毫无疑问，上下文工程的优化，仍然是 Agent 创业公司在新一年都在「卷」的重点。 在实际落地开发中，上下文信息的质量，很大程度上决定了 Agent 的表现。 Manus 的首席科学家季逸超在之前访谈中提到过一个观点： 初创公司真的应该尽可能长时间地依赖通用模型和上下文工程，而不是过早地构建专用模型，也包括微调。 上下文工程是应用层和模型层之 间最清晰、最实用的边界。 做好上下文工程，开发者能够在不触及模型底层权重的前提下，灵活驾驭模型，同时还能适应快速变化的产品需求。 最近，Cursor 也发表了一篇文章《Dynamic context discovery》，分享了他们是怎么做上下文管理的。 结合 Manus、Cursor 这两家 Agent 领域头部团队的思路，我们整理了如何做好上下文工程的一些关键要点。 Cursor 原文：https://cursor.com/cn/blog/dynamic-context-discovery 此前 Founder Park 分享的文章 《来自 Manus 的一手分享：如何构建 AI Agent 的上下文工程？》 ⬆️关注 Founder Park，最及时最干货的创业分 ...

Core Insights - OpenAI has launched a new "group chat" feature that allows users to collaborate with friends and colleagues while integrating ChatGPT into discussions, marking a shift towards a collaborative SaaS platform [2][8] - The group chat feature is currently available to users in Japan, New Zealand, and South Korea, indicating a strategic expansion into new markets [2][3] Group 1: Functionality and User Experience - The group chat allows users to invite up to 20 members through a link, with data kept separate from personal ChatGPT memories to ensure privacy [3] - ChatGPT's interaction logic has changed; it can now autonomously determine when to respond based on the context of the conversation, enhancing user experience [4][5] - The AI can provide multimodal support, including real-time information searches, image generation, and document analysis, which enriches discussions [6][7] Group 2: Strategic Intent and Market Positioning - OpenAI's motivation for developing the group chat feature is to transition from merely selling APIs to creating a collaborative platform that retains user relationships and data [9] - The group chat is seen as a way to create network effects, making it harder for users to switch to competitors as their projects and data become integrated into the platform [9][10] Group 3: Future Implications - The group chat feature may serve as a testing ground for multi-agent collaboration, where different AI agents could work together on projects, enhancing productivity [11][12] - OpenAI aims to redefine human-machine collaboration in the AI era, positioning itself as a leader in this emerging paradigm [12][13]