{"mode": "selected", "count": 200, "items": [{"url": "https://blog.okturtles.org/2026/07/short-leash-ai-method", "source": "Hacker News 热门", "title": "《Fable》通关指南：短绳AI编程法", "excerpt": "专业开发者经过一年多研究，总结出使用AI编码代理的\"短绳方法\"。该方法要求开发者全程参与：先规划并分解任务，从不使用YOLO模式，每次变更前审查差异并拒绝不想要的更改，每个子任务后提交以防止AI误操作（如Opus曾出现破坏性行为）。最终需进行人工与AI双重PR审查，PR须注明使用模型，提交者须亲自审查自己PR的代码。即便不用前沿模型，此法也能产出超越Fable 5的代码质量。", "published_at": "2026-07-03T06:16:11Z", "score": 8.0, "summary_zh": "专业开发者总结用AI编码代理的“短绳方法”：规划任务、审查差异、频繁提交、双重PR审查，避免AI误操作。", "category": "教程观点", "tags": "[\"AI编程\", \"短绳方法\", \"工作流\", \"Agent安全\"]", "why_zh": "可转化为“普通人如何安全使用AI写代码”的实操教程，适合社群作业。", "idea_zh": "选题：Codex/Claude Code 实操避坑指南——“短绳方法”怎么用。小红书图文形式，分步骤说明：规划、审查、提交、PR。附真实案例对比，强调“别让AI乱改代码”。可布置社群作业：用此方法完成一个小项目。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/da4c8ca57d40b074bdc1b5b1218851100150c56b", "source": "OpenAI Codex Commits", "title": "[codex] Add configurable multi-agent mode hint text (#30493)", "excerpt": "[codex] Add configurable multi-agent mode hint text (#30493) ## Why Multi-agent V2 normally derives its mode instructions from reasoning effort: Ultra enables proactive delegation, while other efforts require an explicit request. Some deployments need to provide one configured delegation policy that replaces those built-ins and remains stable when reasoning effort changes. ## What changed - Add `features.multi_agent_v2.multi_agent_mode_hint_text` alongside the existing root and subagent hint set", "published_at": "2026-07-03T01:44:34Z", "score": 7.0, "summary_zh": "Codex 多智能体模式新增可配置的委派策略提示文本，允许用户覆盖默认的推理驱动行为。", "category": "产品动态", "tags": "[\"Codex\", \"多智能体\", \"功能更新\", \"委派策略\"]", "why_zh": "Codex 多智能体功能向可定制化演进，适合测试自动化工作流中的委派场景。", "idea_zh": "选题：Codex 多智能体模式更新，教你设置固定委派策略。可做图文对比 Ultra 主动委派与显式请求模式的差异，并布置社群作业。", "title_zh": ""}, {"url": "https://mp.weixin.qq.com/s/L6R_SPWlOBv6dI0wWWHQrg", "source": "公众号：数字生命卡兹克", "title": "Claude Fable 5 自主优化 AIHOT 网站 SEO/GEO 全记录", "excerpt": "作者用 Claude Fable 5 优化 AIHOT 网站的 SEO 与 GEO。模型自主启动 22 个 Agent 调研 40 分钟，发现豆包 App 每天六千多次访问未被统计等异常。规划境外加速时，否定 Claude Opus 4.8 的 Cloudflare 方案（无法国内直连/国外分流，且 2025 年起默认拦截 AI 爬虫），改用火山引擎 CDN。因需白名单，模型自行找到工单入口提交专业工单，22 分钟开通；发现工程师漏答回源 IP 网段问题，礼貌追问并补充备选方案；发现官方方案有安全漏洞，自行加暗号验证。23：30 切换域名解析，10 分钟后 616 个海外请求走新线路。最终生成运维文档，提醒边缘证书 10 月 2 日到期并附续期步骤。", "published_at": "2026-07-03T00:16:00Z", "score": 8.0, "summary_zh": "Claude Fable 5自主启动22个Agent优化网站SEO，发现流量统计异常并处理CDN迁移等复杂任务。", "category": "教程观点", "tags": "[\"Claude F\", \"多Agent\", \"SEO优化\", \"网站自动化\"]", "why_zh": "Claude Fable多Agent自主工作流实操，可对比Codex能力，适合做成避坑指南", "idea_zh": "小红书图文：Claude Fable 5是如何72分钟自主搞定网站优化的？对比Codex Code的任务委派，哪些活它能干？屿鹿可以出实测对比视频，讨论Agent的边界。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/07/02/meet-alibabas-page-agent-a-javascript-in-page-gui-agent-that-controls-web-interfaces-with-natural-language-through-the-dom", "source": "MarkTechPost", "title": "阿里巴巴发布 Page Agent：开源 JavaScript 库实现网页 DOM 自然语言操控", "excerpt": "阿里巴巴发布 Page Agent，一个开源的 JavaScript 客户端库，嵌入网页后可通过自然语言指令直接操作 DOM 元素。与 Playwright、Puppeteer 等外部浏览器自动化工具不同，Page Agent 不依赖截图或多模态模型，而是将实时 DOM 脱水压缩为 FlatDomTree 文本映射，让纯文本模型精准执行点击、表单填写等操作。它继承用户 cookies 和会话，无需独立后端，并支持任意 OpenAI 兼容端点的模型（示例使用 `qwen3.5-plus`）。项目采用 MIT 许可证，适合在自有应用内构建 AI 副驾、智能表单填充或无障碍控制等场景，但限于单页面范围，风险操作仍需服务端验证。", "published_at": "2026-07-02T20:51:39Z", "score": 7.0, "summary_zh": "阿里开源Page Agent，通过自然语言直接操作网页DOM，无需截图和多模态模型。", "category": "产品动态", "tags": "[\"开源\", \"DOM操作\", \"浏览器自动化\", \"自然语言\"]", "why_zh": "自然语言操控网页的新工具，可与Codex/Claude等配合构建浏览器自动化工作流，适合教程", "idea_zh": "小红书图文：阿里Page Agent，用说话就能自动填表点击？对比Selenium/Playwright，普通人怎么用？屿鹿社群可探讨集成到AI工作流的方法。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jul/2/llm-coding-agent/#atom-everything", "source": "Simon Willison", "title": "llm-coding-agent 0.1a0", "excerpt": "Release: llm-coding-agent 0.1a0 Another Fable 5 experiment. Now that my LLM library has evolved into more of an agent framework it's time to see what a simple coding agent would look like built on it. I started a new Python library using my python-lib-template-repository GitHub template repository, then ran these two prompts (here's the Claude Code for web transcript ): Write a spec.md for this project - it will depend on the latest “llm” alpha from PyPI and implement a Claude code style coding ", "published_at": "2026-07-02T19:33:12Z", "score": 7.0, "summary_zh": "Simon Willison 发布实验性编码代理 llm-coding-agent 0.1a0，基于其 LLM 库构建。", "category": "产品动态", "tags": "[\"AI编程代理\", \"Codex替代\", \"开源工具\"]", "why_zh": "可对比 Codex，为屿鹿社群提供 AI 编码代理的新选项和实操教程素材。", "idea_zh": "标题：又一个 AI 编码代理来了！实测 llm-coding-agent 和 Codex 谁更适合非程序员？ 可做图文对比，重点看安装难易、命令执行、上下文管理。引导社群讨论哪个更适合自己的一人公司提效。", "title_zh": ""}, {"url": "https://github.com/HUANGCHIHHUNGLeo/claude-real-video", "source": "Hacker News", "title": "Claude-real-video － any LLM can watch a video", "excerpt": "Article URL: https://github.com/HUANGCHIHHUNGLeo/claude-real-video Comments URL: https://news.ycombinator.com/item?id=48766005 Points: 5 # Comments: 0", "published_at": "2026-07-02T19:10:12Z", "score": 8.0, "summary_zh": "开源工具 Claude-real-video 让任意 LLM（含 GPT）能够观看视频", "category": "教程观点", "tags": "[\"多模态\", \"视频理解\", \"GPT工具\", \"实测对比\"]", "why_zh": "可实测对比 GPT-4o 原生视频能力，产出普通用户能照做的教程，契合「AI提效」定位", "idea_zh": "制作「让 GPT 看懂视频的新方法」实测：对比 claude-real-video 工具与 ChatGPT 直接上传视频的差异，演示安装、使用、效果，给出适用场景，做成小红书图文教程或视频号演示。", "title_zh": ""}, {"url": "https://www.lmsys.org/blog/2026-07-02-agent-assisted-sglang-development", "source": "LMSYS：Blog", "title": "Agent辅助的SGLang开发：初步探索", "excerpt": "SGLang团队将LLM服务、分布式运行时、GPU内核、扩散管道等工作流编码为可执行的SKILL.md文件、脚本、基准合约和审查循环。现有技能包括：SGLang .claude/skills（CUDA调试、内核集成、性能分析等）、SGLang diffusion .claude/skills（扩散模型添加与调优）、BBuf/AI-Infra-Auto-Driven-SKILLS（跨框架SOTA循环）、KDA（MLSys 2026 FlashInfer内核竞赛获胜方案）以及BBuf/KDA-Pilot（已合并三个SGLang集成PR）。Profile证据是性能工作的核心，长期优化转向Loop Engineering--SGLang SOTA Performance Loop将追求SOTA分解为公平基准测试、差距决策、性能分析、补丁和再验证，Humanize/RLCR添加外部审查，Codex Goal以更低协调开销运行相同循环。评审重要性提升，开发者需定义问题、选择证据、设计工作流并判断结果是否可用于生产。", "published_at": "2026-07-02T18:37:56Z", "score": 8.0, "summary_zh": "SGLang团队用Skill文件指导AI辅助开发，包括CUDA调试、内核集成和性能分析。", "category": "教程观点", "tags": "[\"Skill\", \"AI编程\", \"SGLang\", \"开发工作流\"]", "why_zh": "Skill驱动的AI开发工作流，可直接启发Codex Skill编写与任务委派，可做教程", "idea_zh": "小红书图文：如何用Skill文件让AI帮你写高性能代码？SGLang团队的实战经验，屿鹿可以出Codex Skill编写教学，比较与Claude Code Skills的异同。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-07-02-copilot-agent-session-streaming-is-now-in-public-preview", "source": "GitHub Changelog", "title": "Copilot agent session streaming is now in public preview", "excerpt": "GitHub Enterprise Cloud customers with enterprise managed users can now access GitHub Copilot agent session data across all Copilot clients, including: Cloud agents operating on github.com and data resident deployments… The post Copilot agent session streaming is now in public preview appeared first on The GitHub Blog .", "published_at": "2026-07-02T18:05:05Z", "score": 8.0, "summary_zh": "GitHub Copilot 代理会话流式数据进入公开预览，企业用户可访问所有客户端代理会话记录。", "category": "产品动态", "tags": "[\"GitHub C\", \"Agent 会话\", \"公开预览\"]", "why_zh": "Copilot 新增代理会话监控，可直接对比 Codex 工作区记忆与任务委派，适合做差异选题。", "idea_zh": "小红书图文：Copilot 能回放代理操作了，和 Codex 的“记住上下文”有什么不同？社群作业：用 Copilot 代理完成一个简单任务，导出会话记录观察 AI 的决策过程。", "title_zh": ""}, {"url": "https://getquality.md", "source": "Hacker News", "title": "Show HN: QUALITY.md – open format/specification, agent skill, and CLI", "excerpt": "Hello all, I created QUALITY.md to help build a holistic quality evaluation process for my projects. Turns out it's also ideal for loop engineering. I'm hoping this provides a valuable contribution to the conversation around quality and craft and having AI help us in the effort. I hope to shift the mindset from a reactive/review/repair mindset to a proactive care mindset. Give it a go. I look forward to your thoughts/comments/feedback! Website: https://getquality.md GitHub: https://github.com/qu", "published_at": "2026-07-02T16:36:12Z", "score": 7.0, "summary_zh": "QUALITY.md 开放规范，结合 agent skill 与 CLI，用于项目质量评估和循环工程。", "category": "产品动态", "tags": "[\"Agent Sk\", \"自动化工作流\", \"开源工具\"]", "why_zh": "与 Codex Skill 类似，可对比展示 AI 代理技能化工作流。", "idea_zh": "实测「用 Agent Skill 自动审查项目质量」，对比 Codex Skill 的写法和触发方式，整理成小红书图文教程。", "title_zh": ""}, {"url": "https://github.com/ctxrs/ctx", "source": "Hacker News", "title": "Show HN: ctx – Search the coding agent history already on your machine", "excerpt": "Coding agents don't have long-term memory. But you do have months of full-fidelity agent transcripts stored on your machine. A simple solution that goes a long way: ingest those transcripts and logs into a structured SQLite database, then search them with ranked text match. Everything is fully local and doesn't require anything fancy like a graph database or hosted memory service. This is the idea behind ctx, a Rust CLI that handles the ingestion and searching. We give our agents a skill that te", "published_at": "2026-07-02T15:58:32Z", "score": 8.0, "summary_zh": "开源工具 ctx：将本地 coding agent 记录导入 SQLite 数据库，支持全文搜索和历史回溯。", "category": "产品动态", "tags": "[\"Agent 工作\", \"搜索工具\", \"开发提效\"]", "why_zh": "直接提升 Agent 编程可追溯性，适合演示如何在 Codex 工作流中复用历史会话。", "idea_zh": "选题：实测 ctx 工具，演示如何搜索 Codex/Cursor 等 agent 的历史命令，快速找回解决方案。可做小红书图文教程，标题：“你的 AI 编程记录别浪费！用这个工具一秒搜回历史方案”，社群作业：配置 ctx 并尝试搜索自己的 agent 日志。", "title_zh": ""}, {"url": "https://joeyh.name/blog/entry/no_LLM_code_in_dependencies/", "source": "Hacker News", "title": "No LLM Code in Dependencies", "excerpt": "Article URL: https://joeyh.name/blog/entry/no_LLM_code_in_dependencies/ Comments URL: https://news.ycombinator.com/item?id=48762008 Points: 15 # Comments: 0", "published_at": "2026-07-02T14:17:23Z", "score": 7.0, "summary_zh": "博主主张避免在软件依赖中使用 LLM 生成的代码，引发讨论。", "category": "教程观点", "tags": "[\"LLM代码\", \"依赖安全\", \"AI编程争议\"]", "why_zh": "可讨论 AI 编程的信任度，与 Codex 用户的代码审查习惯相关。", "idea_zh": "小红书图文：AI 写的代码能直接用在项目里吗？3 个检查步骤。可布置社群作业：用 Codex 生成代码并手动审查，分享发现。", "title_zh": ""}, {"url": "https://x.com/shao__meng/status/2072644710523691110", "source": "X：邵猛", "title": "browser-use 发布开源 AI 视频剪辑 Skill「video-use」", "excerpt": "browser-use 团队推出面向 Codex、Claude Code 等 AI 编码智能体的开源 Skill「video-use」，让 LLM 通过 ElevenLabs Scribe 将音频转写为约 12KB 文本（含逐词时间戳、说话人分离、事件标记），仅在决策点调用 timeline_view.py 生成 PNG 帧图。技术流水线包括转写、打包、生成 JSON 格式 EDL、ffmpeg 渲染及最多 3 轮自评估。渲染关键细节：分段提取 + `-c copy` 拼接、30ms 音频淡入淡出、PTS 时移、字幕最后叠加、HDR 自动映射、竖屏缩放、两-pass loudnorm。动画支持 HyperFrames、Remotion、Manim 等引擎。项目附带 12 条硬规则确保生产正确性。", "published_at": "2026-07-02T11:33:01Z", "score": 9.0, "summary_zh": "browser-use 发布开源 AI 视频剪辑 Skill video-use，支持 Codex/Claude Code 自动剪辑", "category": "产品动态", "tags": "[\"Codex Sk\", \"视频剪辑\", \"开源\", \"工作流\"]", "why_zh": "直接赋能 Codex 用户实现 AI 自动剪辑，可制作实测教程", "idea_zh": "立即上手 video-use Skill，教社群用 Codex 完成短视频剪辑全流程，出小红书图文和视频号演示", "title_zh": ""}, {"url": "https://intrepidkarthi.com/writing/the-gauge-broke/", "source": "Hacker News", "title": "The gauge broke: devs felt 20% faster with AI, measured 19% slower", "excerpt": "Article URL: https://intrepidkarthi.com/writing/the-gauge-broke/ Comments URL: https://news.ycombinator.com/item?id=48757440 Points: 15 # Comments: 1", "published_at": "2026-07-02T06:44:34Z", "score": 8.0, "summary_zh": "研究显示开发者使用AI编程时自感快20%，实测却慢19%，感觉与实情反差大。", "category": "论文研究", "tags": "[\"AI编程\", \"效率实测\", \"开发者体验\", \"错觉\"]", "why_zh": "颠覆认知的实测结果，可警示盲目依赖AI，结合Codex做实操指南。", "idea_zh": "小红书图文/视频号口播：标题如'AI写代码真的更快？实测打脸'，用此研究引出实测对比，教普通人如何用Codex真正提效，避免自欺欺人。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-07-01-kimi-k2-7-is-now-available-in-github-copilot/", "source": "Hacker News", "title": "Kimi K2.7 Code is generally available in GitHub Copilot", "excerpt": "Article URL: https://github.blog/changelog/2026-07-01-kimi-k2-7-is-now-available-in-github-copilot/ Comments URL: https://news.ycombinator.com/item?id=48756602 Points: 7 # Comments: 0", "published_at": "2026-07-02T04:32:41Z", "score": 8.0, "summary_zh": "Kimi K2.7 Code 模型现已在 GitHub Copilot 中正式可用。", "category": "产品动态", "tags": "[\"GitHub C\", \"Kimi\", \"代码助手\", \"模型更新\"]", "why_zh": "Kimi 作为新兴模型加入 Copilot，可与 GPT/Codex 做对比评测。", "idea_zh": "实测 Kimi K2.7 Code 在 Copilot 中的表现，与 ChatGPT Codex 对比代码生成、补全等，适合制作小红书图文教程或视频号对比演示。", "title_zh": ""}, {"url": "https://senior-swe-bench.snorkel.ai/", "source": "Hacker News", "title": "Senior SWE-Bench: open-source benchmark that assesses agents as senior engineers", "excerpt": "Article URL: https://senior-swe-bench.snorkel.ai/ Comments URL: https://news.ycombinator.com/item?id=48755928 Points: 3 # Comments: 0", "published_at": "2026-07-02T02:55:16Z", "score": 7.0, "summary_zh": "Senior SWE-Bench开源，评估AI代理作为高级软件工程师的多文件修复等复杂任务能力。", "category": "论文研究", "tags": "[\"AI编程\", \"基准测试\", \"Agent\", \"软件工程\"]", "why_zh": "可作为实测对比Codex/GPT/Claude等高级编程能力的切入点。", "idea_zh": "实测对比Codex、Claude Code、Cursor在Senior SWE-Bench上的表现，分析谁更适合复杂项目，做成小红书图文或视频号避坑指南。", "title_zh": ""}, {"url": "https://x.com/shao__meng/status/2072484635955900792", "source": "X：邵猛", "title": "Emil Kowalski 发布设计工程师 Skills，让 AI 编码工具具备 UI 动画审美", "excerpt": "Emil Kowalski 将多年 UI/动画原则沉淀为三个 Skill，使 Codex、Claude Code、Cursor 等 Coding Agents 具备资深设计工程师的审美判断。核心规则：动画必须有理由；每天 100+ 次的高频操作禁用动画；UI 动画控制在 300ms 内；只动画 transform 和 opacity；入口从 scale（0.95）+opacity：0 开始；尊重 prefers-reduced-motion（仅移除位移动画）。review-animations 以严格标准审查动画代码，输出 Before/After/Why 表格。