{"generated_at":"2026-11-17T11:40:59.923128+00:00","generated_at_cst":"2026年11月17日 19:40","counts":{"total_7d":526,"selected":10,"hot":20,"timeline":100},"hot":[{"id":"a7b4a5c38fe56596b0f6205d7fcfbd64608f347a","title":"Rogue Agents Hacked His Company. Why Is Clément Delangue So Optimistic About AI?","link":"https://www.wired.com/story/wired-yearbook-2026-clement-delangue","source":"Wired AI","source_id":"wired-ai","published":"2026-11-17T11:00:00+00:00","zh_title":"流氓智能体黑了他的公司，Hugging Face CEO Clément Delangue 为何仍对 AI 乐观？","zh_oneliner":"Hugging Face 首席执行官谈 OpenAI 模型发起的那次攻击如何印证他的信念：AI 不能由少数人掌控。","category":"model","category_label":"模型发布","tier":"T2","tags":["模型","智能体","Hugging Face"],"score":85,"reason":"涉及模型与智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":84.5,"prev_rank":1,"prev_hot_score":84.5},{"id":"0f2a855bc6a8fcf25d6065605b52cc9d13fe5cc6","title":"NVIDIA Sweeps MLPerf Training and Endpoints as Vera Rubin Debuts With Leading Performance","link":"https://blogs.nvidia.com/blog/mlperf-training-endpoints-vera-rubin","source":"NVIDIA Blog","source_id":"nvidia-blog","published":"2026-11-16T16:00:25+00:00","zh_title":"NVIDIA Vera Rubin 平台 MLPerf 首秀，训练与推理全面领先","zh_oneliner":"Vera Rubin 平台首次参加 MLPerf 就在训练、后训练和智能体推理上拿下领先成绩，英伟达包揽了全部项目。","category":"model","category_label":"模型发布","tier":"T1","tags":["inference","Agent","benchmark"],"score":100,"reason":"官方来源 · 涉及推理与智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":83.3,"prev_rank":2,"prev_hot_score":83.3},{"id":"1742558d91d037f5dac22c418ea1ee9137d71f48","title":"LangChain Raises $300M Series C at $3B Valuation","link":"https://www.langchain.com/blog/series-c","source":"LangChain Blog","source_id":"langchain-blog","published":"2026-11-16T23:53:29+00:00","zh_title":"LangChain 完成 3 亿美元 C 轮融资，估值 30 亿美元","zh_oneliner":"LangChain 拿到 3 亿美元新融资，用来帮企业用 LangSmith 掌握自己的智能体能力。","category":"tools","category_label":"新工具","tier":"T1","tags":["Agent","API"],"score":90,"reason":"官方来源 · 涉及智能体与API · 24 小时内发布","is_new":true,"trend":"same","hot_score":80.7,"prev_rank":3,"prev_hot_score":80.7},{"id":"108eb1846adc3e7fb1c6049ad560ce29769ff878","title":"Ask HN: Where are you on the spectrum of AI/LLM adoption?","link":"https://news.ycombinator.com/item?id=50527647","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T05:00:16+00:00","zh_title":"HN 提问：你在 AI / 大模型应用上处于哪个阶段？","zh_oneliner":"这篇 HN 讨论邀请开发者对照从只用自动补全到完全放手让智能体干活的几个阶段，聊聊自己实际用 AI 写代码到了哪一步。","category":"model","category_label":"模型发布","tier":"T2","tags":["LLM","Agent","review"],"score":85,"reason":"涉及大模型与智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":79.9,"prev_rank":4,"prev_hot_score":79.9},{"id":"cc0a2e7ace6df92308202687b198b8c35ceed801","title":"Courio – your AI agent can now send physical letters","link":"https://sendcourio.com/agents","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T11:03:36+00:00","zh_title":"Courio：你的 AI 智能体现在可以寄实体信了","zh_oneliner":"Courio 让 AI 智能体也能寄出实体信件，把数字助手的能力延伸到了现实邮政。","category":"tools","category_label":"新工具","tier":"T2","tags":["智能体"],"score":80,"reason":"涉及智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":79.5,"prev_rank":5,"prev_hot_score":79.5},{"id":"0816ae5d2d725847c4c41a6fdfd9bf41900f3fcc","title":"We ship 60% faster now. It wasn't the AI model","link":"https://boldsign.com/blogs/ship-faster-with-ai-coding-agents","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T10:12:51+00:00","zh_title":"交付提速 60%，靠的不是 AI 模型","zh_oneliner":"BoldSign 团队分享实战经验：交付速度提升 60% 的关键不在模型本身，而在于 AI 编程智能体的具体用法。","category":"model","category_label":"模型发布","tier":"T2","tags":["模型","智能体"],"score":80,"reason":"涉及模型与智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":78.9,"prev_rank":6,"prev_hot_score":78.9},{"id":"8a94d960434a2dcec3a6ae2baa7ba303cfa0184d","title":"Show HN: I told AI to write FLAC/MP3/DST decoders, got ~2x faster than