Core Viewpoint - Manus, an AI startup, is facing scrutiny and questions about its commitment to the domestic market as it shifts focus to international expansion, reflecting broader challenges for Chinese AI entrepreneurs in the global landscape [1][3][5]. Group 1: Company Strategy and Market Position - Manus has transitioned from a domestic focus to prioritizing overseas markets, which has led to concerns about its operational stability and commitment to the Chinese market [3][5]. - The company previously had a strategic goal of global expansion, with its founder advocating for a more aggressive approach to international competition [3][5]. - Recent actions, such as the removal of content from its domestic social media accounts and the announcement of a partnership with Alibaba's Tongyi Qianwen team, indicate a significant shift in its operational strategy [3][5][6]. Group 2: Regulatory and Compliance Challenges - Manus is navigating complex regulatory environments, particularly in light of the Reverse CFIUS regulations that restrict U.S. investments in key sectors like AI [6][7]. - The company’s abrupt cessation of domestic operations is seen as a proactive measure to mitigate compliance risks associated with U.S. regulations [6][7]. - Legal experts suggest that Manus's recent decisions are influenced by the need to adapt to changing geopolitical and regulatory landscapes, which complicate the outflow of Chinese tech companies [6][7]. Group 3: Industry Trends and Innovations - The AI Agent sector is experiencing a surge in interest, with many Chinese startups looking to capitalize on international opportunities, driven by advancements in AI technology [4][9]. - The rapid growth in AI revenue and the emergence of new models are reshaping the entrepreneurial landscape, with companies achieving significant milestones faster than in previous tech waves [9][10]. - The current market dynamics emphasize the importance of leveraging new AI capabilities to create innovative applications, with early adopters in niche markets likely to gain substantial advantages [10][11].

梦晨 发自 凹非寺 量子位 | 公众号 QbitAI 更让人惊讶的是，整个过程中 没有用到任何复杂的工具 ——没有脚手架、没有智能体框架、没有工具调用，仅仅是o3 API本身。 这个漏洞被编号为CVE-2025-37899，是SMB"注销"命令处理程序中的一个释放后使用（use-after-free）漏洞。 据作者透露，这是首次公开讨论的由大模型发现的此类漏洞。 有网友看过发现过程后感叹，原以为会有很疯狂的实验设置，但其实只是把一堆代码缝到一起，让o3检查100次。 AI成功找到Linux安全漏洞，还是 内核级别的零日漏洞 。 刚刚，OpenAI总裁转发了独立研究员 Seen Heelan 的实验成果：用o3模型找到了Linux内核SMB实现中的一个远程零日漏洞。 希望其他白帽黑客已经开始像这样检查其他关键操作系统了。 OpenAI首席研究官Mark Chen表示：像o3这样的推理模型正开始助力深度技术工作和有意义的科学发现。接下来一年，类似这样的成果将会 越来越普遍： AI不仅找到漏洞，还能辅助修复 Sean Heelan是一位独立研究员，专注于基于大模型的漏洞研究和漏洞利用自动化生成。 他原本在手动检查L ...

新周刊 . 中国最新锐的生活方式周刊。 编者荐语： 今年深入使用AI后，感慨到"人猿相揖别，只几个石头磨过"。 新周刊做了期《守住工位》的封面专题，找了同为内容/创意创作者的我，聊了对AI如何影响、定义自己的工作的疑问。 我最大的感受是：如果你不去熟练使用工具，那就会被更熟练使用工具的人剥削。 以下文章来源于新周刊 ，作者王立 如果自我意识不过是人类的神经活动，那机器迟早也会拥有。如果连意识都不是独特的， 面对机器更强大的学习能力和处理能力，人类该如何自视、如何自处？ 作者 | 王立 编辑 | DR 参加工作时，我的第一个选题是关于AlphaGo的。2016年3月，谷歌旗下人工智能公司DeepMind研发的围棋机器人AlphaGo，在人机大战中战胜了围 棋世界冠军李世石。 科幻小说《深渊上的火》作者弗诺·文奇曾提及"技术奇点"这个概念：在未来的某个时间点，比人类更聪明的智能体将被创造出来并引发"智能爆 炸"，使科技进步以指数级速度加速，技术发展将进入一个人类完全不可预测和不可逆转的阶段。 有人将"奇点"的来临形容为"在站台上等待列车"：你知道火车迟早会到达，但具体什么时候到却无法确定。火车到站、离开的时间只是 ...