animation-vocabulary 将模糊描述（如\"弹一下的效果\"）转为专业动效术语。", "published_at": "2026-07-02T00:56:56Z", "score": 9.0, "summary_zh": "Emil Kowalski发布三个UI动画设计Skill，让Codex等Coding Agent自动审查动画代码并输出优化建议。", "category": "教程观点", "tags": "[\"Codex\", \"Skill\", \"UI设计\", \"动画审查\"]", "why_zh": "直接提升Codex输出质量，可做“一行命令让AI帮你审查动画”的实操教程，适合社群分享。", "idea_zh": "实测将Skill安装到Codex，演示对同一段代码审查前后对比，做成小红书图文教程；社群留作业让成员用Skill审查自己项目的动画。", "title_zh": ""}, {"url": "https://twitter.com/Gavriel_Cohen/status/2072432348805669139", "source": "Hacker News", "title": "Fable open sourced NanoClaw's agent factory. It cost $800", "excerpt": "Article URL: https://twitter.com/Gavriel_Cohen/status/2072432348805669139 Comments URL: https://news.ycombinator.com/item?id=48753508 Points: 12 # Comments: 5", "published_at": "2026-07-01T21:41:06Z", "score": 7.0, "summary_zh": "Fable 开源 NanoClaw agent factory，成本 800 美元，可生成多 Agent 协作系统。", "category": "产品动态", "tags": "[\"Agent\", \"开源\", \"工作流\", \"NanoClaw\"]", "why_zh": "开源 Agent 工厂，可实测与 Codex Agent 模式对比，产出低门槛教程。", "idea_zh": "实测 NanoClaw agent factory 与 Codex 的任务委派差异，制作「普通人如何用开源 Agent 搭建工作流」图文/视频，搭配社群作业。", "title_zh": ""}, {"url": "https://github.com/anthropics/claude-code/releases/tag/v2.1.198", "source": "Claude Code：GitHub Rel", "title": "Claude Code v2.1.198 发布", "excerpt": "Claude Code v2.1.198 更新。Claude in Chrome 现已全面可用。为 claude agents 新增后台智能体通知（agent_needs_input / agent_completed）。新增 /dataviz 技能，提供图表与仪表盘设计指导及配色验证器。Gateway 增加 AWS 上的 Claude Platform 作为上游提供商。后台智能体在 worktree 中完成代码后自动提交、推送并创建草稿 PR。内置 Explore 智能体现继承主会话模型（上限 opus）。修复网络短暂断开导致响应中断、后台任务卡在\"Running\"状态、智能体团队队友因 API 错误失败等问题。", "published_at": "2026-07-01T20:45:36Z", "score": 8.0, "summary_zh": "Claude Code更新：新增后台智能体通知、/dataviz技能、自动提交PR，Chrome全面可用。", "category": "产品动态", "tags": "[\"Claude C\", \"Agent通知\", \"数据可视化\"]", "why_zh": "后台Agent自动处理代码并提PR，可对比Codex自动化能力，做差异化实测。", "idea_zh": "短视频演示Claude Code后台Agent全自动修Bug、提PR的流程，对比Codex类似功能，突出各自适用场景。", "title_zh": ""}, {"url": "https://zcode.z.ai/cn", "source": "Hacker News", "title": "ZCode: Claude Code from the Makers of GLM", "excerpt": "Article URL: https://zcode.z.ai/cn Comments URL: https://news.ycombinator.com/item?id=48751752 Points: 10 # Comments: 1", "published_at": "2026-07-01T19:11:38Z", "score": 8.0, "summary_zh": "智谱发布类 Claude Code 工具 ZCode，定位 AI 编程助手。", "category": "产品动态", "tags": "[\"ZCode\", \"AI编程助手\", \"GLM\", \"Claude C\"]", "why_zh": "新竞品可对标 Codex/ChatGPT Codex，适合做对比实测。", "idea_zh": "实测 ZCode 与 Codex/ChatGPT Codex 的代码生成、任务委派等能力差异，做成小红书图文教程或视频号口播，社群可布置动手作业。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-07-01-kimi-k2-7-is-now-available-in-github-copilot", "source": "GitHub Changelog", "title": "Kimi K2.7 Code is generally available in GitHub Copilot", "excerpt": "Kimi K2.7 Code, an open-weight model, is now generally available in GitHub Copilot. This is the first open-weight model offered as a selectable option in the Copilot model picker, giving… The post Kimi K2.7 Code is generally available in GitHub Copilot appeared first on The GitHub Blog .", "published_at": "2026-07-01T19:00:21Z", "score": 8.0, "summary_zh": "GitHub Copilot 首度集成开放权重模型 Kimi K2.7 Code。", "category": "产品动态", "tags": "[\"GitHub C\", \"Kimi\", \"AI编程\", \"开放权重\"]", "why_zh": "Copilot 模型选择变化影响编程工具对比，可策 Codex 竞品分析。", "idea_zh": "对比 Kimi K2.7 Code 与 Codex/GPT 在 Copilot 中的代码建议、上下文理解，测评是否适合非技术用户，输出为社群教程或避坑指南。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-07-01-copilot-vision-is-generally-available", "source": "GitHub Changelog", "title": "Copilot vision is generally available", "excerpt": "Copilot vision is now generally available. You can attach images and PDFs directly to your chat prompts so Copilot can reason about what it sees alongside your code. Supported file… The post Copilot vision is generally available appeared first on The GitHub Blog .", "published_at": "2026-07-01T18:39:17Z", "score": 7.0, "summary_zh": "Copilot视觉功能正式发布，支持在聊天中附加图片和PDF进行推理。", "category": "产品动态", "tags": "[\"Copilot\", \"视觉理解\", \"多模态\", \"编程助手\"]", "why_zh": "可对比ChatGPT Codex的视觉能力，制作竞品评测内容。", "idea_zh": "实测对比：Copilot Vision vs ChatGPT图像理解，看谁更能帮非技术人员快速排查代码截图中的问题。适合小红书图文。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-07-01-browser-tools-for-github-copilot-in-vs-code-are-generally-available", "source": "GitHub Changelog", "title": "Browser tools for GitHub Copilot in VS Code are generally available", "excerpt": "Browser tools for GitHub Copilot in VS Code are now generally available. Agents can now drive a real browser, navigate live web apps, and feed what they find back into… The post Browser tools for GitHub Copilot in VS Code are generally available appeared first on The GitHub Blog .", "published_at": "2026-07-01T16:01:00Z", "score": 7.0, "summary_zh": "GitHub Copilot浏览器工具正式GA，AI编程代理可实际操控浏览器处理网页任务。", "category": "产品动态", "tags": "[\"GitHub C\", \"AI编程\", \"浏览器自动化\", \"竞品对比\"]", "why_zh": "Copilot新浏览器能力可直接与Codex CLI对比，适合做实测教程和答疑。", "idea_zh": "实测Copilot浏览器工具 vs Codex CLI浏览器控制，对比在自动填表、网页数据抓取的表现，做成小红书图文教程或社群作业。", "title_zh": ""}, {"url": "https://x.com/Zai_org/status/2072349453361557898", "source": "X：智谱 Z.ai", "title": "智谱推出GLM-5.2官方开发环境ZCode", "excerpt": "推出 ZCode，GLM-5.2 的官方开发环境 - GLM Coding Plan 订阅用户：现可在 ZCode 获得 1.5 倍使用配额 - 支持 BYOK：可与您现有的订阅和 API 配合使用 - 适用于 macOS、Windows 和 Linux 立即下载：http://zcode.z.ai/en", "published_at": "2026-07-01T15:59:46Z", "score": 7.0, "summary_zh": "智谱推出ZCode，GLM-5.2官方开发环境，支持BYOK，跨平台。", "category": "产品动态", "tags": "[\"智谱\", \"ZCode\", \"GLM-5.2\", \"AI编程\"]", "why_zh": "ZCode作为AI编程助手新选择，可与Codex等工具做对比测评。", "idea_zh": "实测ZCode vs Codex/Cursor：国产编码助手体验如何？适合哪些人？可做视频对比或图文教程。", "title_zh": ""}, {"url": "https://the-decoder.com/openai-paper-reveals-three-gpt-5-6-pro-models-breaking-with-single-top-tier-strategy", "source": "The Decoder：AI News", "title": "OpenAI论文揭示GPT-5.6三个Pro变体，打破单一顶级策略", "excerpt": "OpenAI论文首次列出GPT-5.6的三个Pro变体：Luna Pro、Terra Pro和Sol Pro，取代以往单一Pro模式。在基因组学基准中，Sol Pro通过率31.5%居60个测试模型之首，领先标准Sol（28.7%）和Claude Opus 4.8（16.0%）。Pro相比标准版本提升逐级递减：Luna Pro提升7.1个百分点（16.5%→23.6%），Terra Pro提升5.2（23.3%→28.5%），Sol Pro仅提升2.8（28.7%→31.5%）。Terra Pro（28.5%）几乎与标准Sol（28.7%）持平。论文未披露Pro运行的token用量，也不清楚该分层是否会在ChatGPT中实际推出。", "published_at": "2026-07-01T10:03:41Z", "score": 8.0, "summary_zh": "OpenAI论文披露GPT-5.6三Pro变体，Sol Pro在基因组学基准中领先。", "category": "模型发布", "tags": "[\"GPT-5.6\", \"Pro变体\", \"OpenAI\"]", "why_zh": "GPT-5.6 Pro变体细节曝光，可供社群讨论模型选择与Pro版本价值。", "idea_zh": "GPT-5.6 Pro变体值得买吗？普通用户需要Pro吗？对比标准版与Pro版差距，做社群讨论帖。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/07/441240.html", "source": "量子位", "title": "OpenSquilla 发布 0.4.0：AI 写代码首次能“自我验证”", "excerpt": "OpenSquilla 上线后数周内 GitHub star 增至数千量级", "published_at": "2026-07-01T07:32:23Z", "score": 8.0, "summary_zh": "OpenSquilla 0.4.0发布，AI写代码首次支持自我验证，自动检查并修正生成代码。", "category": "产品动态", "tags": "[\"AI编程\", \"自我验证\", \"OpenSqui\", \"代码生成\"]", "why_zh": "AI编程工具新能力，可与Codex做对比实测，适合产出差异对比教程。", "idea_zh": "实测OpenSquilla自我验证功能，与Codex/Cursor对比代码修复表现，制作教程图文或视频，突出“谁更省心”。", "title_zh": ""}, {"url": "https://twitter.com/anthropicai/status/2072163884430229756", "source": "Hacker News", "title": "Claude Fable 5 available globally tomorrow", "excerpt": "Article URL: https://twitter.com/anthropicai/status/2072163884430229756 Comments URL: https://news.ycombinator.com/item?id=48742236 Points: 19 # Comments: 1", "published_at": "2026-07-01T04:25:50Z", "score": 7.0, "summary_zh": "Anthropic的Claude Fable 5将于明天全球可用。", "category": "产品动态", "tags": "[\"Claude\", \"模型更新\", \"竞品对比\"]", "why_zh": "Claude重大更新，可与GPT直接对比，吸引受众。", "idea_zh": "制作Claude Fable 5与GPT-4o的对比实测，聚焦代码生成、长上下文、工具调用能力，为社群提供选模型建议，适合小红书图文和视频号口播。", "title_zh": ""}, {"url": "https://x.com/shao__meng/status/2072126769986220157", "source": "X：邵猛", "title": "mattpocockuk 的 /writing-great-skills：编写可预测 AI Skill 的指南", "excerpt": "mattpocockuk 的 /writing-great-skills 成为其最常调用的 Skill，指导如何编写稳定可预测的 AI Skill。核心：以过程可预测为目标；区分 model-invoked（自动触发）与 user-invoked（用户调用），description 应作触发器；采用三层信息结构（主步骤、参考、外部文件）实现渐进式披露；每步骤需明确完成标准；拆分 Skill 是为了控制模型注意力；利用 leading word 压缩行为要求。同时诊断五种失败模式：Premature completion、Duplication、Sediment、Sprawl、No-op，并提供 No-op 测试作为判断句子是否有效的标准。", "published_at": "2026-07-01T01:14:54Z", "score": 9.0, "summary_zh": "mattpocockuk发布编写可预测AI Skill指南：描述触发器、三层信息结构等策略。", "category": "教程观点", "tags": "[\"AI Skill\", \"编写指南\", \"Codex\", \"可预测性\"]", "why_zh": "直接对应Codex Skill编写，是实操性极强的教程，可转化为社群教学和作业。", "idea_zh": "Codex Skill怎么写才稳定？解读mattpocockuk指南，出图文教程和社群实战作业，帮学员写出可预测的Skill。", "title_zh": ""}, {"url": "https://mp.weixin.qq.com/s/yLb4T2UC16ebKHApdBbgWw", "source": "公众号：数字生命卡兹克", "title": "Anthropic在Claude Code中植入隐写术代码识别中国用户", "excerpt": "Anthropic在Claude Code中植入隐写术：读取本地时区（Asia/Shanghai或Asia/Urumqi）和ANTHROPIC_BASE_URL环境变量，与一份经base64+XOR（密钥91）加密的147个域名列表（含美团、字节跳动、月之暗面等）比对，识别中国用户。识别后，在请求发送前将系统提示词中日期字符串的单引号（U+0027）替换为其他Unicode字符，连字符改为斜杠，作为2-3比特分类标记传回服务器。该隐蔽行为被社区逆向发现后引发争议，被认为破坏用户信任。", "published_at": "2026-07-01T00:11:00Z", "score": 7.0, "summary_zh": "Anthropic在Claude Code中识别中国用户并隐写标记，引发信任争议。", "category": "产品动态", "tags": "[\"Claude C\", \"隐私争议\", \"信任危机\", \"竞品对比\"]", "why_zh": "竞品隐私争议可做安全对比选题，提醒用户关注工具选型。", "idea_zh": "小红书图文：AI编程助手隐私大对比（Claude Code vs Codex vs Cursor），附避坑与替代方案。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/30/claude-sonnet-5/#atom-everything", "source": "Simon Willison", "title": "What's new in Claude Sonnet 5", "excerpt": "What's new in Claude Sonnet 5 Claude Sonnet 5 came out this morning . I always head straight for the \"what's new\" developer docs because they tend to have more actionable information than the official announcement post. Anthropic say of Sonnet 5 that \"its performance is close to that of Opus 4.8, but at lower prices\". The system card helps explain how they were able to release the model without being blocked by the US government: Sonnet 5 is significantly less capable at cyber tasks than My", "published_at": "2026-06-30T21:23:02Z", "score": 7.0, "summary_zh": "Claude Sonnet 5 发布，性能接近 Opus 4，开发者文档已更新。", "category": "产品动态", "tags": "[\"Claude\", \"模型更新\", \"竞品对比\"]", "why_zh": "可与 Codex 编程能力对比，为社群提供选型参考。", "idea_zh": "实测对比 Claude Sonnet 5 和 ChatGPT Codex 的代码任务表现，做成小红书图文/视频号，帮助不懂技术的创始人决策。", "title_zh": ""}, {"url": "https://artificialanalysis.ai/models/claude-sonnet-5", "source": "Hacker News", "title": "Claude Sonnet 5 – benchmark results", "excerpt": "Article URL: https://artificialanalysis.ai/models/claude-sonnet-5 Comments URL: https://news.ycombinator.com/item?id=48738528 Points: 11 # Comments: 2", "published_at": "2026-06-30T20:09:25Z", "score": 8.0, "summary_zh": "Anthropic 发布 Claude Sonnet 5 基准测试结果，多项指标超越 GPT-4o", "category": "模型发布", "tags": "[\"Claude S\", \"基准测试\", \"模型对比\"]", "why_zh": "Claude Sonnet 5 基准出炉，可与 GPT-4o 直接对比，为普通用户提供选型依据", "idea_zh": "制作「Claude Sonnet 5 vs GPT-4o」实测对比，聚焦日常写作、代码、数据分析，提炼「谁更适合普通人」的结论，适合小红书图文或视频号口播", "title_zh": ""}, {"url": "https://x.com/NotebookLM/status/2072043680442245276", "source": "X：NotebookLM", "title": "NotebookLM Short Video Overviews 全量上线 Web 英文版", "excerpt": "NotebookLM 正式向 Web 英文用户全量推出 Short Video Overviews（短视频概览）功能。该功能可将复杂资料自动转化为 60 秒竖屏视频，深入讲解任意概念。此前，这一功能已面向 Google AI Ultra 和 Pro 订阅者（移动端及 Web）推出，免费用户即将可用。", "published_at": "2026-06-30T19:44:44Z", "score": 7.0, "summary_zh": "NotebookLM上线Short Video Overviews，可自动生成60秒竖屏讲解视频。", "category": "产品动态", "tags": "[\"Notebook\", \"短视频\", \"内容创作\"]", "why_zh": "短视频自动生成功能可为创作者提效，适合社群实战教程。", "idea_zh": "小红书视频实测：NotebookLM一键生成讲解视频，附适用场景与操作教程。", "title_zh": ""}, {"url": "https://www.anthropic.com/news/claude-sonnet-5", "source": "Hacker News", "title": "Claude Sonnet 5", "excerpt": "Article URL: https://www.anthropic.com/news/claude-sonnet-5 Comments URL: https://news.ycombinator.com/item?id=48736605 Points: 57 # Comments: 7", "published_at": "2026-06-30T17:59:52Z", "score": 7.0, "summary_zh": "Anthropic 发布 Claude Sonnet 5，可能在编程和推理上超越前代，引发社区讨论。", "category": "模型发布", "tags": "[\"Claude S\", \"竞品对比\", \"编程能力\"]", "why_zh": "可对比 Codex/ChatGPT，为受众提供选型参考和迁移评估。", "idea_zh": "做一期“Claude Sonnet 5 vs Codex”实操对比，聚焦日常编程任务，做成小红书图文或视频口播，帮助技术小白决策。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-30-claude-sonnet-5-is-generally-available-for-github-copilot", "source": "GitHub Changelog", "title": "Claude Sonnet 5 is generally available for GitHub Copilot", "excerpt": "Claude Sonnet 5 is Anthropic’s latest Sonnet-class model, now available in GitHub Copilot. It brings strong coding performance to everyday development and agentic workflows, giving developers a new Sonnet-class option… The post Claude Sonnet 5 is generally available for GitHub Copilot appeared first on The GitHub Blog .", "published_at": "2026-06-30T17:25:51Z", "score": 8.0, "summary_zh": "Claude Sonnet 5模型在GitHub Copilot全面可用，强化编码与代理工作流。", "category": "产品动态", "tags": "[\"Claude S\", \"GitHub C\", \"AI编程\"]", "why_zh": "Claude Sonnet 5进入Copilot，可与Codex形成对比，适合做竞品实测选题。", "idea_zh": "实测Claude Sonnet 5在Copilot中的代码生成与代理任务能力，与ChatGPT Codex进行差异对比，做成小红书图文教程，指出谁更适合哪个模型，并布置社群对比作业。", "title_zh": ""}, {"url": "https://developers.googleblog.com/announcing-adk-go-20", "source": "Google Developers Blog", "title": "ADK Go 2.0 发布：构建可靠的多智能体应用，新增基于图的工作流引擎、人工参与循环与动态编排", "excerpt": "Agent Development Kit （ADK） for Go 2.0 发布，引入了一类基于图的工作流引擎，用于组合复杂多智能体应用。新版本内置人工参与循环（HITL）编排、使用纯 Go 代码的动态执行、以及指数退避重试等自动弹性特性。