄露"？Altman：未来几个月将发布大量东西 整理 | 傅宇琪、褚杏娟 8 月 3 日，Sam Altman 发布 x 表示，OpenAI 未来几个月将发布大量新东西，包括：新模型、新产品和新性能等。 最近，OpenAI 不断有新模型"泄露"。 7 月 19 日，工程师 Tibor Blaho 在 X 上分享了一个截图："gpt-5-reasoning-alpha-2025-07-13"。OpenAI 研究员 Alexander ...

采访嘉宾｜蒋思源，硅心科技（aiXcoder) 算法专家；黄宁，硅心科技（aiXcoder) 产品专家 作者 | 冬梅 当 GitHub Copilot、CodeRabbit 等代码审查在 2024 年掀起行业讨论热潮时，距离 Meta 工程师 Marius Eriksen2021 年预言 AI 代码助手的潜力 不过三年。 如今，据不完全统计，市面辅助编码相关工具已超过 20 种，涵盖通用代码审查、安全漏洞检测等多个细分领域。这些工具已形成割据之势，宣称 能将代码审查效率最高提升 300%。 然而，这些工具的实际表现却引发诸多争议。有工程团队反馈，部分 AI 代码审查工具与传统的静态代码分析工具（linters）功能高度重叠，只是 披上了 AI 的外衣。Echios 首席技术官 Jon Freedman 表示，在小规模团队中， 部分工具每月 30 美元的成本物有所值，但在复杂项目中，其作用 有限 。 这种局限性在处理复杂业务逻辑或跨模块交互的代码时尤为明显。一些 AI 编码工具虽标榜拥有"完整代码库上下文"，但在实际应用中，仍难以穿透 项目特有的业务规则和架构设计。 误报问题也让开发者头疼。Reddit 上有 ...

编译 | Tina Meta 首席执行官马克·扎克伯格（Mark Zuckerberg）在周三分享了他对"个人超级智能"（personal superintelligence）的愿景——也就是每个人都能够 借助 AI 实现自己的个人目标。 不过，在这封信中也悄然透露出一个信号：Meta 正在调整其 AI 模型的发布策略，以更好地推动"超级智能"的发展。 扎克伯格写道："我们相信，超级智能带来的好处应尽可能广泛地惠及全球。但与此同时，它也将带来前所未有的安全挑战。我们必须严谨地管理这些风 险，并慎重考虑哪些内容适合开源。" 这段关于开源的表述耐人寻味。一直以来，扎克伯格都将 Meta 的 Llama 开源模型系列视为公司区别于 OpenAI、xAI 和 Google DeepMind 等竞争对手 的关键优势。Meta 的目标是构建性能媲美甚至优于闭源模型的开源 AI 模型。 不过他此前也曾留下余地。"我们当然非常支持开源，但我并没有承诺会发布我们做的每一项成果。"他在去年的一档播客节目中这样说，"如果某个阶段 模型的能力发生了质的变化，而我们觉得开源它是不负责任的，那我们就不会开源。这一切都很难预测。" 扎克伯 ...

编译 | 平川、Tina 近期，AI 编程领域又一匹 AI Coding 黑马正在快速崛起。在一次 对主流 AI 编程产品的评级分类里 ，唯一与 Claude Code 并列 S 级的，是 Sourcegraph 最新推出的 AmpCode，而爆火的 Cursor 也只排在了第二档次的 A 级。 那么，AmpCode 究竟有何独特之处？Sourcegraph 工程师 Thorsten Ball 在近期一档播客中分享了这款产品背后的理念与 AI 编程的范式转变。 Thorsten 透露，AmpCode 的研发实际上早于 Claude Code 的发布。两者虽独立发展，但在智能编程助手的核心设计理念上却不谋而合。在他看 来，AmpCode 和 Claude Code 目前代表了最具"代理性"（agentic）的 AI 编程产品：它们不仅能调用工具，还真正"参与"开发流程，具备高度自 治能力。 而与 Cursor、Windsurf 等交互过程不够直接的产品不同，AmpCode 在架构设计上做出了关键决策： "我们选择了放权——把对话记录、工具访问权限、文件系统权限全都交给模型，然后放手让它去做。" Thorsten ...

