🤖 AI Agent 研究Research
【arXiv】代理人工智能（ AI ）系统在材料发现和设计的复杂工作流程自动化方面拥有巨大的潜力，但将其扩展到领域密集型机械研究仍然具有挑战性。在这里，我们展示GBClaw ，一个
【arXiv】Agentic artificial intelligence (AI) systems hold immense potential for automating complex workflows in materials discovery and design, but extending them to domain-intensive mechanistic investigations remains challenging. Here, we present GBClaw, an
【arXiv】软件工程代理通过迭代工具使用来解决存储库级别的任务，但它们不断增长的交互痕迹会产生大量的代币消耗和推理开销。现有的压缩方法通常需要模型TRA
【arXiv】Software engineering agents solve repository-level tasks through iterative tool use, but their continuously growing interaction traces incur substantial token consumption and inference overhead. Existing compression approaches often require model tra
图形用户界面（ GUI ）代理越来越多地通过跨屏幕截图、音频、视频、交互历史记录和设备状态的流式全模式观察进行操作。然而，在每一步均匀地暴露这个完整的上下文忽略了相关性是暂时的和稀疏的：决定一个动作的信号在下一步可能会变得混乱。
Graphical User Interface (GUI) agents increasingly act over streaming omni-modal observations spanning screenshots, audio, video, interaction history, and device states. However, exposing this complete context uniformly at every step ignores that relevance is transient and sparse: the signal determining one action can become clutter at the next.
编码代理继续致力于验证失败的运行，要求前沿模型对此进行监控本身就很昂贵。我们引入了StopLoss ，这是一种小型外部监视器，可从已完成的轨迹及其验证器结果中学习故障识别。
Coding agents keep working on runs that will fail verification, and asking a frontier model to watch for this is expensive in its own right. We introduce StopLoss, a small external monitor that learns failure recognition from completed trajectories and their verifier outcomes.
⭐ GitHub 热门项目GitHub Trending
【GitHub】仅为数据库和Claude Code的开源理赔台。（⭐ 0 ）
【GitHub】The open-source claims desk that is just a database and Claude Code. （⭐ 0）
【GitHub】Claude Code 的调度大师，将每个任务路由到合适的模型（Haiku、Sonnet、Opus、Fable）。开源修改版，无网络、无遥测。（⭐ 1）
【GitHub】A maestro for Claude Code that routes each task to the right model (Haiku, Sonnet, Opus, Fable). Open-source mod with no network and no telemetry. （⭐ 1）
【GitHub】人工智能代理实时与您的团队一起工作的开源实时工作空间：实时文档、工作表和幻灯片、共享浏览器、Linux桌面和全双工语音。带上Claude Code、Codex、Gemin （⭐ 4 ）
【GitHub】Open-source live workspace where AI agents work alongside your team in real time: live docs, sheets and slides, a shared browser, a Linux desktop and full-duplex voice. Bring Claude Code, Codex, Gemin （⭐ 4）
【GitHub】一个3D虚拟代理办公室，四个人工智能机器人代理（ Evan、Lummi、Athena、Zed ）在这里聊天、协作并完成工作。接待经理Zed接管您的命令并委派任务。包括代理状态和权限的实时活动面板、Zed自动运行的任务日历以及Gmail、Drive、GitHub等插件。（⭐ 14 ）
【GitHub】A virtual agentic office in 3D where four AI robot agents (Evan, Lummi, Athena, Zed) chat, collaborate and get work done. Zed, the reception manager, takes your commands and delegates tasks. Includes a live activity panel for agent status and permissions, a task calendar Zed runs automatically, and plugins like Gmail, Drive, GitHub. （⭐ 14）
【GitHub】防滑设计+适用于AI编码代理的GSAP运动技能包，带有82规则UI检测器。（⭐ 9 ）
【GitHub】Anti-slop design + GSAP motion skill pack for AI coding agents, with an 82-rule UI detector. （⭐ 9）
🚀 模型与行业动态Models & Industry
拟议的政策似乎提供了针对人工智能疯狂运行的保护措施，但实际上禁止联邦机构在未经法官许可的情况下阻止模型的发布。
The proposed policy seems to offer protections against AI run amok but actually forbids federal agencies from blocking a model's release without the permission of a judge.
OpenAI表示，它正在增加Orchestra的团队和技术，以开发其用于生命科学研究的人工智能模型Rosalind ，以及其相关的Rosalind Workbench工具，该工具旨在帮助研究人员计划实验和分析结果。
OpenAI says it's adding Orchestra's team and technology to work on Rosalind, its AI model for life sciences research, and to its related Rosalind Workbench tool, a harness that aims to help researchers plan experiments and analyze results.
该公司表示，该功能专为用户可能正在讨论个人或计划惊喜的情况而设计，并且在将设备交给朋友或家人以在另一个聊天中向他们展示某些内容时，希望保持这些对话的私密性。
The company says the feature is designed for situations where users might be discussing something personal or planning a surprise and want to keep those conversations private when handing their device to friends or family to show them something in another chat.
Figma表示，它计划在不暴露所有数据的情况下支持尽可能多的标准。
Figma said it plans to support as many standards as possible without exposing all of its data.
🔥 社区热议Community
【HN】热度: 2142 分 | 835 评论
【HN】热度: 2142 分 | 835 评论
【HN】热度: 28 分 | 0 评论
【HN】热度: 28 分 | 0 评论
【HN】热度: 45 分 | 18 评论
【HN】热度: 45 分 | 18 评论
【HN】热度: 94 分 | 63 评论
【HN】热度: 94 分 | 63 评论
【HN】热度: 129 分 | 66 评论
【HN】热度: 129 分 | 66 评论