统一执行模型后，单智能体应用与复杂图均运行在同一运行时上，简化了遥测与状态持久化。", "published_at": "2026-06-30T17:17:40Z", "score": 7.0, "summary_zh": "Google发布ADK Go 2.0，新增图工作流引擎、人工参与循环与动态编排。", "category": "产品动态", "tags": "[\"多智能体\", \"Google A\", \"工作流引擎\"]", "why_zh": "可与Codex Agent工作流对比，帮助开发者选型。", "idea_zh": "社群教程+小红书：Google ADK 2.0上手实测，对比Codex SDK，多智能体工作流怎么选？", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/30/shot-scraper-video/#atom-everything", "source": "Simon Willison", "title": "Have your agent record video demos of its work with shot-scraper video", "excerpt": "shot-scraper video is a new command introduced in today's shot-scraper 1.10 release which accepts a storyboard.yml file defining a routine to run against a web application and uses Playwright to record a video of that routine. I've written before about the importance of having coding agents produce demos of their work; this is my latest attempt at enabling them to do that. Here's an example video created using shot-scraper video , exercising a still in development feature adding the ability to c", "published_at": "2026-06-30T16:54:26Z", "score": 7.0, "summary_zh": "shot-scraper 1.10 新增 video 命令，可通过 storyboard 文件录制浏览器操作视频。", "category": "产品动态", "tags": "[\"shot-scr\", \"录制视频\", \"浏览器自动化\", \"AI代理\"]", "why_zh": "可结合 Codex 代理录制工作演示，制作实测教程。", "idea_zh": "制作教程：用 Codex 生成 storyboard，shot-scraper 录制网页操作，展示给社群作为自动化案例。适合小红书图文+社群教程。", "title_zh": ""}, {"url": "https://formicarium.es", "source": "Hacker News", "title": "Show HN: My 13-year-old built an ant colony tracker", "excerpt": "He's 13 years old. He wanted to track his own ant colonies — growth, feeding, humidity, and other metrics. He built the whole app himself with some help from AI tools; I just helped him deploy it to a server. Would love to hear your feedback! Comments URL: https://news.ycombinator.com/item?id=48735446 Points: 12 # Comments: 4", "published_at": "2026-06-30T16:48:47Z", "score": 7.0, "summary_zh": "13岁少年用AI工具独立开发蚂蚁群落追踪应用。", "category": "教程观点", "tags": "[\"AI辅助编程\", \"素人开发\", \"动手实测\"]", "why_zh": "可展示普通人用AI编程工具实现创意的完整案例。", "idea_zh": "小红书图文：用ChatGPT/Codex从零做小工具的全过程，重点标出适合不懂代码人群的提问技巧。社群作业：让成员试着用AI生成一个自己的实用小网页。", "title_zh": ""}, {"url": "https://twitter.com/IntCyberDigest/status/2071971609183678544", "source": "Hacker News", "title": "Anthropic has embedded hidden spyware-like code in Claude Code", "excerpt": "Article URL: https://twitter.com/IntCyberDigest/status/2071971609183678544 Comments URL: https://news.ycombinator.com/item?id=48735113 Points: 13 # Comments: 2", "published_at": "2026-06-30T16:29:36Z", "score": 8.0, "summary_zh": "安全研究员指控Anthropic在Claude Code中隐藏间谍代码。", "category": "产品动态", "tags": "[\"代码代理\", \"安全\", \"隐私\", \"Claude\"]", "why_zh": "Claude Code安全争议重大，直接关系代码代理工具信任，可做避坑指南。", "idea_zh": "立即出‘Claude Code有后门？代码代理工具安全怎么查’的紧急测评/分析，小红书图文或视频号口播，社群内讨论替代方案。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-30-copilot-agent-is-now-available-in-jetbrains-ai-assistant", "source": "GitHub Changelog", "title": "Copilot Agent is now available in JetBrains AI Assistant", "excerpt": "Today, JetBrains and GitHub are announcing a deeper integration between JetBrains AI Assistant and GitHub Copilot. Millions of developers already rely on the GitHub Copilot plugin as their AI pair… The post Copilot Agent is now available in JetBrains AI Assistant appeared first on The GitHub Blog .", "published_at": "2026-06-30T16:00:19Z", "score": 7.0, "summary_zh": "GitHub Copilot Agent 模式现已集成至 JetBrains AI Assistant，开发者可在 IDE 中使用代理式编程辅助。", "category": "产品动态", "tags": "[\"Copilot\", \"JetBrain\", \"代码代理\", \"AI 编程\"]", "why_zh": "Copilot Agent 与 GPT/Codex Agent 直接对标，可做差异化对比，帮助普通用户选择适合自己的 AI 编程工具。", "idea_zh": "标题：「Copilot Agent 杀进 JetBrains，跟 Codex 比谁更香？」内容：实测两者在任务委派、自动修 Bug、浏览器控制等场景的体验差异，总结「不会代码的人用哪个更省心」。形式：小红书图文对比 + 视频号口播演示。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/30/shot-scraper/#atom-everything", "source": "Simon Willison", "title": "shot-scraper 1.10", "excerpt": "Release: shot-scraper 1.10 The big new feature is shot-scraper video storyboard.yml , described in detail in Have your agent record video demos of its work with shot-scraper video . Tags: shot-scraper", "published_at": "2026-06-30T15:10:14Z", "score": 7.0, "summary_zh": "shot-scraper 1.10 发布，核心功能为 shot-scraper video 录制浏览器操作。", "category": "产品动态", "tags": "[\"shot-scr\", \"视频录制\", \"浏览器自动化\"]", "why_zh": "可结合 Codex 代理录制工作演示，制作实测教程。", "idea_zh": "实测 shot-scraper video，演示如何让 AI 代理自动生成操作视频，适合社群作业。避免与 Simon 博文重复，可强调版本发布要点。", "title_zh": ""}, {"url": "https://github.com/anthropics/claude-code/issues/62476", "source": "Hacker News", "title": "Beware, Claude Code deletes >30 day old transcripts. Anthropic won't fix it", "excerpt": "Article URL: https://github.com/anthropics/claude-code/issues/62476 Comments URL: https://news.ycombinator.com/item?id=48732846 Points: 15 # Comments: 14", "published_at": "2026-06-30T13:58:21Z", "score": 7.0, "summary_zh": "Claude Code被曝自动删除超过30天的会话记录，Anthropic拒绝修复。", "category": "产品动态", "tags": "[\"Claude C\", \"数据删除\", \"避坑\"]", "why_zh": "可作为与Codex对比的会话管理避坑点，提醒社群成员注意。", "idea_zh": "实测对比Codex与Claude Code的会话保存机制，制作避坑清单，小红书图文呈现“这些AI工具会偷偷删你数据”。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/440599.html", "source": "量子位", "title": "Claude Code之父版「职场MBTI」：AI洗牌后只剩5类人，你选哪种？", "excerpt": "未来是属于这5种职业的", "published_at": "2026-06-30T09:34:05Z", "score": 7.0, "summary_zh": "Claude Code之父预测AI时代只剩5种职业，引发职场MBTI讨论。", "category": "教程观点", "tags": "[\"Claude C\", \"职业分化\", \"AI编程\", \"观点\"]", "why_zh": "Claude Code创始人观点，可做AI工具对一人公司影响的对照选题。", "idea_zh": "小红书图文/视频号口播：拆解Claude Code之父的5类职业，结合Codex/GPT实操，讨论普通人如何借AI工具转型，适合社群引发「未来职业」讨论。", "title_zh": ""}, {"url": "https://openai.com/index/how-chatgpt-adoption-has-expanded", "source": "OpenAI", "title": "How ChatGPT adoption has expanded", "excerpt": "New OpenAI Signals data shows how ChatGPT adoption is growing globally, with users increasing usage, exploring more capabilities, and driving growth across regions and languages.", "published_at": "2026-06-30T09:00:00Z", "score": 7.0, "summary_zh": "OpenAI 内部数据：ChatGPT 全球采用率、使用频率和功能探索均显著增长。", "category": "产品动态", "tags": "[\"ChatGPT\", \"用户增长\", \"采用率\"]", "why_zh": "可了解用户行为趋势，为选题提供数据支撑，启发社群讨论不同阶段的用法。", "idea_zh": "小红书图文「ChatGPT 用户都在偷偷用这些功能」：基于增长数据解读热门功能，社群作业让用户分享自己使用频次最高的功能。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/440162.html", "source": "量子位", "title": "破天荒！DeepSeek V4正式版居然要涨价，而且翻着倍地涨", "excerpt": "好消息是V4正式版能力有升级", "published_at": "2026-06-30T04:08:11Z", "score": 7.0, "summary_zh": "DeepSeek V4正式版价格翻倍上涨，同时能力有所升级。", "category": "产品动态", "tags": "[\"DeepSeek\", \"模型涨价\", \"竞品对比\"]", "why_zh": "直接影响预算有限的个人用户，可与ChatGPT/GPT-4o做性价比实测对比。", "idea_zh": "短视频/图文：「DeepSeek V4涨价后还值得用吗？和ChatGPT比谁更划算？」实测编码、写作等任务，给出替代建议。", "title_zh": ""}, {"url": "https://github.com/SMJAI/open-memory-protocol", "source": "Hacker News", "title": "Open Memory Protocol – One Memory Store for Claude, ChatGPT, Curso", "excerpt": "Article URL: https://github.com/SMJAI/open-memory-protocol Comments URL: https://news.ycombinator.com/item?id=48726966 Points: 4 # Comments: 0", "published_at": "2026-06-30T00:05:33Z", "score": 7.0, "summary_zh": "Open Memory Protocol为Claude、ChatGPT、Cursor等提供统一记忆存储", "category": "产品动态", "tags": "[\"记忆管理\", \"AI工具\", \"开源\"]", "why_zh": "统一记忆方案可能简化多AI工作流，适合社群教程", "idea_zh": "实测协议能否让ChatGPT与Claude共享记忆，制作图文教程“不同AI共享记忆了？”，社群作业：搭建记忆服务并配置到常用工具", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.31", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.31", "excerpt": "Release 0.143.0-alpha.31", "published_at": "2026-06-29T22:35:24Z", "score": 8.0, "summary_zh": "OpenAI Codex rust 发布 v0.143.0-alpha.31 版本更新。", "category": "产品动态", "tags": "[\"Codex更新\", \"Rust\", \"Alpha版本\"]", "why_zh": "Codex 新版本发布，需跟进实测变化，更新社群教程。", "idea_zh": "快速测试新版本的任务委派、终端自动化等核心能力，与旧版对比，制作“Codex 新版本速览”小红书图文或短视频。", "title_zh": ""}, {"url": "https://www.jonathanbeard.io/blog/2026/06/27/the-80-percent-problem.html", "source": "Hacker News", "title": "The 80% Problem: The Last 20% Is Where the Engineer Used to Live", "excerpt": "Article URL: https://www.jonathanbeard.io/blog/2026/06/27/the-80-percent-problem.html Comments URL: https://news.ycombinator.com/item?id=48723276 Points: 10 # Comments: 0", "published_at": "2026-06-29T18:35:23Z", "score": 8.0, "summary_zh": "探讨AI编程工具能快速生成80%代码，但剩余20%的精细工作仍需工程师介入。", "category": "教程观点", "tags": "[\"AI编程\", \"Codex\", \"工作流\", \"提效局限\"]", "why_zh": "直接对应Codex实操中的“最后20%”痛点，可转化为社群案例。", "idea_zh": "小红书图文：《我用Codex写网站，前80%只花10分钟，最后20%花了2小时》——用真实案例展示AI能做什么、哪里必须手动改，给出避坑清单；社群可布置“修复AI生成代码的常见bug”作业。", "title_zh": ""}, {"url": "https://github.com/deepreinforce-ai/Ornith-1", "source": "Hacker News", "title": "Ornith-1.0: self-improving open-source models for agentic coding", "excerpt": "Article URL: https://github.com/deepreinforce-ai/Ornith-1 Comments URL: https://news.ycombinator.com/item?id=48722052 Points: 3 # Comments: 0", "published_at": "2026-06-29T17:16:17Z", "score": 8.0, "summary_zh": "Ornith-1.0：面向自主编程的开源自改进模型发布。", "category": "模型发布", "tags": "[\"AI编程\", \"Agent\", \"开源模型\"]", "why_zh": "开源自主编程模型，可直接对比Codex任务委派与修Bug能力。", "idea_zh": "实测Ornith-1.0 vs Codex：布置相同编码任务，对比成功率、代码质量与易用性；小红书图文展示差异，社群布置作业。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-29-claude-opus-4-8-fast-mode-is-now-in-preview-for-github-copilot", "source": "GitHub Changelog", "title": "Claude Opus 4.8 (fast mode) is now in preview for GitHub Copilot", "excerpt": "Claude Opus 4.8 (fast mode) is now rolling out in preview on GitHub Copilot. Fast mode delivers significantly faster output token speeds while maintaining the same intelligence as Claude Opus… The post Claude Opus 4.8 (fast mode) is now in preview for GitHub Copilot appeared first on The GitHub Blog .", "published_at": "2026-06-29T16:47:16Z", "score": 7.0, "summary_zh": "Claude Opus 4.8 快速模式预览版上线 GitHub Copilot，输出速度提升。", "category": "产品动态", "tags": "[\"Claude O\", \"GitHub C\", \"快速模式\", \"AI编程\"]", "why_zh": "直接对标屿鹿的 Codex 测评，可做速度对比教程，吸引编程助手用户。", "idea_zh": "短视频口播：Claude Opus 4.8 快速模式 vs ChatGPT Codex CLI，谁更懂非技术用户的自然语言指令？附实际任务测试。小红书图文：三分钟看懂新功能与迁移成本。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/80f54d1266b4571ef649e7e5ecc382dd4e670937", "source": "OpenAI Codex Commits", "title": "[codex] Treat max as a first-class reasoning effort (#30467)", "excerpt": "[codex] Treat max as a first-class reasoning effort (#30467) ## Why The Bedrock GPT-5.6 catalog advertises `max`, but Codex treated it as an opaque custom effort. That made the reasoning picker render it as lowercase `max` while known efforts use productized labels. Making `max` a known effort aligns catalog data, parsing, and UI presentation without changing the `max` wire value or persisted representation. ## What changed - Add first-class `ReasoningEffort::Max` parsing and serialization. - Us", "published_at": "2026-06-29T16:38:49Z", "score": 7.0, "summary_zh": "Codex将“max”作为一级推理级别，与GPT-5.6目录对齐。", "category": "产品动态", "tags": "[\"Codex\", \"推理级别\", \"更新\"]", "why_zh": "Codex新特性：明确支持“max”推理级别，影响任务性能。", "idea_zh": "可测试Codex的“max”模式对复杂任务的效果，制作对比图文/视频，面向进阶用户。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/29/ornith/#atom-everything", "source": "Simon Willison", "title": "Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding", "excerpt": "Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding This is an interesting new open weights (MIT licensed) model, the first model release from DeepReinforce. [...] with variants including 9B Dense, 31B Dense, 35B MoE, and 397B MoE. Built on top of pretrained Gemma 4 and Qwen 3.5, it achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks. As far as I can tell the licenses of those underlying models is compatible with being used in this way - Gemma ", "published_at": "2026-06-29T16:17:59Z", "score": 8.0, "summary_zh": "开源模型 Ornith-1.0 发布，基于 Gemma 和 Qwen，专为代理编码设计，有 9B 到 397B 多种规模。", "category": "模型发布", "tags": "[\"Ornith-1\", \"代理编码\", \"开源模型\", \"AI编程\"]", "why_zh": "首个专注自我脚手架代理编码的开源模型，可与 Codex 等对比，适合做成实测选题。", "idea_zh": "实测 Ornith-1.0 在简单编程任务上的表现，对比 Codex/Claude Code，突出各自优劣；做成小红书图文或视频号教程，教社群成员部署试用。", "title_zh": ""}, {"url": "https://htmx.org/essays/working-with-ai/", "source": "Hacker News", "title": "Working With AI: A concrete example", "excerpt": "Article URL: https://htmx.org/essays/working-with-ai/ Comments URL: https://news.ycombinator.com/item?id=48720064 Points: 41 # Comments: 9", "published_at": "2026-06-29T14:53:41Z", "score": 7.0, "summary_zh": "htmx作者分享用AI辅助编程的具体实例，演示大模型加速开发流程。", "category": "教程观点", "tags": "[\"AI编程\", \"工作流\", \"实操案例\"]", "why_zh": "提供了AI辅助编程的实操案例，可转化为普通人能看懂的提效流程。", "idea_zh": "拆解该案例流程，制作图文/视频教程，对比Codex与Claude Code用法，布置社群作业：用AI助手完成一个小项目。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.30", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.30", "excerpt": "Release 0.143.0-alpha.30", "published_at": "2026-06-29T05:17:22Z", "score": 7.0, "summary_zh": "OpenAI Codex 发布 rust-v0.143.0-alpha.30 版本。", "category": "产品动态", "tags": "[\"Codex\", \"版本更新\", \"Alpha\"]", "why_zh": "Codex 新版本可能引入功能变化或 bug 修复，值得关注。", "idea_zh": "如果新版有显著新功能，可做一期“Codex 新版本实测”的短视频或图文；若无明显变化，可简单提醒社群成员更新。", "title_zh": ""}, {"url": "https://github.com/itsthelore/rac-core", "source": "Hacker News", "title": "Lore – give your coding agent the decisions your team made", "excerpt": "Article URL: https://github.com/itsthelore/rac-core Comments URL: https://news.ycombinator.com/item?id=48714880 Points: 7 # Comments: 0", "published_at": "2026-06-29T04:40:35Z", "score": 7.0, "summary_zh": "一个工具让AI编码代理能读取团队架构决策记录，更好地遵循项目规范。", "category": "产品动态", "tags": "[\"AI编程\", \"决策记录\", \"团队协作\"]", "why_zh": "可转化为「让Codex遵守团队代码规范」的实测教程。", "idea_zh": "演示如何将团队架构决策喂给Codex，对比有无规范时的代码输出差异；适合小红书图文「一句话让Codex听你指挥」。