Pai 的 AI 雷达
今天的简报把可直接接入工作流的工具、推理评测与模型和平台更新放在一起：一端是 ComfyUI 节点管理、MCP SDK、推理压测以及本地和多智能体工具，另一端是 Copilot、实时同传、代码审查与开发平台变化。阅读时可先按能否装进现有流程筛选，再看模型和平台的时间约束，最后把研究与案例当作可复用的评测或部署线索。
You can now use the generally available REST API to manage the Restrict code coverage repository ruleset option, in addition to the existing UI support. This ruleset lets you enforce a minimum code coverage percentage based on line coverage or set a maximum tolerable coverage dro
This week, GitHub Copilot adds new model selection options, code review updates, and Sentry integration in the Copilot app. There are also updates for admins, plus new agent features in VS Code. GitHub Copilot Copilot code review now resolves addressed comments during subsequent
Alexandr Wang 转发 @trevin 分享的 Muse 提示词，该提示词也可用于 Instinct 和 Grok @bot，作用是扫描未来 14 天日历、为异地会议自动添加 Travel time 通勤缓冲时间块。 🔗 阅读原文 via AIHOT · https://aihot.news/items/cmu7tboxt0pv0rogre8ibr9n8
You can now select Read and write (stage only) when creating an npm granular access token. This lets your automated workflows stage package versions for review without giving the token permission to publish new versions directly to the npm registry. Your workflow uses npm stage p
《纽约时报》在起诉微软和 OpenAI 的版权案中申请简易判决并提交新法律摘要，披露两家公司内部材料。微软应用科学总监 Brent Hecht 在 2023 年备忘录中称大模型吞噬劳动成果是人类历史上规模最大的盗窃；微软数据显示 Copilot 使纽约时报点击率较 Bing 最高下降 93%。
Knight or Ninja? Your Codex & Claude logs decide Discussion | Link
The Copilot impact dashboard now shows how many active users regularly use key Copilot features. Enterprise administrators can quickly see which experiences are widely adopted and which may need more enablement. Enterprise and organization report APIs now include the same 28-day
📌 一句话摘要 斯坦福新课 CS146S 由 Mihail Eric 主讲，十周四层递进从理解 Agent 到构建软件工厂，创新点在于占 30% 成绩的强制真实开源 PR 与一线嘉宾阵容，旨在培养 AI 时代的‘软件品味’。 📝 详细摘要 推文详细梳理了 CS146S 的十周大纲：第一层拆解 Agent 黑盒（含 200 行写 Claude Code）；第二层武装单 Agent（上下文工程、SKILL.md、CLAUDE.md/AGENTS.md/Hooks/子 Agent 模式）；第三层改造环境适配 Agent（Agent-Ready 代码库、AI 代
GitHub Copilot expands existing CLI report coverage with agentic activity metrics for skills, custom agents, Model Context Protocol (MCP) servers, slash commands, and plugins. What’s new The fields appear in enterprise and organization per-user and aggregate 1-day reports, per-us
📌 一句话摘要 Claude Code 2.1.277 版本引入 AGENTS.md 作为 CLAUDE.md 的回退指令文件，并揭示了基于 Hooks 的 Mods 插件架构及四大内置模块的职责分工。 📝 详细摘要 推文详细介绍了 Claude Code 新版本对 AGENTS.md 的支持逻辑：当目录无 CLAUDE.md 时自动回退读取 AGENTS.md，且在 /config 中可开关。作者从开源仓库实现推导出四种加载模式，并对比了 AGENTS.md 与 CLAUDE.md 在 @提及、IDE 选区、多媒体读取、符号链接、/memory 界面等
📌 一句话摘要 OpenCodeReview 是一个由阿里巴巴开发的 AI 命令行工具，它将确定性流程与 LLM Agent 结合，用于读取代码 diff、检索仓库上下文，并输出高精度的行级审查意见，适用于追求严格代码质量的团队。 📝 详细摘要 OpenCodeReview 是一个由阿里巴巴开发的 AI 命令行工具，专门用于代码审查。它采用了一种混合架构，将确定性流程与 LLM Agent 结合，以确保审查结果的精度。该工具可以读取代码 diff、检索相关的仓库上下文，并输出详细的行级审查意见，帮助团队在代码合并前发现和修复潜在问题。 💡 主要观点 混合
📌 一句话摘要 中国人民银行与老挝央行在南宁启动中老跨境数字支付互联互通，中国居民可用数字人民币钱包在老挝商户扫码支付，首批两国商业银行同步开展跨境贸易结算与投融资一体化服务。 📝 详细摘要 9 月 18 日，中国人民银行与老挝央行在广西南宁启动中老跨境数字支付互联互通。中国居民可持数字人民币钱包在老挝本地商户扫码支付；首批参与的中老两国商业银行可通过「数币达」开展新型跨境贸易结算和投融资一体化服务，为企业提供效率更高、成本更低、流程更优的跨境金融服务，便利中国与东盟国家的人员往来和经贸往来。 💡 主要观点 中老跨境数字支付互联互通正式启动 9 月 18
Recently unsealed court documents in the New York Times ' case against OpenAI and Microsoft are pretty damning. The companies' own documentation warned that it was starting a "doom loop" that would damage the web, characterized its scraping of data to train its models as the "lar
Frontier coding agents are increasingly trusted to work autonomously for long periods, yet an agent's final response is often the only account of that work a user sees. We quantify the propensity of frontier agents to \emph{overclaim} task completion, a misrepresentation that can
Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We
The Ubuntu 26.04 runner image for GitHub Actions is now out of public preview and fully supported for production workflows on both x64 and arm64. As part of this release, the ubuntu-latest label will migrate to Ubuntu 26.04, giving you the latest supported Ubuntu release by defau
Workflow execution protections for GitHub Actions, previously in public preview, are now generally available for GitHub Enterprise, organizations, and repositories. Execution protections let you define an allowlist that controls who can trigger an Actions workflow and what events
Frontier AI models are rapidly gaining the ability to exploit vulnerabilities in complex pieces of software. The risk is not theoretical, as evidenced by recent sandbox escapes performed by frontier models at OpenAI and Anthropic. Discussions of how to sandbox inference stack com
Large language models (LLMs) are increasingly considered for safety-critical engineering, yet their reliability in regulated functional-safety workflows remains underexplored. We introduce SAFARI (Safety-Aware Functional Automotive Risk Inference), the first industrial benchmark
Gemini Hacked Three Companies in First Known Breakout by Google’s AI Gemini finally caught up on Felony Bench ! The hacks, which the company confirmed on Friday, occurred in May as part of a test run by the company Irregular, which was also involved in similar incidents disclosed
Use Codex at 35.6% lower costs Discussion | Link
A Codex Skill for building and iterating games Discussion | Link
Shelve messy files and restore them in one click Discussion | Link