作者 | 华卫 当大模型的聚光灯照向实体经济，一个"必答题"浮出水面：数字世界里惊艳的大模型技术，怎样才能变成现实世界中实实在在的生产力？ 在 2025 世界人工智能大会（WAIC 2025）期间，蘑菇车联（MOGOX）发布首个物理世界 AI 大模型——MogoMind。在蘑菇车联展区，MogoMind 作 为首个深度理解物理世界 AI 大模型，成为本届大会最受关注的人工智能技术应用之一。 通过深度整合实时、海量的多模态交通数据，MogoMind 能够从物理世界的复杂数据中抽取意义、从经验中学习规则、在不同场景中灵活决策，形成对 交通环境的全局感知、深度认知和实时推理决策能力，可以为多类型智能体提供实时数字孪生与深度理解服务，成为城市和交通高效运行的"AI 数字基 座"。 依托 MogoMind 大模型能力，蘑菇车联推出多款 L4 级前装量产自动驾驶车辆，包括 RoboBus、RoboSweeper 和 RoboTaxi，深度融入全局感知、深 度认知和实时推理决策能，推动自动驾驶技术在公共交通、城市环卫、无人零售等多场景应用。 其中，自动驾驶巴士 MOGOBUS 搭载端到端"MogoAutoPilot+Mog ...

整理 | 华卫 刚刚，中国人工智能初创公司 Manus 推出了一项名为"Wide Research"的新功能。在发布视频中，Manus 联合创始人兼首席科学家季逸峰（Yichao "Peak" Ji）亲自介绍并进行了使用演示，就像他当初首次推出 Manus 时那样。 据悉，这一突破性功能可让用户借助多个 AI Agent 来同时处理大规模任务，单次甚至可调用 100 多个 Agent，所有 Agent 都专注于完成某一单一任务 （或一系列有助于实现上述总体目标的子任务），旨在对人工智能行业中传统的深度研究方法构成挑战。 最初，Wide Research 将向 Manus Pro 计划用户开放，定价为每月 199 美元。之后，该公司计划逐步向 Plus 和 Basic 计划的订阅用户开放这一功能。 走不同于"Deep Research"的路径 OpenAI、谷歌、xAI 等多家主要竞争对手推出的"深度研究"（Deep Research）或"深度研究员"（Deep Researcher）人工智能 Agent，能进行长达数分 钟乃至数小时的广泛、深入的网络研究，并为用户撰写引用规范、内容详尽的报告。 与之不同的是， ...

Core Viewpoint - The article discusses the rapid development of AI in China, highlighting the importance of global cooperation in AI governance and the potential risks associated with technology misuse [1][3]. Group 1: AI Development in China - In the past two years, China's AI technologies, particularly large models like DeepSeek, Mini Max, and Kimi, have achieved remarkable global recognition [3][5]. - Chinese AI models are characterized by their open-weight approach, contrasting with the closed strategies of many leading models in the U.S. [5]. Group 2: Global Cooperation and Governance - Eric Schmidt emphasizes the necessity of open dialogue between China and the U.S. to navigate the challenges posed by AI and to foster a responsible and sustainable future [3][8]. - The establishment of a continuous dialogue mechanism is crucial for both sides to define issues clearly and seek collaborative solutions [8][10]. Group 3: Risks and Ethical Considerations - There are concerns regarding the potential misuse of AI technologies, including issues of deception and harmful behaviors that AI systems might learn [11]. - The need for a balance between open-source technology and regulatory measures is highlighted, as open-source can lead to rapid dissemination of technology, which may pose risks [10][11]. Group 4: Future Outlook - The next two years are expected to witness the emergence of intelligent agents that can perform tasks and interact within various workflows, significantly impacting businesses and governance [14][15]. - There is optimism about the potential for AI to bring about profound societal changes, provided that key concerns are addressed through dialogue and cooperation [15].