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/8dac605901a7a5a91a6a16c381543af070d3c0c6", "source": "OpenAI Codex Commits", "title": "[codex] Restore v1 delegation guidance (#30511)", "excerpt": "[codex] Restore v1 delegation guidance (#30511) ## Summary - restore the v1 clarification that requests for depth, research, or investigation do not authorize subagent spawning - restore guidance for keeping critical-path, urgent, tightly coupled, or difficult work local - update the focused v1 tool-search and spawn-description coverage ## Why PR #27919 simplified the v1 `spawn_agent` prompt by removing its delegation decision guidance. That left the authorization rule intact, but removed the in", "published_at": "2026-06-29T03:34:47Z", "score": 9.0, "summary_zh": "Codex恢复v1委托指导，深度研究类任务不再允许自动生成子代理", "category": "产品动态", "tags": "[\"Codex\", \"Agent协作\", \"任务委派\", \"工作流\"]", "why_zh": "委托规则直接影响Codex多Agent行为，可做实测对比教程", "idea_zh": "对比新规下Codex如何处理深度研究需求，制作「何时能用子代理」避坑指南，发小红书图文并布置社群作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/6b5f5743b3169ef463155241da2bab6888a3cbe4", "source": "OpenAI Codex Commits", "title": "[codex] Use model metadata for skills usage instructions (#29740)", "excerpt": "[codex] Use model metadata for skills usage instructions (#29740) ## Summary - add a false-by-default `include_skills_usage_instructions` model metadata field - enable the field for the bundled `gpt-5.5` model metadata - consume the metadata in both core and extension skill rendering - remove hardcoded legacy-model matching and its marker plumbing", "published_at": "2026-06-29T00:44:36Z", "score": 7.0, "summary_zh": "Codex 提交更新：新增模型元数据字段以包含技能使用说明，并针对 gpt-5.5 启用。", "category": "产品动态", "tags": "[\"Codex\", \"技能系统\", \"模型元数据\", \"GPT-5.5\"]", "why_zh": "Codex 技能渲染机制变化，可能影响社群中技能编写与触发方式。", "idea_zh": "可做一期短视频解析：Codex 这次更新如何让技能指令更适配特定模型？标题拟为「Codex 新更新：技能指令现在能看人下菜碟了？」。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/e428a12d2235fe2bc10b10bc45d245d1f491f3c7", "source": "OpenAI Codex Commits", "title": "[codex] Enable remote plugins by default (#30297)", "excerpt": "[codex] Enable remote plugins by default (#30297) ## Summary - enable the remote plugin feature by default - promote the remote plugin feature from under development to stable - preserve the existing `features.remote_plugin` override for explicitly disabling it - keep legacy disabled-path coverage explicit in TUI and app-server tests ## Impact Remote plugin functionality is enabled by default for configurations that do not set the feature flag. The existing Codex backend authentication gate stil", "published_at": "2026-06-28T18:46:25Z", "score": 9.0, "summary_zh": "Codex 将远程插件功能设为默认启用，从开发阶段升为稳定特性。", "category": "产品动态", "tags": "[\"Codex\", \"远程插件\", \"MCP\", \"安全提示\"]", "why_zh": "直接影响 Codex 用户的使用方式和安全设置，是重大默认行为变更。", "idea_zh": "可做实测对比：开启/关闭远程插件的区别、如何显式禁用、安全注意事项，适合小红书图文和社群避坑指南。", "title_zh": ""}, {"url": "https://12gramsofcarbon.com/p/agentics-tech-things-tokenmaxxing", "source": "Hacker News", "title": "Tokenmaxxing is dead, long live Tokenmaxxing", "excerpt": "Article URL: https://12gramsofcarbon.com/p/agentics-tech-things-tokenmaxxing Comments URL: https://news.ycombinator.com/item?id=48708795 Points: 3 # Comments: 0", "published_at": "2026-06-28T16:24:10Z", "score": 7.0, "summary_zh": "讨论AI代理时代的Token最大化策略演变", "category": "教程观点", "tags": "[\"Token优化\", \"AI代理\", \"工作流\", \"成本控制\"]", "why_zh": "探讨AI代理中Token使用效率，可指导Codex/GPT工作流中的成本与上下文管理", "idea_zh": "提炼Token优化技巧，对比Codex、Claude Code等工具的上下文窗口策略，为社群制作成本控制避坑指南；适合小红书图文清单。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.29", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.29", "excerpt": "Release 0.143.0-alpha.29", "published_at": "2026-06-27T23:43:24Z", "score": 7.0, "summary_zh": "OpenAI Codex 发布 Rust 版本 v0.143.0-alpha.29 更新", "category": "产品动态", "tags": "[\"Codex\", \"更新\", \"AI编程\"]", "why_zh": "Codex 新版本可能带来功能变化，值得实测验证", "idea_zh": "快速检查该版本的更新日志，对比之前版本，制作「Codex 新版本实测」短视频或图文，突出实用变化或隐藏技巧。", "title_zh": ""}, {"url": "https://github.com/kageroumado/adrafinil", "source": "Hacker News", "title": "Show HN: Adrafinil – keep a lid-closed Mac awake only while agents work", "excerpt": "A month ago there was a wave of posts and tweets about engineers walking around cafes and parks with their MacBooks propped half-open, as fully closing the lid forces sleep that stops their AI agents. Some people made snarky comments about using tmux or Amphetamine, and some defended their choice with “but I only need it sometimes, and forgetting to disable Amphetamine and finding my laptop discharged in my bag is worse.” This is a solution to this problem. Unlike caffeinate, it will prevent you", "published_at": "2026-06-27T20:34:53Z", "score": 8.0, "summary_zh": "Adrafinil 让 Mac 合盖后仅当 AI 代理工作时保持唤醒，避免任务中断。", "category": "产品动态", "tags": "[\"工具推荐\", \"Agent 工作\", \"Mac 不间断运\"]", "why_zh": "直接解决 Mac 合盖休眠导致 AI 代理任务中断的常见痛点。", "idea_zh": "实测教程：安装 Adrafinil 并对比 Amphetamine，演示在合盖下运行 Codex Agent 的长时间任务。适合小红书图文或视频号口播，标题突出“合盖也能跑 Agent”。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/439393.html", "source": "量子位", "title": "BrowserBC：克隆人类点击，让一次网页操作转化为所有Agent的能力", "excerpt": "人类一次录制，Agent就能模拟", "published_at": "2026-06-27T12:29:33Z", "score": 8.0, "summary_zh": "BrowserBC可克隆人类点击，将一次网页操作转化为Agent的模拟执行能力。", "category": "产品动态", "tags": "[\"浏览器自动化\", \"Agent\", \"工作流\", \"录屏克隆\"]", "why_zh": "直接对应Agent工作流自动化，可与Codex联动制作实测教程。", "idea_zh": "标题：「不会写代码？让AI学你点网页」BrowserBC+Codex 实操。录制你的重复操作，教AI帮你自动填表、抓数据，社群作业：录制一个自己常做的网页任务并让Agent复现。小红书图文展示步骤，视频号口播对比手动vs自动。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/438895.html", "source": "量子位", "title": "GPT-5.6突然发布！Fable5痛失最强基模王座", "excerpt": "一口气端出三款GPT 5.6系列模型", "published_at": "2026-06-27T01:53:27Z", "score": 8.0, "summary_zh": "第三方基准测试出现GPT-5.6模型，性能超越Fable5等对手，可能为OpenAI新模型。", "category": "模型发布", "tags": "[\"GPT-5.6\", \"基准测试\", \"OpenAI\"]", "why_zh": "OpenAI新模型动向，可做性能对比和实操选题。", "idea_zh": "实测GPT-5.6在文案、翻译、编程等任务中的表现，与GPT-4o对比，分析是否值得升级，适合小红书图文。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/a107b84967eb9a3444fd2d4de03f200337acd52b", "source": "OpenAI Codex Commits", "title": "feat(protocol): define missing rollout turn items (#30282)", "excerpt": "feat(protocol): define missing rollout turn items (#30282) ## Description This PR adds canonical core `TurnItem` shapes for command execution, dynamic tool calls, collab agent tool calls, and sub-agent activity, to be stored in the rollout file soon. It also teaches app-server protocol / `ThreadHistoryBuilder` how to render those items, and adds the small legacy fanout helpers needed for existing event-based consumers. No core producer or rollout persistence behavior changes here, that will be d", "published_at": "2026-06-26T23:44:34Z", "score": 7.0, "summary_zh": "OpenAI Codex 仓库新增协议 TurnItem，定义命令执行、工具调用等协作代理行为。", "category": "产品动态", "tags": "[\"Codex\", \"Agent协作\", \"协议更新\"]", "why_zh": "Codex 底层能力扩展，支持更丰富的多 Agent 与工具交互，预示更强自动化工作流。", "idea_zh": "可做「Codex 悄悄更新了什么」图文解读，点出对普通人未来 Agent 工作流的潜在影响，适合社群预告。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/6509f3148a248f5da6c03cdf008f2ad70526cd31", "source": "OpenAI Codex Commits", "title": "[codex] Support npm marketplace plugin sources (#29375)", "excerpt": "[codex] Support npm marketplace plugin sources (#29375) ## Why Marketplace source deserialization treated `{"source":"npm", ...}` as unsupported. The loader logged and skipped the entry, so npm-backed plugins never appeared in `plugin list --available` and `plugin add` returned "plugin not found". Codex plugins are installed from a plugin root, not from an npm dependency tree. For npm-backed marketplace entries, Codex should fetch the published package contents with", "published_at": "2026-06-26T21:24:46Z", "score": 8.0, "summary_zh": "Codex 支持 npm 市场插件源，此前 npm 插件被忽略，现在可正常列出和添加。", "category": "产品动态", "tags": "[\"Codex\", \"插件工作流\", \"npm\"]", "why_zh": "Codex 新增 npm 插件源支持，可做成实操教程。", "idea_zh": "制作 Codex npm 插件安装与使用评测，对比不同插件源，给社群布置‘安装一个 npm 插件并分享效果’的作业。可做小红书图文/视频号口播。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/69596f0e428ae6698b0b9f1174e05bac0a023ded", "source": "OpenAI Codex Commits", "title": "feat: add GPT-5.6 variants to Bedrock catalog (#30285)", "excerpt": "feat: add GPT-5.6 variants to Bedrock catalog (#30285) ## Summary - add Sol (`openai.gpt-5.6-sol`), Terra (`openai.gpt-5.6-terra`), and Luna (`openai.gpt-5.6-luna`) to the Amazon Bedrock static model catalog - derive all three entries from the bundled GPT-5.5 metadata and add the Bedrock-only `max` reasoning effort - keep the new entries below the current GPT-5.5 and GPT-5.4 models at priorities 2, 3, and 4, preserving GPT-5.5 as the default - add deep-equality coverage for inherited model confi", "published_at": "2026-06-26T20:32:49Z", "score": 9.0, "summary_zh": "OpenAI Codex 仓库提交显示添加 GPT-5.6 系列模型（Sol/Terra/Luna）至 Amazon Bedrock", "category": "模型发布", "tags": "[\"GPT-5.6\", \"OpenAI\", \"模型发布\", \"Amazon B\"]", "why_zh": "透露 OpenAI 即将推出 GPT-5.6 新模型，可能带来能力与价格变化，值得跟踪", "idea_zh": "做“GPT-5.6 前瞻：新模型可能带来哪些能力升级？对普通人意味着什么？” 追踪后续进展，对比现有模型，在社群或小红书出预测与解读", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/ac85409b7b137843e1ec59d9ecaa063e88c60e6c", "source": "OpenAI Codex Commits", "title": "Let Codex consult user-level code-review-* skills. (#30143)", "excerpt": "Let Codex consult user-level code-review-* skills. (#30143) ## Why I use the `$code-review` skill a lot and it'd be nice to add my own additional review criteria in `$CODEX_HOME/skills/code-review-*`. ## What Removes phrasing about "code-review-* skills in this repository" which in practice seems like enough to get Codex to consult my user-level code review skills in addition to the repo-level ones.", "published_at": "2026-06-26T19:36:40Z", "score": 9.0, "summary_zh": "Codex支持用户自定义代码审查技能，路径$CODEX_HOME/skills/code-review-*", "category": "产品动态", "tags": "[\"Codex\", \"Skill\", \"代码审查\", \"自定义\"]", "why_zh": "用户可扩展Codex审查规则，是Skill玩法新进展，适合教普通人定制AI助手", "idea_zh": "实测教程：如何编写、安装、触发自定义审查Skill。对比Claude Code的hooks差异。小红书图文：“给Codex加一条专属审查规则”，社群作业：为自身项目写一条审查Skill", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/79a8ffdbf7ca08820e0b44cab0261eb61ff72465", "source": "OpenAI Codex Commits", "title": "[codex] allow AGENTS.md and skills to authorize delegation (#30274)", "excerpt": "[codex] allow AGENTS.md and skills to authorize delegation (#30274) Prompt update of MAv2 to include agents.md and skills more explicitly should mimic: https://github.com/openai/codex/pull/27919", "published_at": "2026-06-26T19:17:26Z", "score": 8.0, "summary_zh": "Codex 允许 AGENTS.md 和技能文件声明授权委派，强化多代理协作。", "category": "产品动态", "tags": "[\"AGENTS.m\", \"技能\", \"任务委派\", \"Codex工作流\"]", "why_zh": "拓展了技能和代理配置文件对任务委派的授权控制，利于构建多智能体工作流，可直接做成实操教程。", "idea_zh": "实测：如何用 AGENTS.md 和 Skill 让 Codex 自动分配任务？对比 Claude Code 的代理协作差异，适合哪些普通人？制作小红书图文教程，社群布置搭建练习。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/26/openai/#atom-everything", "source": "Simon Willison", "title": "Quoting OpenAI", "excerpt": "We're beginning a limited preview of the GPT‑5.6 series: Sol, our flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable model. Terra has competitive performance to GPT‑5.5 while being 2x cheaper and Luna brings strong capability at our lowest cost. [...] We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. As part of our ongoing engagement with the U.S. government, we previewed our plans and the", "published_at": "2026-06-26T17:10:43Z", "score": 10.0, "summary_zh": "OpenAI 预览 GPT-5.6 系列，包括旗舰 Sol、均衡 Terra 和低成本 Luna，性能更强且更便宜。", "category": "模型发布", "tags": "[\"GPT-5.6\", \"OpenAI 更\", \"模型预览\", \"价格对比\"]", "why_zh": "GPT-5.6 系列发布，包含旗舰和性价比模型，创作者可立即对比实测，判断对提效的价值。", "idea_zh": "实测 GPT-5.6 Sol/Terra/Luna 对比 GPT-5.5 和旧版，做差异评测，适合“普通人是否值得升级”的选题，可出小红书图文或社群教程。", "title_zh": ""}, {"url": "https://openai.com/index/previewing-gpt-5-6-sol/", "source": "Hacker News", "title": "Previewing GPT‑5.6 Sol: a next-generation model", "excerpt": "Article URL: https://openai.com/index/previewing-gpt-5-6-sol/ Comments URL: https://news.ycombinator.com/item?id=48689028 Points: 63 # Comments: 42", "published_at": "2026-06-26T17:06:55Z", "score": 9.0, "summary_zh": "OpenAI 官方预览下一代模型 GPT-5.6 Sol，性能与新功能或大幅提升。", "category": "模型发布", "tags": "[\"GPT模型\", \"OpenAI\", \"新模型发布\"]", "why_zh": "重大模型更新，适合做实测对比、社群教程，吸引目标受众关注。", "idea_zh": "制作“GPT-5.6 Sol 实测”对比图文/视频，展示新能力与旧版差异，设计社群作业让成员体验，突出对普通人的提效价值。", "title_zh": ""}, {"url": "https://www.theverge.com/ai-artificial-intelligence/957845/openai-gpt-5-6-trump-administration-ai-preview", "source": "The Verge AI", "title": "OpenAI unveils GPT-5.6 amid US AI regulatory drama", "excerpt": "Less than 24 hours after news broke that OpenAI would stagger its next model release at the request of the Trump administration, that model, GPT-5.6, is here. On Friday, the company unveiled the limited preview of its new GPT 5.6 model suite: Sol, the flagship; Terra, a medium-tier model for \"high-volume work\"; and Luna, a […]", "published_at": "2026-06-26T17:01:03Z", "score": 10.0, "summary_zh": "OpenAI 在监管压力下发布 GPT-5.6 系列 Sol、Terra、Luna，延续模型迭代节奏。", "category": "模型发布", "tags": "[\"GPT-5.6 \", \"OpenAI\", \"监管\", \"Sol/Terr\"]", "why_zh": "监管背景提供话题性，可结合发布做“合规压力下的模型发布”角度，但核心仍是新模型。", "idea_zh": "做“监管压力下 OpenAI 仓促发布 GPT-5.6？