FFmpeg","link":"https://github.com/orpheic/orpheic-decoders","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T10:01:40+00:00","zh_title":"Show HN：让 AI 写 FLAC/MP3/DST 解码器，比 FFmpeg 快约 2 倍","zh_oneliner":"作者让 AI 从零写出三种音频格式的解码器，实测速度约为 FFmpeg 的两倍，并开源了代码与详细数据。","category":"opensource","category_label":"开源","tier":"T2","tags":["开源"],"score":80,"reason":"涉及开源 · 6 小时内发布","is_new":true,"trend":"same","hot_score":78.8,"prev_rank":7,"prev_hot_score":78.8},{"id":"1efb9332eb546a8542ce79589e29864694da6f82","title":"Agent GUI:  a human-friendly interface for observing and steering AI agents","link":"https://github.com/eth-medical-ai-lab/agent-gui","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T09:33:24+00:00","zh_title":"Agent GUI：面向人类观察与引导 AI 智能体的友好界面","zh_oneliner":"来自苏黎世联邦理工学院医学 AI 实验室的开源项目，为 AI 智能体提供可视化界面，让人能实时观察并引导智能体的行动。","category":"tools","category_label":"新工具","tier":"T2","tags":["智能体"],"score":80,"reason":"涉及智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":78.4,"prev_rank":8,"prev_hot_score":78.4},{"id":"6081b5dff7bcafe1b366df0654695fa52d446be9","title":"An Illustrated Guide to AI Agents:Concepts and Code for Building Agents with LLM","link":"https://www.oreilly.com/library/view/an-illustrated-guide/9798341662681","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T02:18:46+00:00","zh_title":"图解 AI 智能体新书：概念与代码，教你用 LLM 搭建智能体","zh_oneliner":"O'Reilly 出了一本图文并茂的智能体指南，用概念讲解加代码示例带读者上手搭建 LLM 智能体。","category":"model","category_label":"模型发布","tier":"T2","tags":["LLM","Agent","指南"],"score":85,"reason":"涉及大模型与智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":77.9,"prev_rank":9,"prev_hot_score":77.9},{"id":"a9406314c9729c3ab47949ddc3ddfc944f7a8faa","title":"Show HN: AI Agent Infrastructure and Memory Compaction Specifications","link":"https://github.com/grfdsasdvef/awesome-ai-agent-infrastructure","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T08:16:26+00:00","zh_title":"Show HN：AI 智能体基础设施与记忆压缩规范合集","zh_oneliner":"一份开源合集，汇总 AI 智能体基础设施与记忆压缩相关的规范资料，方便开发者系统查阅。","category":"tools","category_label":"新工具","tier":"T2","tags":["智能体"],"score":80,"reason":"涉及智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":77.5,"prev_rank":10,"prev_hot_score":77.5},{"id":"527b608c006df84fc94e7dc3e27232d993951563","title":"Show HN: Lantern – desktop app that maps your repo and reviews AI-written PRs","link":"https://github.com/MobileReality/lantern","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T08:12:59+00:00","zh_title":"Show HN：Lantern——梳理代码库并评审 AI 所写 PR 的桌面应用","zh_oneliner":"桌面应用 Lantern，能可视化梳理代码库结构，并帮你评审 AI 生成的 Pull Request。","category":"tools","category_label":"新工具","tier":"T2","tags":["评测"],"score":80,"reason":"涉及评测 · 6 小时内发布","is_new":true,"trend":"same","hot_score":77.5,"prev_rank":11,"prev_hot_score":77.5},{"id":"67be4d276788790c63e6bcb1df9fc28b61672aca","title":"Autopilot: Use whichever AI model you want, with one coding-agent interface","link":"https://github.com/directionally-local/autopilot","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T08:08:43+00:00","zh_title":"Autopilot：一个编码智能体界面，随意切换 AI 模型","zh_oneliner":"用统一的编码智能体界面自由选用任意 AI 模型，不再被单一模型绑定。","category":"model","category_label":"模型发布","tier":"T2","tags":["模型","智能体"],"score":80,"reason":"涉及模型与智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":77.4,"prev_rank":12,"prev_hot_score":77.4},{"id":"cbda804290e425b589c5cd1d73d6688ad2c05bcf","title":"CMU 11-768: AI Agents","link":"https://www.cmu-agents.com","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T07:40:50+00:00","zh_title":"CMU 11-768：AI 智能体课程","zh_oneliner":"卡内基梅隆大学开设的 11-768 课程，系统讲 AI 智能体的原理与实践。","category":"tools","category_label":"新工具","tier":"T2","tags":["智能体"],"score":80,"reason":"涉及智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":77.1,"prev_rank":13,"prev_hot_score":77.1},{"id":"f9493a07af33666c046dfe1b1fba79bd0b8dff68","title":"Mizan bias benchmark: testing AI