Core Viewpoint - The article highlights the significant achievements of Chinese authors in the field of computational linguistics, particularly focusing on the award-winning paper from DeepSeek that introduces a novel sparse attention mechanism for long-context modeling, showcasing its efficiency and performance improvements over traditional methods [1][17]. Group 1: Award and Recognition - The ACL announced that over 51% of the award-winning papers for 2025 had Chinese authors, with the USA at 14% [1]. - A paper by DeepSeek, led by author Liang Wenfeng, won the Best Paper award, which has generated considerable discussion [1]. Group 2: Technical Innovations - The paper introduces a Natively Trainable Sparse Attention (NSA) mechanism, which combines algorithmic innovation with hardware optimization for efficient long-context modeling [4][6]. - NSA employs a dynamic hierarchical sparse strategy that balances global context awareness with local precision through token compression and selection [11]. Group 3: Performance Evaluation - NSA demonstrated superior performance in various benchmarks, outperforming traditional full attention models in 7 out of 9 metrics, particularly in long-context tasks [8][10]. - In a "needle in a haystack" test with 64k context, NSA achieved perfect retrieval accuracy and significant speed improvements in decoding and training processes [9][15]. Group 4: Future Implications - The upcoming DeepSeek model is expected to incorporate NSA technology, generating anticipation for its release [17]. - There are speculations regarding the delay of DeepSeek R2's release, attributed to the founder's dissatisfaction with its current performance [17].

Core Insights - The core argument presented by Jared Kaplan emphasizes the significance of Scaling Law in the development of AI models, suggesting that the majority of AI's value comes from the most powerful models, and that the current rapid evolution of AI is unbalanced, focusing more on capabilities than costs [1][6][50]. Group 1: Scaling Law and AI Development - Scaling Law is derived from fundamental questions about the importance of data size and model scale, revealing a consistent trend where increasing the scale of pre-training leads to improved model performance [10][13]. - Both pre-training and reinforcement learning phases exhibit clear Scaling Laws, indicating that as computational resources increase, model performance continues to enhance [14][17]. - The ability of AI models to handle longer tasks is increasing, with research indicating that the time span of tasks AI can autonomously complete doubles approximately every seven months [20][23]. Group 2: Future Implications and Recommendations - The future of AI may involve models capable of completing complex tasks that currently require extensive human effort, potentially revolutionizing fields like theoretical physics [25]. - Companies are encouraged to build products that are not yet fully operational, as rapid advancements in AI capabilities may soon enable these products to function effectively [29]. - Integrating AI into existing workflows and identifying new areas for large-scale application are crucial for maximizing the potential of AI technologies [30][31]. Group 3: Claude 4 and Its Enhancements - Claude 4 has improved its performance in programming tasks and has enhanced its memory capabilities, allowing it to retain information over longer interactions [34][35]. - The model's ability to understand nuanced supervision signals has been refined, making it more responsive to user instructions and improving the quality of its outputs [34][36]. Group 4: Challenges and Considerations - The current rapid advancement of AI presents challenges, as the focus on capability may overshadow the need for cost efficiency and balance in AI development [50][51]. - The potential for AI to replace human tasks raises questions about the future roles of individuals in the workforce, emphasizing the importance of understanding AI's workings and integrating it effectively into practical applications [52].

Group 1 - The core viewpoint of the article emphasizes that AI is not just a flashy technology but is fundamentally restructuring products and user experiences [1] - The article highlights a live event featuring key figures from Plaud, Rokid, and Fuxi Technology, focusing on the underlying logic of AI hardware evolution and commercialization [2][4] - The discussion will cover how companies like Plaud and Rokid have managed to stand out in the AI hardware sector and the secrets behind the sustainable commercialization of AI hardware [4] Group 2 - The live event is scheduled for July 30, from 20:00 to 21:30, under the theme "Beyond Tools: The Underlying Logic and Breakthrough Path of AI Hardware Advancement" [2] - Key speakers include Mo Zihua, CEO of Plaud China, Duan Ran, CEO of Fuxi Technology, and Zhao Weiqi, Global Development Ecosystem Leader at Rokid [3] - Participants are encouraged to submit questions for the speakers, which will be addressed during the live session [5]