实测对比是否值得用”的选题，突出监管故事和新模型能力，适合视频号口播。", "title_zh": ""}, {"url": "https://github.com/workweave/router", "source": "Hacker News", "title": "Show HN: Smart model routing directly in Claude, Codex and Cursor", "excerpt": "We built a model router that plugs into coding agents (e.g. Claude Code, Codex, Cursor, etc.) and intelligently sends requests to the best model to serve them. Here's a quick demo of running it locally: https://www.youtube.com/watch?v=isKhAyivtfM . At Weave, we write ~all our code with AI, and it's been getting more expensive. This came to a head when Opus 4.7 was released and, thanks to its tokenizer changes, our costs shot up. We knew we didn't need Opus for everything but we didn't want to lo", "published_at": "2026-06-26T16:40:11Z", "score": 9.0, "summary_zh": "开源智能模型路由器，可集成 Claude Code、Codex 和 Cursor，实现请求的最佳模型分发。", "category": "产品动态", "tags": "[\"模型路由\", \"Codex 插件\", \"Agent 编程\", \"多模型协作\"]", "why_zh": "直接连接 Codex 和 Cursor，展示多模型路由在编码代理中的应用，屿鹿可实测并提供工作流教程。", "idea_zh": "实测该路由器在 Codex 和 Cursor 中的效果，对比直接使用单模型的差异，形成“如何让 Codex 自动选择最强模型”的教程，适合社群作业和图文。", "title_zh": ""}, {"url": "https://openai.com/index/previewing-gpt-5-6-sol", "source": "OpenAI", "title": "Previewing GPT-5.6 Sol: a next-generation model", "excerpt": "OpenAI previews GPT-5.6 Sol, a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with its most advanced safety stack.", "published_at": "2026-06-26T10:00:00Z", "score": 10.0, "summary_zh": "OpenAI 正式预览 GPT-5.6 Sol，强化编码、科学和网络安全能力，并配备最强安全措施。", "category": "模型发布", "tags": "[\"GPT-5.6\", \"Sol 模型\", \"编码能力\", \"安全\"]", "why_zh": "官方详细说明 GPT-5.6 编码能力，屿鹿可直接据此制作 Codex 相关课程，探讨新模型对自动编程的提升。", "idea_zh": "结合官方演示，实测 GPT-5.6 Sol 在 Codex 任务中的表现，与 GPT-5 对比，给出“升级是否提升编程体验”的结论，适合社群直播或视频。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/7d8906b4785d9e2d6ebe8c75a163f97de4f7ca35", "source": "OpenAI Codex Commits", "title": "[codex] wire process-owned code mode host into core (#30142)", "excerpt": "[codex] wire process-owned code mode host into core (#30142) ## Summary - add the `code_mode_host` feature flag and select `ProcessOwnedCodeModeSessionProvider` in `CodeModeService` when enabled - initialize code-mode sessions lazily so a missing host reports a tool error without failing thread startup - resolve `codex-code-mode-host` beside the running Codex binary by default while preserving `CODEX_CODE_MODE_HOST_PATH` as an override - add unit and end-to-end coverage for host resolution and g", "published_at": "2026-06-26T07:23:33Z", "score": 7.0, "summary_zh": "Codex仓库新增process-owned代码模式托管功能，支持特性标志与延迟初始化。", "category": "产品动态", "tags": "[\"Codex\", \"代码模式\", \"特性更新\", \"编程助手\"]", "why_zh": "新代码模式可提升编程代理稳定性，后续适合做实测与对比教程。", "idea_zh": "待功能正式上线，实测process-owned代码模式的任务委派与错误处理，与当前模式对比，产出小红书图文/视频教程。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/0d4351c1b80cae3a40a2d0d64d4e87ca1d5c4afa", "source": "OpenAI Codex Commits", "title": "[codex] allow CCA image generation and web search extensions (#29909)", "excerpt": "[codex] allow CCA image generation and web search extensions (#29909) ## Summary - allow the standalone image-generation and web-search extensions for the actor-authorized provider shape used by CCA - preserve builtin `image_generation` and `web_search` for older models and existing flows - keep ordinary non-OpenAI providers excluded from both extensions - remove only the image extension local managed-AuthManager requirement that CCA cannot satisfy - share actor-authorization detection through `", "published_at": "2026-06-26T01:34:35Z", "score": 9.0, "summary_zh": "Codex新增对CCA图片生成和网络搜索扩展的支持。", "category": "产品动态", "tags": "[\"Codex\", \"功能更新\", \"图片生成\", \"网页搜索\"]", "why_zh": "Codex新增图片生成和搜索能力，可实测对比ChatGPT，适合实操测评。", "idea_zh": "实测Codex的图片生成和网页搜索功能，与ChatGPT内置功能对比，分析差异和适用场景，制作视频号口播或小红书图文，社群布置作业：用Codex生成配图并验证搜索准确性。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/ec300bc7bd7355833f579a603df067d59c1963cb", "source": "OpenAI Codex Commits", "title": "Expose MCP app identity in app context (#29934)", "excerpt": "Expose MCP app identity in app context (#29934) ## Why MCP tool-call events need to expose trusted app identity and action metadata directly so v2 clients do not have to infer it from tool names or resource URIs. ## What changed - Add optional `appName`, `templateId`, and `actionName` fields to MCP tool-call `appContext`. - Populate `appName` and `templateId` from trusted Codex Apps metadata, and derive `actionName` from the trusted app resource metadata. - Preserve all three fields through core", "published_at": "2026-06-26T01:31:10Z", "score": 8.0, "summary_zh": "Codex更新，MCP工具调用可暴露应用身份和动作元数据。", "category": "产品动态", "tags": "[\"Codex\", \"MCP\", \"身份识别\"]", "why_zh": "MCP工具身份暴露提升安全性，对搭建工作流有影响，值得关注。", "idea_zh": "解读更新：Codex的MCP现在能识别工具来源，可以写一篇短文讲这对自动化流程有什么好处，并举例说明如何通过模板ID追踪工具调用，发社群日报。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/3095ea9c3d155bfc89197d2628eb818a55c2755d", "source": "OpenAI Codex Commits", "title": "Project selected plugin runtime by environment availability (#30093)", "excerpt": "Project selected plugin runtime by environment availability (#30093) ## Why Selected plugin metadata is stable, but MCP processes are live runtime state. They need different lifetimes: - the MCP extension caches manifest, MCP, and connector declarations for each stable selected root; - each model step projects that cached metadata through the roots that resolved as ready for that exact step; - the MCP manager is rebuilt only when that availability projection changes. This matches executor skills", "published_at": "2026-06-26T00:36:44Z", "score": 8.0, "summary_zh": "Codex 更新 MCP 插件处理逻辑：元数据与运行时状态分离，按环境可用性选择插件。", "category": "产品动态", "tags": "[\"Codex\", \"MCP\", \"插件管理\", \"环境配置\"]", "why_zh": "Codex MCP 插件机制有细节变化，影响工作流配置，可做实测解读。", "idea_zh": "做「Codex MCP 插件环境配置」实测：解释元数据 vs 运行时状态，演示如何在不同环境中启用/禁用插件，社群布置环境切换作业。形式：小红书图文教程/短视频讲解。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/c9e6d9783dfe99a819f00318f753b6f5cbcc38e8", "source": "OpenAI Codex Commits", "title": "Let extensions contribute World State sections (#30100)", "excerpt": "Let extensions contribute World State sections (#30100) ## Why #29856 already owns the durable thread intent and exact environment binding. This PR adds only the small missing extension boundary: an extension can contribute one named World State section, while core still owns persistence, diffing, and model-visible fragment types. This lets skills stay in the skills extension instead of moving their runtime into core. ## Shape ```text extension-owned state | | contribute section id + JSON snapsh", "published_at": "2026-06-25T21:23:51Z", "score": 8.0, "summary_zh": "Codex 开放扩展接口，支持自定义“世界状态”上下文，让 AI 更懂项目背景。", "category": "产品动态", "tags": "[\"Codex\", \"扩展\", \"世界状态\", \"工作流\"]", "why_zh": "能直接做成 Codex 实操教程，教用户定制 AI 编程助手的上下文。", "idea_zh": "小红书图文“Codex 终于能记住你的项目了”：演示如何用新扩展添加项目特定信息，如 API 文档、代码规范，让 Codex 生成代码更贴合；社群布置作业：为自己的项目创建一个世界状态扩展。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.23", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.23", "excerpt": "Release 0.143.0-alpha.23", "published_at": "2026-06-25T20:28:01Z", "score": 7.0, "summary_zh": "OpenAI Codex 发布 Rust 版 v0.143.0-alpha.23。", "category": "模型发布", "tags": "[\"Codex\", \"Rust\", \"Alpha 发布\"]", "why_zh": "Codex 新版本，可关注 Rust 语言支持对开发者的意义。", "idea_zh": "可以做一篇「Codex 支持 Rust 了，对普通用户意味着什么？」的解读，说明 Codex 是什么、Rust 版可能带来的新能力，适合图文或口播。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/e8d4a1a411bbab9be85e704ae3847915916bf347", "source": "OpenAI Codex Commits", "title": "[codex] add current time reminder delivery mode config (#30031)", "excerpt": "[codex] add current time reminder delivery mode config (#30031) ```python delivery_mode = "any_inference" # default delivery_mode = "after_user_or_tool_output" # new mode ``` ## Validation - just test -p codex-core load_config_resolves_current_time_reminder - just test -p codex-core lock_contains_prompts_and_materializes_features", "published_at": "2026-06-25T19:06:43Z", "score": 7.0, "summary_zh": "OpenAI Codex 新增时间提醒传递模式配置，支持在用户或工具输出后触发。", "category": "产品动态", "tags": "[\"Codex\", \"配置更新\", \"时间提醒\", \"工作流\"]", "why_zh": "直接涉及 Codex 功能变化，可讲解新配置对自动化提醒工作流的影响。", "idea_zh": "可以做一期短图文或视频，演示 Codex 新增的 'after_user_or_tool_output' 配置如何让提醒更精准，对比默认模式，给社群布置实操作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.22", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.22", "excerpt": "Release 0.143.0-alpha.22", "published_at": "2026-06-25T17:20:57Z", "score": 8.0, "summary_zh": "Codex Rust 版本发布 v0.143.0-alpha.22，底层更新", "category": "产品动态", "tags": "[\"Codex\", \"Rust\", \"版本更新\"]", "why_zh": "Codex 新版本可能带来能力变化，可做上手实测", "idea_zh": "录制一期“Codex 新版本实测”，对比旧版任务执行、速度、稳定性，做成小红书图文，突出是否值得升级", "title_zh": ""}, {"url": "https://github.com/inkeep/open-knowledge", "source": "Hacker News", "title": "Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion", "excerpt": "Hi HN, Nick here. We’re launching OpenKnowledge ( https://openknowledge.ai/ ), a “what you see is what you get” markdown editor that has direct integrations with Claude, Codex, and Cursor. Available as MacOS app or CLI. Fully free/local and OSS ( https://github.com/inkeep/open-knowledge ). We built this because we wanted a “Google docs” like experience for writing and sharing markdown files across our team. Obsidian is the best alternative we tried, but found it doesn’t have a true “what you see", "published_at": "2026-06-25T16:04:46Z", "score": 7.0, "summary_zh": "开源笔记 OpenKnowledge 集成 Claude、Codex、Cursor，支持 Markdown 编辑，免费本地使用。", "category": "产品动态", "tags": "[\"开源笔记\", \"Codex集成\", \"AI笔记\", \"知识管理\"]", "why_zh": "集成 Codex 的开源笔记工具，可对比 Obsidian，为创作者提供新的知识管理方案。", "idea_zh": "测评 OpenKnowledge：与 Obsidian/Notion 对比，突出 Codex 集成的特色功能，演示如何在笔记中调用 Codex 辅助写作，适合小红书图文对比。", "title_zh": ""}, {"url": "https://news.ycombinator.com/item?id=48673194", "source": "Hacker News", "title": "Tell HN: OpenAI has started putting ads on paid programs", "excerpt": "I was on the £6.99/month program-- mainly because I didnt use it much. Last few days I've started seeing ads: 1. Ad for Financial Times 2. Add for Shein 3. Add for Amazon prime day All 3 were on a chat where I was asking about tips on a mobile game. Needless to say, I cancelled my plan. Im not paying money to see ads Comments URL: https://news.ycombinator.com/item?id=48673194 Points: 13 # Comments: 3", "published_at": "2026-06-25T13:37:36Z", "score": 7.0, "summary_zh": "OpenAI 在 ChatGPT 付费计划中展示广告，用户投诉", "category": "产品动态", "tags": "[\"ChatGPT\", \"广告\", \"付费计划\", \"用户体验\"]", "why_zh": "直接影响 ChatGPT 付费用户的使用体验和信任度，可引发用户讨论是否迁移", "idea_zh": "小红书图文：实测「付费版 ChatGPT 惊现广告」，展示广告位置与内容，对比免费版，讨论是否还值得续费，或推荐无广告替代工具", "title_zh": ""}, {"url": "https://twitter.com/mfts0/status/2070080422482977095", "source": "Hacker News", "title": "You didn't vibe code it, you stole it from open source&enterprise-licensed code", "excerpt": "Article URL: https://twitter.com/mfts0/status/2070080422482977095 Comments URL: https://news.ycombinator.com/item?id=48672328 Points: 33 # Comments: 9", "published_at": "2026-06-25T12:23:05Z", "score": 8.0, "summary_zh": "AI编程面临版权争议：用户被指从开源和企业许可代码中抄袭", "category": "教程观点", "tags": "[\"AI编程\", \"版权风险\", \"Codex\", \"开源代码\"]", "why_zh": "提醒AI编程用户注意代码版权问题，适合做避坑指南", "idea_zh": "做一期短视频：用Codex生成代码时，如何避免无意中侵权？实测演示，给出安全使用建议。适合视频号口播+社群讨论。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/b215961a56b2553a4612a22a812747403c08a58b", "source": "OpenAI Codex Commits", "title": "Support OAuth for HTTP MCP servers from selected executor plugins (#2…", "excerpt": "Support OAuth for HTTP MCP servers from selected executor plugins (#28529) ## Why #28522 routes selected-plugin HTTP MCP traffic through the owning executor, but OAuth bootstrap and refresh still used host-local clients. Executor-only servers therefore cannot complete discovery or login through the same network boundary as the MCP connection. ## What changed - adapt `codex_exec_server::HttpClient` to RMCP 1.8's `OAuthHttpClient` contract - let RMCP own discovery, dynamic registration, PKCE, ", "published_at": "2026-06-25T09:31:17Z", "score": 8.0, "summary_zh": "Codex 为选定执行器插件添加 HTTP MCP 服务器的 OAuth 认证支持", "category": "产品动态", "tags": "[\"Codex 更新\", \"MCP 服务器\", \"OAuth 认证\"]", "why_zh": "Codex 新增 OAuth 能力，普通用户连接第三方 MCP 服务更安全，可做新手教程", "idea_zh": "制作「Codex 如何用 OAuth 连接 MCP 工具」图文教程，展示实际绑定步骤，突出安全性提升，适合小红书和社群作业", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/6368937939dceb07b4a3c47c4448027d0d1a85a6", "source": "OpenAI Codex Commits", "title": "Support HTTP MCP servers from selected executor plugins (#28522)", "excerpt": "Support HTTP MCP servers from selected executor plugins (#28522) ## Why Selected executor plugins can declare both stdio and Streamable HTTP MCP servers, but only stdio registrations were retained. That silently drops part of the plugin's tool surface and prevents HTTP traffic from using the owning executor's network. ## What changed - retain selected-plugin Streamable HTTP MCP declarations alongside stdio declarations - route their HTTP clients through the owning executor environment - ", "published_at": "2026-06-25T09:10:36Z", "score": 9.0, "summary_zh": "Codex 支持从插件加载 HTTP MCP 服务器，扩展 Codex Skill 的工具获取方式。", "category": "产品动态", "tags": "[\"Codex\", \"MCP\", \"HTTP\", \"Skills\"]", "why_zh": "HTTP MCP 支持让 Codex Skill 能调用远程工具，大幅拓宽自动化场景，是重大功能更新。", "idea_zh": "录制实测视频：如何给 Codex 添加 HTTP MCP 服务器，实现联网搜索或内部 API 调用。