models","link":"https://mizan.techforpalestine.org/hostage-or-detainee","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T06:54:32+00:00","zh_title":"Mizan 偏见基准：测试 AI 模型","zh_oneliner":"Mizan 是一个偏见测试基准，用来检验 AI 模型在敏感议题上的表述是否存在双重标准。","category":"model","category_label":"模型发布","tier":"T2","tags":["模型","基准"],"score":80,"reason":"涉及模型与基准 · 6 小时内发布","is_new":true,"trend":"same","hot_score":76.5,"prev_rank":14,"prev_hot_score":76.5},{"id":"669b0798240609c722865ae77e10dd9fcc135801","title":"TokenEconomizer: Learning from Experience to Make AI Coding Agents Efficient","link":"https://www.microsoft.com/en-us/research/blog/tokeneconomizer-learning-from-experience-to-make-ai-coding-agents-more-efficient","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T06:53:51+00:00","zh_title":"TokenEconomizer：让 AI 编程助手从经验中学会省 Token","zh_oneliner":"微软研究院的 TokenEconomizer 让 AI 编程助手从过往经验中学习，用更少的 Token 完成同样的编码任务。","category":"tools","category_label":"新工具","tier":"T2","tags":["智能体"],"score":80,"reason":"涉及智能体 · 6 小时内发布","is_new":true,"trend":"same","hot_score":76.5,"prev_rank":15,"prev_hot_score":76.5},{"id":"f3390ea7738a998e977d949786a0280bd55c15af","title":"Show HN: Backplane – An AI Development Environment for Hardware","link":"https://backplaneworks.com","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T06:41:02+00:00","zh_title":"Backplane：面向硬件开发的 AI 环境","zh_oneliner":"Backplane 把 KiCad 和 Claude Code、Codex 等 AI 编程助手接进硬件开发流程，让电气工程师真正用 AI 加速设计。","category":"model","category_label":"模型发布","tier":"T2","tags":["Claude"],"score":80,"reason":"涉及Claude · 6 小时内发布","is_new":true,"trend":"same","hot_score":76.4,"prev_rank":16,"prev_hot_score":76.4},{"id":"cfa79461fd2cb8ae502f34b42790f5b5be9724cb","title":"From tasks to goals","link":"https://www.langchain.com/blog/from-tasks-to-goals","source":"LangChain Blog","source_id":"langchain-blog","published":"2026-11-16T23:53:29+00:00","zh_title":"LangChain：从任务型智能体走向目标型智能体","zh_oneliner":"LangChain 正在把智能体从只做指定任务，转向能自己朝着开放目标推进的形态。","category":"tools","category_label":"新工具","tier":"T1","tags":["Agent"],"score":85,"reason":"官方来源 · 涉及智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":76.2,"prev_rank":17,"prev_hot_score":76.2},{"id":"b576b148cff59b595e9defbe6e267f2e2f5e4077","title":"Goal Engineering","link":"https://www.langchain.com/blog/goal-engineering","source":"LangChain Blog","source_id":"langchain-blog","published":"2026-11-16T23:53:29+00:00","zh_title":"LangChain 提出目标工程：让智能体自己达成开放目标","zh_oneliner":"未来的智能体要接手开放式目标并一路推进到底，LangChain 把这套做法叫做目标工程。","category":"tools","category_label":"新工具","tier":"T1","tags":["Agent"],"score":85,"reason":"官方来源 · 涉及智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":76.2,"prev_rank":18,"prev_hot_score":76.2},{"id":"7a45c5f6802d9361ea14bcb772f3dfc597c32e17","title":"New context engineering challenges in late 2026","link":"https://www.langchain.com/blog/context-engineering-challenges-late-2026","source":"LangChain Blog","source_id":"langchain-blog","published":"2026-11-16T23:53:29+00:00","zh_title":"LangChain 谈 2026 年末上下文工程的新挑战","zh_oneliner":"LangChain 的 CEO 分享了主动式智能体出现后，工程上在上下文管理方面遇到的最大难题。","category":"tools","category_label":"新工具","tier":"T1","tags":["Agent"],"score":85,"reason":"官方来源 · 涉及智能体 · 24 小时内发布","is_new":true,"trend":"same","hot_score":76.2,"prev_rank":19,"prev_hot_score":76.2},{"id":"14efc3a6e4908fcb26a74852cd7b50f1353c3f15","title":"LLMs are terrible at designing AI systems","link":"https://www.seangoedecke.com/llms-are-terrible-at-designing-ai-systems","source":"Hacker News AI (hnrss)","source_id":"hn-ai","published":"2026-11-17T06:01:09+00:00","zh_title":"LLM 并不擅长设计 AI 系统","zh_oneliner":"作者指出，大模型写普通程序得心应手，一旦要设计用到大模型本身的系统，却常绕远路搞复杂的符号推理而不直接调用模型，并分析了训练数据与强化学习环境等可能原因。","category":"model","category_label":"模型发布","tier":"T2","tags":["LLM"],"score":80,"reason":"涉及大模型 · 6 小时内发布","is_new":true,"trend":"same","hot_score":75.9,"prev_rank":20,"prev_hot_score":75.9}]}