Core Insights - Anthropic has acquired Vercept, a visual-driven AI automation startup, to enhance its Computer Use functionality for its AI model Claude, enabling it to perform complex tasks like a human [3][4] - The acquisition follows Anthropic's previous purchase of Bun in December 2025, marking its second acquisition aimed at improving AI capabilities [3] - Vercept's expertise in high-precision UI recognition and dynamic scene processing addresses the limitations of Claude's early visual understanding [3][4] Company Developments - Vercept, founded by alumni of the Allen Institute for AI, focuses on vision-first AI agents that automate tasks without APIs [4] - The company raised $16 million in seed funding in January 2025, with a valuation of approximately $67 million, backed by notable investors including former Google CEO Eric Schmidt [4] - Vercept's desktop application Vy will be discontinued within 30 days, encouraging users to transition to Anthropic's Claude as an alternative [5] Industry Context - The acquisition reflects increasing competition in the AI agent space, with both Anthropic and OpenAI actively developing capabilities for AI to control computers and complete tasks [6][7] - OpenAI is enhancing its personal AI agent capabilities through talent acquisition, indicating a strategic response to Anthropic's advancements in enterprise-level AI agents [6][7] - Dario Amodei, co-founder of Anthropic, emphasizes that the ultimate evolution of AI is a closed-loop execution system capable of general computer control, highlighting the industry's focus on improving interaction reliability [7]

Core Viewpoint - Anthropic has acquired Vercept, a visual-driven AI automation startup, to enhance its Computer Use functionality, which allows its AI model Claude to perform complex tasks like a human [3][4]. Group 1: Acquisition Details - The acquisition of Vercept is Anthropic's second purchase following the acquisition of Bun in December 2025 [3]. - Vercept specializes in high-precision UI recognition, spatial reasoning, and low-latency visual processing, addressing the limitations of Claude's early visual understanding capabilities [3][6]. Group 2: Technology and Capabilities - Computer Use enables Claude to operate software in real-time, completing multi-step tasks that cannot be achieved solely through coding [4]. - Vercept's expertise in perception and interaction is expected to directly benefit Anthropic's ongoing challenges in AI task execution [6]. Group 3: Market Context and Competition - The acquisition reflects the increasing competition in the AI Agent space, with both Anthropic and OpenAI actively developing their capabilities [7]. - OpenAI is also making strides in personal AI Agents, indicating a strategic focus on enhancing their operational capabilities in this domain [7][8]. Group 4: Future Implications - The acquisition is seen as a step towards transforming human-computer interaction, with expectations that Claude's capabilities will evolve to match its current coding abilities [7]. - Anthropic's founder emphasizes that the ultimate form of AI will be a closed-loop execution system capable of general computer control, which is currently hindered by reliability in task execution [8].

RT Claude (@claudeai)This is Claude Sonnet 4.6: our most capable Sonnet model yet.It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design.It also features a 1M token context window in beta. https://t.co/TDId3XUSRs ...

23 年 4 月以 AutoGPT 为代表的那一波里，Agent 更像是一个玩具，demo 都很炫，但实际应用价值很有限。 经过两年的发展，这波 Agent 确实能够在实际的工作和生活场景中解决问题，为大家带来价值了。 曲凯： Agent 是当下绝对的风口。关于 Agent 这个话题，我自己有一些核心在思考的问题，相信也是很多人同样会有疑问的地方。所以今天我们请来了长时间对 Agent 有研究和实操的文锋，想就这些问题展开一些讨论。 首先我想问，到底怎么定义 Agent？ 文锋： 我认为最好的就是 Anthropic 的定义：Agent 是让模型基于环境反馈去使用工具的一个程序。 曲凯： 那你怎么看最近这波 Agent 热？ 文锋： 这波 Agent 跟过去非常不一样。 之所以会有这种跃迁，一是因为底层模型能力有了很大的进步，尤其是在结合了 RL 之后，以 o1 为代表的模型还赋予了 Agent 长思维能力。 二是因为 Agent 的工程侧和产品侧也有很大的突破，主要表现就是大家更知道该怎么给 Agent 构建一个合适的 Context，从而更好地解决问题了。 曲凯： 怎么理解这个 Context？ 文锋： ...

编译：haozhen, Cookie AI agent 并不是一个新概念，但从 2024 年到今天，agent 的行动能力和交互方式发生了质变，头部模型厂商也正在将 agentic 能力融入模型，agentic 能 力会成为今年模型竞赛的重点之一， tool use 作为 agent 最重要的能力，一直是头部 AI labs 非常关注的方向。上周，OpenAI 发布了新一代模型 o3， o3 有最丰富的 tool use 方式。 本文是对 OpenAI agent 团队访谈的编译，OpenAI agent 产品和工程负责人分享了 OpenAI 在 agent 开发与工具生态方面的技术细节，以及他们对开发 者实践的观察与见解。他们认为，受益于 CoT 与 tool use 的结合，agent 获取信息的方式已经发生了巨变，agent 的下一步是能够接入数百个工具，并 能够自主判断调用哪个工具并确定如何使用。此外，multi agent 系统的工作效率会更高，且具有更高的可控性和优化潜力。 我们判断， multi agent 系统会在今年有大的突破，vertical agent 会因此直接受益，在 compute ...