对比 stdio 与 HTTP 的配置差异和适用场景。小红书图文拆解步骤，社群布置作业：为自己的 Codex 接入一个 HTTP MCP。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/5579792b3b881b44ba42139f9fcd191c4a944128", "source": "OpenAI Codex Commits", "title": "Parallelize environment skill loading (#29990)", "excerpt": "Parallelize environment skill loading (#29990) ## Why Avoid a request waterfall for loading lots of skills at once by hiding latency in concurrent tasks. ## What changed Poll the per-skill parse futures concurrently with an order-preserving stream capped at 64 in-flight loads. Results retain discovery order, and the existing filtering, warnings, and final catalog sorting are unchanged.", "published_at": "2026-06-25T09:02:07Z", "score": 7.0, "summary_zh": "Codex 并行加载环境技能，减少多技能场景下的请求延迟，提升启动速度。", "category": "产品动态", "tags": "[\"Codex\", \"性能优化\", \"技能加载\"]", "why_zh": "对重度 Skill 用户和大型项目有实操价值，能解释为什么 Codex 变快了。", "idea_zh": "社群日报简讯：Codex 技能加载加速，如果你的工作区有很多 Skill 现在启动更丝滑。可附带一个小测试，启用后对比加载时间，适合视频号口播展示。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/cef5444a80ac5a94d435ab780fba5d6f433c504f", "source": "OpenAI Codex Commits", "title": "Report MCP error codes with server attribution (#29969)", "excerpt": "Report MCP error codes with server attribution (#29969) ## Why MCP error-code telemetry special-cased Codex Apps: its reported error codes were retained, while codes from every other MCP server were replaced with `unknown`. Error reporting should behave consistently for every MCP server. The server name already identifies where an error came from, so telemetry does not need a separate Codex Apps classification. This follows up on [#28976](https://github.com/openai/codex/pull/28976), which introd", "published_at": "2026-06-25T04:08:39Z", "score": 7.0, "summary_zh": "Codex修复MCP错误码上报逻辑，所有MCP服务器错误码将一致上报，不再特殊处理Codex Apps。", "category": "产品动态", "tags": "[\"Codex\", \"MCP\", \"错误码\"]", "why_zh": "Codex MCP功能小改进，可作技术更新解读或社群提醒。", "idea_zh": "社群日报速递：Codex调整MCP错误码上报，可能影响部分自建MCP插件的错误处理，提醒用户更新时注意兼容性。可做小红书图文“Codex又悄悄更新了”系列。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/a74771340db6eb81db39e81445a706346f14c139", "source": "OpenAI Codex Commits", "title": "[3/3] core: replay persisted world state (#29837)", "excerpt": "[3/3] core: replay persisted world state (#29837) ## Why Persisting `WorldState` snapshots and patches is only useful if resume and fork restore that exact comparison baseline. Rebuilding it from `TurnContextItem` loses section state and can either repeat or suppress model-visible updates. This is the third PR in the WorldState persistence stack, built on #29835. ## What - Replay full WorldState snapshots and RFC 7386 patches through the existing rollout reconstruction segments. - Discard state ", "published_at": "2026-06-25T03:32:08Z", "score": 7.0, "summary_zh": "Codex 提交增强世界状态持久化，支持任务恢复与分叉。", "category": "产品动态", "tags": "[\"Codex\", \"工作区状态\", \"代理记忆\", \"技术更新\"]", "why_zh": "Codex 工作区状态持久化能力提升，对多步骤代理工作流和记忆管理很有价值，可做成通俗教程。", "idea_zh": "小红书图文：'Codex 终于能记住你之前做到哪了'，用生活化比喻解释持久化状态对多步骤任务的好处。社群布置作业：尝试用 Codex 做一个需要分步执行的项目，测试恢复能力。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/df1199fddb0c41441b7cd5a1f48bc48514a617dd", "source": "OpenAI Codex Commits", "title": "[codex] Add Ultra reasoning effort (#29899)", "excerpt": "[codex] Add Ultra reasoning effort (#29899) ## Why Ultra should be one user-facing reasoning selection for work that benefits from both maximum reasoning and proactive multi-agent delegation. Without it, clients must coordinate maximum reasoning with the experimental `multiAgentMode` setting, even though the inference backend still expects its existing `max` effort value. This change makes reasoning effort the source of truth: clients select `ultra`, core derives proactive multi-agent behavior w", "published_at": "2026-06-25T03:13:52Z", "score": 9.0, "summary_zh": "Codex 新增 Ultra 推理设置，支持最大推理与多 Agent 主动委派", "category": "产品动态", "tags": "[\"Codex\", \"推理模式\", \"Agent协作\"]", "why_zh": "Codex 新推理模式直接面向复杂任务与多 Agent 协作，可做实测对比与教程", "idea_zh": "实测 Codex Ultra 模式：什么时候用？和普通推理差距有多大？适合谁？可做小红书图文对比 / 社群作业", "title_zh": ""}, {"url": "https://news.ycombinator.com/item?id=48668199", "source": "Hacker News", "title": "Ask HN: Where is our profession (programmer) going?", "excerpt": "I had been running a small (3 people) software company for about 4 years. Since closing down, I recently hung out at a friend's company to see what they were working on (15 ppl). To preface: I'm a heavy user of Claude (rarely write code by hand), but what I'm seeing in person has been rather shocking to me, and I wanted to calibrate with others. In particular: - the code is not the source of truth anymore; it's ask claude to write, and ask claude to explain - LoC, abstractions, and all those \"so", "published_at": "2026-06-25T02:41:47Z", "score": 7.0, "summary_zh": "HN 讨论：AI 时代程序员职业去向，重度用 Claude 手写代码变少。", "category": "行业新闻", "tags": "[\"AI编程\", \"程序员前途\", \"Claude\", \"行业讨论\"]", "why_zh": "热门话题反映 AI 工具对编程习惯的重塑，适合引出 ChatGPT/Codex 对非技术人的机会和对比。", "idea_zh": "视频号口播：'我最近在朋友公司看到，程序员都不怎么手写代码了'，结合自身使用 Codex 经历，对比 Claude 和 ChatGPT，讨论 AI 如何赋能普通人。社群发起'你每天靠 AI 写多少内容'的分享。", "title_zh": ""}, {"url": "https://openai.com/index/how-agents-are-transforming-work", "source": "OpenAI", "title": "How agents are transforming work", "excerpt": "A new OpenAI research paper shows how AI agents are transforming work, enabling longer, more complex tasks and expanding productivity across roles.", "published_at": "2026-06-25T02:00:00Z", "score": 8.0, "summary_zh": "OpenAI 论文显示代理可完成更长更复杂的任务，扩大生产力边界。", "category": "论文研究", "tags": "[\"Agent\", \"生产力\", \"OpenAI\"]", "why_zh": "官方背书代理能力升级，可用论文结论指导普通人构建 Codex/GPT 工作流。", "idea_zh": "解读论文核心发现：代理现在擅长多步骤、长时任务，做一份《一人公司代理任务清单》。社群分享：如何将复杂工作拆解给 Codex Agent，小红书图文对比传统方法与代理效率。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.16", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.16", "excerpt": "Release 0.143.0-alpha.16", "published_at": "2026-06-25T01:04:45Z", "score": 9.0, "summary_zh": "OpenAI Codex 发布 rust-v0.143.0-alpha.16 新版本", "category": "产品动态", "tags": "[\"Codex\", \"版本更新\", \"Rust\"]", "why_zh": "Codex 新版本发布，可测试新功能并制作更新解析内容", "idea_zh": "制作 Codex 0.143.0-alpha.16 更新速览：对比上一版本变化，演示新功能，适合社群成员第一时间上手。小红书图文用'更新清单'形式，口播快速过亮点。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/f4e6aa70e55426b39e560e043228c665be092e23", "source": "OpenAI Codex Commits", "title": "feat(remote-control): add daemon pairing command (#29913)", "excerpt": "feat(remote-control): add daemon pairing command (#29913) ## Why Users who run Codex remote control through daemon mode can keep the daemon running, but they do not have a CLI path to mint the short-lived manual pairing code needed to connect another device. Without this command, they need to speak app-server JSON-RPC directly. Related: #25675 ## What Changed - Added `codex remote-control pair`, which connects to the existing daemon control socket and calls `remoteControl/pairing/start` with `ma", "published_at": "2026-06-25T01:00:06Z", "score": 7.0, "summary_zh": "Codex 远程控制新增守护进程配对命令，简化多设备连接", "category": "产品动态", "tags": "[\"Codex\", \"远程控制\", \"配对\"]", "why_zh": "提升 Codex 远程控制体验，可演示多设备协作场景", "idea_zh": "录制短视频：用 Codex 远程控制功能在两台设备间协作，展示新增的配对命令如何简化流程，适合口播展示实际操作。", "title_zh": ""}, {"url": "https://www.jimmont.com/llm-style-token-costs", "source": "Hacker News", "title": "What I'm Finding About LLM Code Style and Token Costs", "excerpt": "Article URL: https://www.jimmont.com/llm-style-token-costs Comments URL: https://news.ycombinator.com/item?id=48667409 Points: 3 # Comments: 2", "published_at": "2026-06-25T00:52:40Z", "score": 7.0, "summary_zh": "博客探讨不同代码风格对 LLM 令牌消耗和成本的影响，提供实测数据对比。", "category": "教程观点", "tags": "[\"代码风格\", \"Token成本\", \"AI编程\"]", "why_zh": "直接关系到使用 Codex/ChatGPT 编程时的成本优化，可做成省钱技巧内容。", "idea_zh": "可实测不同代码风格（注释多少、命名长段）在 Codex 中的 token 消耗差异，做成小红书图文‘省 token 编程技巧’，并对比 GPT-4o 等模型，引出社群实操作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.142.1", "source": "OpenAI Codex Releases", "title": "rust-v0.142.1", "excerpt": "Release 0.142.1", "published_at": "2026-06-24T23:46:27Z", "score": 9.0, "summary_zh": "OpenAI Codex 发布 v0.142.1 版本。", "category": "产品动态", "tags": "[\"Codex\", \"版本更新\", \"AI编程\"]", "why_zh": "Codex 新版本，可能包含功能变化或修复，适合做实测对比和更新解读。", "idea_zh": "即刻上手 Codex v0.142.1，做一期「新版本速览」小红书图文，重点看对日常编程任务的影响，社群可布置更新适配作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/134646eff0dfaa32d736d49dc550cde82728ea9a", "source": "OpenAI Codex Commits", "title": "Isolate curated plugin sync Git environment (#29785)", "excerpt": "Isolate curated plugin sync Git environment (#29785) ## Why Several users have reported data loss from this bug, including tracked files being deleted or replaced and branches appearing to be reset to the curated plugins repository. This can happen during startup, before the model chooses to edit anything. Ambient repository variables such as `GIT_DIR` and `GIT_WORK_TREE` can override the repository selected by `git -C`, redirecting startup sync's `git reset --hard` and `git clean -fdx` into", "published_at": "2026-06-24T23:04:51Z", "score": 8.0, "summary_zh": "Codex修复插件同步的Git环境bug，该bug可能导致文件删除或分支重置", "category": "产品动态", "tags": "[\"Codex\", \"插件\", \"Git\", \"Bug修复\"]", "why_zh": "Codex插件同步的重大修复，涉及用户数据丢失风险，可做紧急避坑指导", "idea_zh": "小红书图文：‘Codex这个新Bug差点清空我的项目！’ 用修复提交做引子，演示如何检查Codex插件同步设置、手动备份Git环境，社群布置‘检查你的Codex插件安全’作业。", "title_zh": ""}, {"url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-computer-use-gemini-3-5-flash/", "source": "Hacker News", "title": "Computer use in Gemini 3.5 Flash", "excerpt": "Article URL: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-computer-use-gemini-3-5-flash/ Comments URL: https://news.ycombinator.com/item?id=48662999 Points: 4 # Comments: 0", "published_at": "2026-06-24T17:21:34Z", "score": 7.0, "summary_zh": "Google 发布 Gemini 3.5 Flash 的计算机使用功能，可操控浏览器等界面。", "category": "产品动态", "tags": "[\"Gemini\", \"计算机使用\", \"竞品对比\", \"AI Agent\"]", "why_zh": "Gemini Flash 新增计算机使用能力，可与 Codex/Operator 对比，适合做实测和差异分析。", "idea_zh": "标题：Gemini Flash 也能操控电脑了？跟 Codex 比谁更适合普通人？内容：实测 Gemini 的 computer use 功能，对比 ChatGPT Operator 或 Codex 的计算机交互，分析易用性、限制、适合场景。形式：视频号口播+小红书图文对比表。社群作业：让社群成员用两个工具完成同一任务并分享体验。", "title_zh": ""}, {"url": "https://deepmind.google/blog/introducing-computer-use-in-gemini-3-5-flash/", "source": "DeepMind", "title": "Introducing computer use in Gemini 3.5 Flash", "excerpt": "", "published_at": "2026-06-24T16:30:01Z", "score": 8.0, "summary_zh": "Gemini 3.5 Flash新增computer use功能，可操控桌面应用。", "category": "模型发布", "tags": "[\"Gemini\", \"Computer\", \"多模态\", \"自动化\"]", "why_zh": "可与OpenAI的Codex或Operator对比，适合做实测教程。", "idea_zh": "选题：Gemini 3.5 Flash computer use vs Codex CLI/Operator：普通人用哪个自动化电脑操作？可做小红书图文或视频号实测对比，社群可布置作业尝试。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/390b73133b0707ce877ec924b0011c06b776b9e8", "source": "OpenAI Codex Commits", "title": "Cache plugin namespace during executor skill discovery (#29831)", "excerpt": "Cache plugin namespace during executor skill discovery (#29831) ## Why Executor skill discovery runs before the remote skills catalog is available. For a remote environment, each `ExecutorFileSystem` operation becomes an exec-server RPC. Previously, every discovered `SKILL.md` independently resolved its plugin namespace by walking its ancestors and probing both supported manifest locations. In the common `plugin/skills/<skill>/SKILL.md` layout, that repeats 8 RPCs per skill even though eve", "published_at": "2026-06-24T16:14:34Z", "score": 7.0, "summary_zh": "Codex 提交缓存插件命名空间，优化 executor 技能发现过程。", "category": "产品动态", "tags": "[\"Codex\", \"技能发现\", \"性能优化\"]", "why_zh": "Codex 仓库更新，可能影响技能加载速度，可追踪后续体验变化。", "idea_zh": "观察技能发现速度是否提升，制作「Codex 更新速递」小红书图文，简要说明优化点与用户感知。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/3694b48a82343d94b58c62fa9af335710c3a6187", "source": "OpenAI Codex Commits", "title": "[codex] show external import result counts (#29567)", "excerpt": "[codex] show external import result counts (#29567) ## What changed - Show per-type import counts in the `/import` review UI and started message. - Render completion results as a multi-line summary with total imported/failed counts and one row per import type. - Add snapshot coverage for the updated review and completion output. <img width="537" height="322" alt="Screenshot 2026-06-23 at 9 41 20 PM" src="https://github.com/user-attachments/assets/166542eb-20", "published_at": "2026-06-24T15:56:57Z", "score": 9.0, "summary_zh": "Codex CLI 新增外部导入计数显示，/import 界面展示各类型导入数量。", "category": "产品动态", "tags": "[\"Codex\", \"CLI更新\", \"导入功能\"]", "why_zh": "Codex CLI 新迭代，导入计数优化可做实测演示。", "idea_zh": "创作小红书图文：Codex /import 新功能实测，对比旧版导入体验，教普通人如何用 Codex 导入外部代码模组，提高复用效率。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/69b76e9d075198b14beff200a2ecf53364cd2c40", "source": "OpenAI Codex Commits", "title": "Use fs/walk for environment skill discovery (#29842)", "excerpt": "Use fs/walk for environment skill discovery (#29842) Stack 2 of 3. Base: #29841. Follow-up: #29844. ## What changes Environment skill discovery currently walks remote filesystems through repeated `readDirectory` and `getMetadata` calls. This switches that scan to the bounded `fs/walk` operation from the base PR. ```text Before: readDirectory(root) -> getMetadata(...) -> readDirectory(child) -> ... After: fs/walk(root, limits) -> filter the result for SKILL.md ``` This makes environme", "published_at": "2026-06-24T15:32:35Z", "score": 9.0, "summary_zh": "Codex 改用 fs/walk 进行环境技能发现，优化远程文件系统扫描。", "category": "产品动态", "tags": "[\"Codex\", \"技能发现\", \"文件系统\", \"工作流优化\"]", "why_zh": "Codex 技能发现机制变化，直接影响技能加载和用户工作流稳定性，可做技术解读。", "idea_zh": "拆解 Codex 技能发现如何从 readDirectory 切换到 fs/walk，对比性能与限界处理，用示意图说明对「技能突然找不到」这类用户痛点的改进，适合小红书图文。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/b4f0f3eff1303ea445a218afb039e0079df0afce", "source": "OpenAI Codex Commits", "title": "Persist agent messages as response items (#29829)", "excerpt": "Persist agent messages as response items (#29829) ## Why Inter-agent messages are recorded in live history as `ResponseItem::AgentMessage`, but rollouts stored `InterAgentCommunication` and rebuilt the response item during resume. This made the rollout differ from the actual Responses history. ## What changed - store the prepared `agent_message` response item directly - keep `trigger_turn` in a small local metadata record for fork truncation - keep reading older `inter_agent_communication` rollo", "published_at": "2026-06-24T14:43:10Z", "score": 7.0, "summary_zh": "Codex 提交：代理消息持久化为响应项，修复回放与实时历史不一致问题。", "category": "产品动态", "tags": "[\"Codex\", \"代理协作\", \"持久化\"]", "why_zh": "Codex 内部改进影响多 agent 协作准确性，可关注后续体验变化。", "idea_zh": "待更新后实测多 agent 回放稳定性，对比改版前后差异，制作教程或小红书图文。", "title_zh": ""}, {"url": "https://haystack.deepset.ai/", "source": "Hacker News", "title": "Haystack: Open-Source AI Framework for Production Ready Agents, RAG", "excerpt": "Article URL: https://haystack.deepset.ai/ Comments URL: https://news.ycombinator.com/item?id=48658095 Points: 3 # Comments: 1", "published_at": "2026-06-24T11:21:32Z", "score": 7.0, "summary_zh": "Haystack：用于构建生产级 AI Agent 和 RAG 的开源框架", "category": "产品动态", "tags": "[\"AI框架\", \"Agent工作流\", \"RAG\", \"开源\"]", "why_zh": "可直接用于搭建基于 GPT 等模型的智能体，契合‘AI 工作流教练’定位，适合做社群实战教程", "idea_zh": "小红书图文：《不用代码也能搭AI助手？Haystack RAG 工作流实测》。拆解如何用 Haystack 连 GPT 做知识库问答，给出模板，布置社群作业：用自家说明书建一个客服 Bot。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/438198.html", "source": "量子位", "title": "“AI填志愿”成高考后新现象，超1400万人在和千问聊高考", "excerpt": "现在仅需几分钟，就能生成一份志愿报告", "published_at": "2026-06-24T10:13:28Z", "score": 7.0, "summary_zh": "高考季超1400万人用通义千问聊志愿填报，AI辅助决策成新现象。", "category": "产品动态", "tags": "[\"通义千问\", \"AI志愿\", \"高考\"]", "why_zh": "可转化为 ChatGPT 志愿填报实操教程，切中普通人用 AI 解决实际问题的场景。", "idea_zh": "小红书图文「用ChatGPT填志愿，我试了这3步，比花钱找人靠谱？」对比千问、ChatGPT效果，给出指令模板，引流社群布置作业「设计自己的志愿咨询Agent」。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/2a320fedb53bc0ce89d2887e5b8c2896bc079ca7", "source": "OpenAI Codex Commits", "title": "[codex] Remove auto-compaction opt-out (#29815)", "excerpt": "[codex] Remove auto-compaction opt-out (#29815) ## Summary - remove the default-on `auto_compaction` feature flag and generated config schema entries - restore unconditional pre-turn, model-switch/hash, and mid-turn automatic compaction - expose `new_context` whenever token-budget tooling is enabled - remove the disabled-auto-compaction integration coverage introduced by #28260 ## Motivation Roll back the internal auto-compaction escape hatch added in #28260. Automatic compaction should no longe", "published_at": "2026-06-24T07:15:04Z", "score": 7.0, "summary_zh": "Codex移除auto-compaction功能标志，恢复自动上下文压缩，影响token预算管理。", "category": "产品动态", "tags": "[\"Codex\", \"自动压缩\", \"Token预算\"]", "why_zh": "自动压缩行为变化直接影响Codex会话长度和成本，可做实测对比。", "idea_zh": "实测Codex自动压缩的触发条件与效果：对比开启/关闭（已移除）时上下文长度与响应质量差异，做成小红书图文教程。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/437734.html", "source": "量子位", "title": "刚刚，Claude Code大升级！卡帕西：LLM第三次变革", "excerpt": "目前A社约65%的产品代码已经由Claude Tag参与完成", "published_at": "2026-06-24T03:56:02Z", "score": 9.0, "summary_zh": "Claude Code升级，被卡帕西称为LLM第三次变革，A社65%代码由AI参与。", "category": "产品动态", "tags": "[\"Claude C\", \"升级\", \"编程助手\", \"竞品对比\"]", "why_zh": "Claude Code作为主要竞品，此次升级可能影响AI编程工具格局，适合做对比测评。", "idea_zh": "制作“Claude Code vs ChatGPT Codex：新升级后谁更强？”的实测图文/视频，对比任务委派、自动修Bug等，布置社群体验作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/ff78e212155af07ec2db2abbcbd037fbc9d0889f", "source": "OpenAI Codex Commits", "title": "[codex] Ignore local curated plugins when remote catalog is active (#…", "excerpt": "[codex] Ignore local curated plugins when remote catalog is active (#29765) ## Summary - suppress configured `openai-curated` plugins when the remote plugin feature is enabled and auth uses the Codex backend - preserve `openai-api-curated` and non-Codex-backend behavior while including remote catalog activation in the plugin load cache key - add core plugin coverage and an app-server integration test for runtime feature enablement ## Why The Codex app enables remote plugins through process-local", "published_at": "2026-06-24T02:51:31Z", "score": 8.0, "summary_zh": "Codex 更新：远程插件启用时自动忽略本地 curated 插件，防止命名冲突。", "category": "产品动态", "tags": "[\"Codex\", \"插件管理\", \"MCP\", \"配置\"]", "why_zh": "影响 Codex 插件加载逻辑，可测试远程/本地 Skill 共存行为，适合做讲解。", "idea_zh": "标题：Codex 远程插件 vs 本地插件，你该用哪个？实测并给出选择指南，发布小红书图文/视频，社群可布置‘配置最优插件环境’作业。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/6f65b9a98c4172fac9ba7200a9c498e91fcb84ec", "source": "OpenAI Codex Commits", "title": "[codex] Update bundled skill installer guidance (#29768)", "excerpt": "[codex] Update bundled skill installer guidance (#29768) ## Summary - Update the bundled skill installer's post-install guidance to say the skill will be available on the user's next turn. - Remove the obsolete instruction to restart Codex. ## Why Codex refreshes its skill catalog between turns. The existing bundled instruction predates that behavior and causes the model to recommend an unnecessary restart. ## Impact Released Codex builds will materialize accurate post-install guidance f", "published_at": "2026-06-24T02:36:17Z", "score": 7.0, "summary_zh": "Codex 更新技能安装引导：安装后无需重启，下次对话即可使用。", "category": "产品动态", "tags": "[\"Codex\", \"Skill安装\", \"使用技巧\"]", "why_zh": "用户体验小改进，易转化为实操小教程。", "idea_zh": "小红书图文：“Codex 装完 Skill 不用重启？新变化实测”，展示安装过程与生效时间。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/c660e2b644a5ca75d9fde9947895679906d29c4d", "source": "OpenAI Codex Commits", "title": "[codex] allow image generation with provider auth (#29513)", "excerpt": "[codex] allow image generation with provider auth (#29513) ## Summary - allow the native Responses API `image_generation` tool when the active provider carries CCA's non-empty `x-openai-actor-authorization` header - preserve the Codex-managed ChatGPT auth path, scoped to providers that actually require OpenAI auth - keep generic custom providers excluded, including when unrelated ChatGPT credentials are cached - retain the existing feature, provider-capability, and image-input-modality gates", "published_at": "2026-06-23T19:40:54Z", "score": 8.0, "summary_zh": "Codex 新增提供商认证的图像生成功能，支持 Responses API 的 image_generation 工具。", "category": "产品动态", "tags": "[\"Codex\", \"图像生成\", \"AI工作流\"]", "why_zh": "Codex 整合图像生成能力，可做实测教程对比原生 ChatGPT。", "idea_zh": "出一期【Codex 图像生成实测】视频：对比 Codex 与 ChatGPT 原生生图，展示鉴权、效果差异，社群布置作业「用 Codex 生成配图」。小红书图文拆解步骤。", "title_zh": ""}, {"url": "https://arxiv.org/abs/2606.13175", "source": "Hacker News", "title": "The End of Code Review: Coding Agents Supersede Human Inspection", "excerpt": "Article URL: https://arxiv.org/abs/2606.13175 Comments URL: https://news.ycombinator.com/item?id=48649183 Points: 5 # Comments: 0", "published_at": "2026-06-23T18:24:16Z", "score": 9.0, "summary_zh": "论文声称编程代理将取代人工代码审查，引发HN热议", "category": "论文研究", "tags": "[\"代码审查\", \"编程代理\", \"Codex\", \"AI取代人工\"]", "why_zh": "直接命中Codex实操对比：是否能做代码审查？适合做实测对比教程", "idea_zh": "视频号口播：用Codex实际审查代码，对比传统review，给出“AI能不能替代人审代码”的结论，附社群作业：让Codex审查你自己的项目", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.143.0-alpha.9", "source": "OpenAI Codex Releases", "title": "rust-v0.143.0-alpha.9", "excerpt": "Release 0.143.0-alpha.9", "published_at": "2026-06-23T18:14:15Z", "score": 8.0, "summary_zh": "OpenAI Codex的Rust客户端发布v0.143.0-alpha.9版本更新。", "category": "产品动态", "tags": "[\"Codex更新\", \"AI编程助手\", \"Rust客户端\"]", "why_zh": "Codex新版发布，可实测功能变化并制作上手教程。", "idea_zh": "对比新旧版本，梳理Codex CLI在Rust端的改进点，实测对普通人编程提效的实际影响，适合做成小红书图文或视频号口播教程。", "title_zh": ""}, {"url": "https://www.anthropic.com/news/introducing-claude-tag", "source": "Hacker News", "title": "Claude Tag", "excerpt": "Article URL: https://www.anthropic.com/news/introducing-claude-tag Comments URL: https://news.ycombinator.com/item?id=48648039 Points: 21 # Comments: 2", "published_at": "2026-06-23T17:09:18Z", "score": 7.0, "summary_zh": "Anthropic 推出 Claude Tag 功能，用于标记和组织对话。", "category": "产品动态", "tags": "[\"Claude\", \"Tag\", \"竞品动态\"]", "why_zh": "Claude 新功能可能影响用户对聊天组织方式的选择，可对比 ChatGPT 类似功能。", "idea_zh": "实测：Claude Tag vs ChatGPT 项目/标签，谁更适合管理多任务工作流？可做小红书图文对比，落到普通人的选择指南。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/437503.html", "source": "量子位", "title": "刚刚，豆包2.1发布！Agent自己跑18个小时搞定芯片设计代码", "excerpt": "编程比肩Opus 4.7", "published_at": "2026-06-23T13:37:17Z", "score": 8.0, "summary_zh": "豆包2.1发布，Agent自主完成芯片设计代码，编程能力媲美Opus 4.7。", "category": "模型发布", "tags": "[\"豆包\", \"AI编程\", \"Agent\", \"竞品对比\"]", "why_zh": "可与Codex做编程对比实测，为社群提供选型参考。", "idea_zh": "实测豆包2.1 vs Codex在自动化编程任务上的效率与质量，制作对比图文或短视频，给出“普通人选哪个更划算”的结论。", "title_zh": ""}, {"url": "https://github.com/openai/codex/commit/49614a0391d83eec442ffeca1d4aa0fdeb119818", "source": "OpenAI Codex Commits", "title": "Namespace multi-agent v2 tools under collaboration (#29067)", "excerpt": "Namespace multi-agent v2 tools under collaboration (#29067) ## Summary Multi-agent v2 tools now use the fixed `collaboration` namespace when namespace tools are available. This keeps the model-visible hint and the actual tool surface aligned around `functions.collaboration.*`, without exposing an unshipped namespace knob to users. The PR also removes the old `features.multi_agent_v2.tool_namespace` config/schema surface, updates the MAv2 test fixtures for namespaced calls, and fixes stale `TurnC", "published_at": "2026-06-23T12:15:20Z", "score": 8.0, "summary_zh": "Codex 多智能体 v2 工具统一使用 'collaboration' 命名空间，确保工具提示与实际接口一致。", "category": "产品动态", "tags": "[\"Codex\", \"多Agent\", \"工作流\", \"命名空间\"]", "why_zh": "Codex 多 Agent 协作机制更新，可关注其如何影响任务委派和工具触发。", "idea_zh": "实测新命名空间下多 Agent 协作稳定性，拍摄对比视频或撰写图文教程，展示用 'collaboration' 规划工作流，布置为社群实践作业。", "title_zh": ""}, {"url": "https://github.com/shumaiOne/shumai", "source": "Hacker News", "title": "Show HN: Shumai – open-source Frame.io alternative for creative work", "excerpt": "Shumai is an open source platform for uploading creative files, managing projects, collecting precise feedback, sharing work, and collaborating with AI agents, all in one simple creative-first workspace. I’ve always liked the product design of Frame.io, and I wanted to build an alternative that feels just as polished, while being open source and easy to self host. You can deploy Shumai with docker compose in just a few minutes. It can also be installed from npm, though you'll need to provide you", "published_at": "2026-06-23T10:02:54Z", "score": 7.0, "summary_zh": "Shumai 开源替代 Frame.io，支持创意文件协作与 AI 代理反馈。", "category": "产品动态", "tags": "[\"开源\", \"协作平台\", \"AI代理\", \"创意工作\"]", "why_zh": "结合 AI 代理的创意协作工具，可探讨如何融入视频剪辑审阅流程。", "idea_zh": "演示 Shumai 配合 AI agent 进行视频反馈，对比 Frame.io，制作视频号口播《剪辑审稿新利器？AI 代理帮你自动收集反馈》，关联屿鹿的剪辑服务。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3865554251584774?f=rss", "source": "36氪", "title": "腾讯企业微信将推出“大圆”AI agent，目前已进入内测阶段", "excerpt": "知情人士表示，在微信的ai内测消息曝出后，企业微信也即将推出名为“大圆”的ai agent，目前已进入内测阶段。值得关注的是，企业微信天然连接微信，这意味着“大圆”未来可能不仅服务企业内部协作，还将进一步切入客户经营场景。（财联社）", "published_at": "2026-06-23T09:43:56Z", "score": 10.0, "summary_zh": "企业微信将推出AI agent“大圆”，已内测，未来可能切入客户经营场景。", "category": "产品动态", "tags": "[\"企业微信\", \"AI Agent\", \"大圆\", \"内测\"]", "why_zh": "企业微信推出AI agent，直接对标个人创作者和中小企业提效，可实测测评。", "idea_zh": "第一时间申请内测，做「大圆」上手实测：对比钉钉/飞书AI，演示如何用企业微信AI agent自动处理客户咨询、生成营销文案，突出「一人公司用一个大圆顶半个团队」的降本增效。小红书图文+视频号口播。", "title_zh": ""}, {"url": "https://news.ycombinator.com/item?id=48642229", "source": "Hacker News", "title": "Gemini models increasingly stucking in thinking loop", "excerpt": "Using Google Antigravity and Google AI Studio, I have noticed that Gemini models, which includes 3.1 Pro and 3.5 Flash, are increasingly stucking in thinking loops. It has become much more common. So I did a 100 tasks test and out of them, 3.5 Flash got stuck in a loop 23 times, while 3.1 Pro was a lower 16 times. Comments URL: https://news.ycombinator.com/item?id=48642229 Points: 3 # Comments: 4", "published_at": "2026-06-23T09:00:28Z", "score": 7.0, "summary_zh": "实测显示 Gemini 3.5 Flash 在 100 次任务中 23 次陷入思考循环，3.1 Pro 也存在问题", "category": "产品动态", "tags": "[\"Gemini\", \"思考循环\", \"模型实测\", \"翻车\"]", "why_zh": "实测揭露 Gemini 可靠性问题，适合做警示或对比内容，吸引关注 AI 工具实效的受众", "idea_zh": "视频号口播：实测Gemini 100次任务23次卡死，对比其他模型稳定性，提醒多模型备份", "title_zh": ""}, {"url": "https://36kr.com/p/3865513055540481?f=rss", "source": "36氪", "title": "氪星晚报｜索尼集团计划近三十年来首次发行美元债券；甲骨文上财年裁员约2.1万；豆包发布2.1 Pro模型", "excerpt": "大公司： 甲骨文上财年裁员约2.1万 甲骨文上财年裁员约2.1万，减员规模大于此前已知水平，其中包括因使用人工智能（AI）而被取消的岗位。甲骨文周一在年度财务监管申报文件中称：“我们在各项业务中采用和部署AI技术，已经导致并且未来可能继续导致员工人数减少。” 英伟达宣布Vera Rubin NVL4系统Q4起供货 6月22日，英伟达宣布推出Vera Rubin超级计算平台，用于支撑气候建模、计算流体力学、能源勘探等高强度HPC与AI融合工作负载。该平台结合Rubin GPU与Vera CPU，并通过NVLink、InfiniBand与液冷架构实现一体化设计，单系统可提供超过7 exaflops级AI算力与约5 PF FP64科学计算能力，并支持最高144 GPU/机架密度。戴尔、HPE、Supermicro等厂商也将推出基于该架构的高密度超算系统，预计2026年Q4起陆续上市。 达美乐提升首席运营官乔丹为CEO 达美乐披萨周一宣布任命首席运营官乔·乔丹（Joe Jordan）为首席执行官，寄希望于他丰富的经验能够带领公司渡过当前充满不确定性的经济环境，尤其是在消费者谨慎消费的情况下。 ", "published_at": "2026-06-23T08:59:08Z", "score": 8.0, "summary_zh": "豆包发布2.1 Pro模型，中文综合能力提升，支持128K上下文。", "category": "模型发布", "tags": "[\"豆包\", \"大模型\", \"2.1 Pro\", \"发布\"]", "why_zh": "字节跳动的豆包新模型，适合做国内大模型实测对比选题。", "idea_zh": "快速实测豆包2.1 Pro：用它写文案、做选题、总结长文，对比Claude/GPT，拍成视频口播（突出哪项任务可替代人工），落点「选对模型省一半成本」。", "title_zh": "豆包发布2.1 Pro模型，中文综合能力再升级"}, {"url": "https://36kr.com/newsflashes/3865468521567236?f=rss", "source": "36氪", "title": "火山引擎谭待：豆包还是会保持免费，专业版办公任务模式搭载2.1Pro模型", "excerpt": "火山引擎force原动力大会期间，总裁谭待谈到豆包收费问题时称，“豆包不是火山的业务范畴，但据我所知，豆包还是会保持免费，而且高质量地服务广大用户。同时它最近会推出面向生产力场景的专业版办公任务模式，搭载的就是我们最新推出的2.1Pro。”小财注：6月3日豆包发布说明称，计划针对专业人群的生产力需求推出豆包专业版，对于广大用户日常使用的豆包功能将保持目前的免费服务。（财联社）", "published_at": "2026-06-23T08:28:52Z", "score": 8.0, "summary_zh": "豆包保持免费，即将推出专业版办公任务模式，搭载2.1Pro模型。", "category": "产品动态", "tags": "[\"豆包\", \"大模型\", \"办公提效\"]", "why_zh": "豆包新模型可实测办公提效效果，契合“一人公司提效”选题。", "idea_zh": "实测豆包2.1Pro办公模式：用写报告、做表格等真实任务对比免费版，突出降本增效，适合小红书图文+视频号口播。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-23-github-copilot-app-support-for-byok", "source": "GitHub Changelog", "title": "GitHub Copilot app support for BYOK", "excerpt": "The GitHub Copilot app now supports bring your own key (BYOK), so you can run agent sessions against your own model providers, including OpenAI, Azure OpenAI, Microsoft Foundry, Anthropic, LM… The post GitHub Copilot app support for BYOK appeared first on The GitHub Blog .", "published_at": "2026-06-23T08:00:02Z", "score": 7.0, "summary_zh": "GitHub Copilot应用支持自带密钥，可连接OpenAI等自有模型。", "category": "产品动态", "tags": "[\"GitHub C\", \"BYOK\"]", "why_zh": "Copilot支持BYOK能与Codex工作流对比，帮助用户选择模型。", "idea_zh": "实测Copilot BYOK vs Codex在不同模型下的编程效果，制作对比表格或短视频，指导社群成员做选择。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/22/glm-5-2-openai-compatible-api-a-hands-on-guide-to-reasoning-effort-function-calling-and-long-context-retrieval/", "source": "MarkTechPost", "title": "GLM-5.2 OpenAI-Compatible API: A Hands-On Guide to Reasoning Effort, Function Calling, and Long-Context Retrieval", "excerpt": "We build a practical GLM-5.2 workflow using its hosted, OpenAI-compatible API instead of running the model locally. We set up multiple providers, load the API key securely, and create a reusable chat wrapper. We then test thinking-effort control, streamed reasoning, function calling, a tool-using agent, structured JSON output, and long-context retrieval. We close with token and cost accounting so every demo stays measurable. The post GLM-5.2 OpenAI-Compatible API: A Hands-On Guide to Reasoning E", "published_at": "2026-06-23T06:35:05Z", "score": 8.0, "summary_zh": "智谱GLM-5.2开放API，支持推理控制、函数调用与长上下文检索。", "category": "模型发布", "tags": "[\"GLM-5.2\", \"函数调用\", \"推理控制\", \"Agent\"]", "why_zh": "新模型API实测教程，直接命中Agent/工作流赛道，可做对比评测吸粉。", "idea_zh": "制作视频号口播，实测GLM-5.2的函数调用与推理模式，对比Claude/GPT在自动化工作流中的表现，突出降本增效。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3865364540331271?f=rss", "source": "36氪", "title": "火山引擎总裁谭待：字节内部已大规模落地AI Coding生产流程", "excerpt": "针对智谱港股走高、Anthropic顶尖模型实现AI Coding突破相关行业问题，火山引擎总裁谭称，AI Coding是大模型泛化能力的核心体现，也是支撑复杂Agent任务的关键能力，当前市场对具备代码生产力的大模型给予高估值具备产业逻辑支撑。他表示，字节新推出的豆包2.1Pro已正式跨过生产级生产力质变点，多项评测稳定超越Anthropic旗下标杆模型，部分指标与海外高阶版本持平，可落地芯片RTL开发等真实工业代码迭代任务，字节内部已大规模落地AI Coding生产流程。（财联社）", "published_at": "2026-06-23T06:13:02Z", "score": 9.0, "summary_zh": "字节豆包2.1Pro在AI Coding评测中超越Anthropic，已内部大规模用于生产，包括芯片RTL开发。", "category": "模型发布", "tags": "[\"豆包2.1Pro\", \"AI Codin\", \"大模型评测\", \"生产级AI\"]", "why_zh": "豆包模型能力跃进，具实测引子，适合做对比测评，吸引老板、技术人关注提效。", "idea_zh": "选题：豆包2.1Pro vs Claude写代码能力实测。拍摄真实编程任务（如写脚本、调试）对比视频，口播评述，强调「一人公司用国产模型提效降本」。落点：程序员或非技术者用AI写代码的可能。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/437381.html", "source": "量子位", "title": "可口可乐的世界杯TVC，居然是prompt生成的", "excerpt": "24小时在线，2秒回话，陪你聊球", "published_at": "2026-06-23T06:10:04Z", "score": 8.0, "summary_zh": "可口可乐2026世界杯广告由AI用prompt生成，展现AI视频工具商业应用潜力。", "category": "行业新闻", "tags": "[\"AI生视频\", \"可口可乐\", \"广告\", \"prompt\"]", "why_zh": "品牌大厂用AI做广告，为创作者提供“AI替代传统拍摄”的强说服力案例。", "idea_zh": "视频选题：《可口可乐世界杯广告竟是用AI生成的？手把手教你用AI做商业TVC》。拆解广告使用的AI工具及prompt技巧，落点：一人公司零成本拍出专业广告，降本增效。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3865258380006404?f=rss", "source": "36氪", "title": "豆包发布2.1 Pro模型", "excerpt": "36氪获悉，6月23日，火山引擎在北京举办2026夏季FORCE原动力大会，集中发布豆包大模型2.1，以及视频、图像、音频等多款新模型，并升级面向Agent的云服务体系。据了解，豆包2.1 Pro在 Coding（编程）、Agent（智能体）、VLM（视觉语言模型）三大核心方向实现能力跃升。", "published_at": "2026-06-23T04:57:04Z", "score": 9.0, "summary_zh": "火山引擎发布豆包2.1 Pro，Coding/Agent/VLM能力跃升", "category": "模型发布", "tags": "[\"豆包\", \"大模型\", \"Agent\", \"火山引擎\"]", "why_zh": "豆包2.1 Pro强化Agent能力，直接关联一人公司提效与AI工作流搭建", "idea_zh": "实测豆包2.1 Pro的Agent功能，设计一个自动生成小红书图文+封面+文案的工作流，对比旧版或其他模型，口播演示或图文拆解，突出「一人顶团队」的提效效果", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3865244034176002?f=rss", "source": "36氪", "title": "英伟达、vivo、红杉中国等公司联合加码B站AI创造公开赛", "excerpt": "36氪获悉，英伟达、智谱、vivo、红杉中国、及真格基金联合加码“build in bilibili · AI创造公开赛”。 智谱与英伟达将分别注入100亿Tokens及NVIDIA RTX AI技术专家与新产品支持，降低创新应用开发门槛，助力灵感走向商业现实；vivo携旗下新折叠旗舰设备及原子工作台进场，提供千万级流量、现金及实物扶持，助推移动端AI终端生态落地。资本端，红杉中国与真格基金也将全面跟进项目落地。", "published_at": "2026-06-23T04:27:03Z", "score": 7.0, "summary_zh": "英伟达、智谱、vivo等联合加码B站AI创造公开赛，注入Tokens、技术及流量扶持。", "category": "产品动态", "tags": "[\"AI创造赛\", \"英伟达\", \"智谱\", \"B站\"]", "why_zh": "比赛提供免费Tokens和技术支持，可为AI创作者提供练手和曝光机会。", "idea_zh": "做一期「如何利用B站AI公开赛免费资源打造爆款内容」教程，实测用英伟达RTX AI和智谱Tokens生成图文/视频，落地提效与涨粉。", "title_zh": ""}, {"url": "https://arxiv.org/abs/2606.16140", "source": "Hacker News", "title": "VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO", "excerpt": "Article URL: https://arxiv.org/abs/2606.16140 Comments URL: https://news.ycombinator.com/item?id=48639240 Points: 4 # Comments: 0", "published_at": "2026-06-23T02:01:25Z", "score": 9.0, "summary_zh": "VibeThinker: 3B模型用新训练方法在推理上击败Opus 4.5。", "category": "论文研究", "tags": "[\"小模型\", \"推理能力\", \"模型训练\", \"降本增效\"]", "why_zh": "小模型击败大模型，适合演示低成本AI推理，契合一人公司提效。", "idea_zh": "制作短视频实测VibeThinker与Opus 4.5在逻辑题、商业分析等场景的表现，突出小模型省钱优势，引导观众思考如何用AI降本增效，结尾引导社群。", "title_zh": ""}, {"url": "https://openai.com/index/daybreak-securing-the-world/", "source": "Hacker News", "title": "OpenAI DayBreak – GPT-5.5-Cyber", "excerpt": "Article URL: https://openai.com/index/daybreak-securing-the-world/ Comments URL: https://news.ycombinator.com/item?id=48639063 Points: 15 # Comments: 0", "published_at": "2026-06-23T01:36:08Z", "score": 9.0, "summary_zh": "OpenAI发布DayBreak项目推出GPT-5.5-Cyber，或为安全领域新模型。", "category": "模型发布", "tags": "[\"OpenAI\", \"GPT-5.5\", \"DayBreak\"]", "why_zh": "若为新模型发布，对AI实操测评极为重要，可第一时间实测出圈。", "idea_zh": "选题：OpenAI新模型DayBreak测评。第一时间上手体验，演示功能，与之前模型对比，做小红书图文+视频号口播，强调对写内容、编程的提效。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/22/porting-moebius/#atom-everything", "source": "Simon Willison", "title": "Porting the Moebius 0.2B image inpainting model to run in the browser with Claude Code", "excerpt": "This morning on Hacker News I saw Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance , describing a small but effective inpainting model - a model where you can mark regions of an image to remove and the model imagines what should fill the space. The released model required PyTorch and NVIDIA CUDA , but since it described itself as 0.2B I decided to try and get it running using WebGPU in a browser. TL;DR: I got it working, and you can try the demo at simonw.github.io", "published_at": "2026-06-22T23:43:51Z", "score": 9.0, "summary_zh": "Moebius 0.2B轻量图像修复模型已移植至浏览器，实现客户端AI修复。", "category": "模型发布", "tags": "[\"AI生图\", \"图像修复\", \"浏览器\", \"轻量模型\"]", "why_zh": "小型模型浏览器运行，无需GPU，适合创作者快速修复图片，可做实测演示。", "idea_zh": "实录：用浏览器版Moebius一键移除照片杂物、修复老照片，对比PS操作，展现AI降本增效。适合小红书图文教程或视频号实操口播，强调“一人公司免费用AI修图”。", "title_zh": ""}, {"url": "https://github.com/openai/codex/releases/tag/rust-v0.142.0", "source": "OpenAI Codex Releases", "title": "0.142.0", "excerpt": "New Features /usage can now show and redeem earned usage-limit reset credits, with confirmation, retry, and refreshed availability states. ( #28154 , #28793 ) /plugins now organizes remote plugins into OpenAI Curated, Workspace, and Shared with me sections, while eligible turns can recommend and install relevant plugins. ( #26703 , #28399 , #28400 , #27704 , #28403 ) Configurable rollout token budgets track usage across agent threads, provide remaining-budget reminders, and abort turns when exha", "published_at": "2026-06-22T22:21:19Z", "score": 9.0, "summary_zh": "Codex 0.142.0 新增 /usage 用量管理、插件分类与推荐安装", "category": "产品动态", "tags": "[\"Codex\", \"插件工作流\", \"用量管理\"]", "why_zh": "直接提供插件管理和额度管理的新功能，可做成普通人能照着操作的教程", "idea_zh": "拍摄视频号口播：演示用 /plugins 整理插件、/usage 兑换额度，对比此功能与 Cursor 的插件管理差异，附带「社群作业：整理你的插件库」", "title_zh": ""}, {"url": "https://unsloth.ai/docs/models/glm-5.2", "source": "Hacker News", "title": "Unsloth GLM-5.2 – How to Run Locally", "excerpt": "Article URL: https://unsloth.ai/docs/models/glm-5.2 Comments URL: https://news.ycombinator.com/item?id=48636377 Points: 4 # Comments: 0", "published_at": "2026-06-22T21:21:01Z", "score": 8.0, "summary_zh": "智谱GLM-5.2模型通过Unsloth本地部署教程，涵盖量化与性能优化。", "category": "教程观点", "tags": "[\"智谱GLM-5.\", \"本地部署\", \"Unsloth\", \"模型优化\"]", "why_zh": "新国产模型GLM-5.2可本地运行，实测教程贴合创作者低成本部署AI需求。", "idea_zh": "录制一期视频或图文，实测GLM-5.2通过Unsloth在本地运行的效果，对比云端API成本和性能，展示一人公司如何用本地模型提效降本。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/22/xai-launches-goal-in-grok-build-adding-long-running-autonomous-execution-with-built-in-verification-for-multi-step-coding-tasks/", "source": "MarkTechPost", "title": "xAI Launches /goal in Grok Build, Adding Long-Running Autonomous Execution With Built-In Verification for Multi-Step Coding Tasks", "excerpt": "xAI introduced /goal in Grok Build, a mode for long-running, autonomous task execution. You hand off one objective, and the agent plans an approach, executes a progress checklist, and verifies the result until the goal completes. The post xAI Launches /goal in Grok Build, Adding Long-Running Autonomous Execution With Built-In Verification for Multi-Step Coding Tasks appeared first on MarkTechPost .", "published_at": "2026-06-22T20:34:48Z", "score": 9.0, "summary_zh": "xAI为Grok推出/goal指令，支持自主执行多步骤编码任务并内置验证。", "category": "产品动态", "tags": "[\"Grok\", \"AI Agent\", \"自主任务执行\", \"xAI\"]", "why_zh": "xAI的Agent新功能，可用于自动化内容工作流，契合创作者提效需求。", "idea_zh": "标题：《实测Grok新指令/goal：AI自主写代码+验证，一人公司开发提效》 形式：小红书图文/视频号口播，展示用/goal自动完成爬虫或报表任务，验证准确度，强调节省时间。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/22/sakana-ai-launches-sakana-fugu-an-orchestration-model-that-routes-tasks-across-a-swappable-pool-of-frontier-llms/", "source": "MarkTechPost", "title": "Sakana AI Launches Sakana Fugu: An Orchestration Model That Routes Tasks Across a Swappable Pool of Frontier LLMs", "excerpt": "Fugu and Fugu Ultra route tasks across a swappable model pool, leading most coding, reasoning, and agentic benchmarks. The post Sakana AI Launches Sakana Fugu: An Orchestration Model That Routes Tasks Across a Swappable Pool of Frontier LLMs appeared first on MarkTechPost .", "published_at": "2026-06-22T18:42:59Z", "score": 8.0, "summary_zh": "Sakana AI发布Fugu和Fugu Ultra，可跨模型池路由任务，在编程、推理和智能体基准测试领先。", "category": "模型发布", "tags": "[\"AI智能体\", \"模型发布\", \"任务路由\", \"Sakana A\"]", "why_zh": "新型任务路由模型Fugu，可跨多个LLM分配任务，提升智能体效率。", "idea_zh": "实测Fugu路由不同LLM执行任务的效果，与单一模型对比，制作小红书图文教程，落点：用AI智能体提效一人公司工作流。", "title_zh": ""}, {"url": "https://www.artificialintelligence-news.com/news/mitigating-vendor-lock-in-sakana-ai-fugu-multi-agent-models/", "source": "AI News", "title": "Mitigating vendor lock-in with Sakana AI Fugu multi-agent models", "excerpt": "Sakana AI launched Fugu to orchestrate multi-agent operations and mitigate single-vendor dependency risks in enterprise deployments. Enterprises face operational vulnerabilities when relying entirely on monolithic AI APIs. Japanese AI firm Sakana AI designed Fugu as a response to these concentration risks by creating an orchestration language model that calls upon a pool of varied models […] The post Mitigating vendor lock-in with Sakana AI Fugu multi-agent models appeared first on AI News", "published_at": "2026-06-22T16:11:21Z", "score": 7.0, "summary_zh": "Sakana AI推出Fugu多智能体模型，旨在减少企业对单一AI API的依赖风险。", "category": "产品动态", "tags": "[\"多智能体\", \"企业AI\", \"去中心化\", \"Agent\"]", "why_zh": "多智能体是当前AI应用热点，可作为实测选题吸引技术创业者。", "idea_zh": "实测Fugu与单模型在内容生产流程中的效率对比，突出多智能体如何帮一人公司降本增效，适合小红书图文 or 视频号口播。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-22-new-features-and-claude-as-agent-provider-preview-in-jetbrains-ides", "source": "GitHub Changelog", "title": "New features and Claude as agent provider preview in JetBrains IDEs", "excerpt": "This update adds support for organization and enterprise agents from GitHub, lets you queue and steer messages in Copilot CLI sessions, introduces a new agent debug logs summary view, and… The post New features and Claude as agent provider preview in JetBrains IDEs appeared first on The GitHub Blog .", "published_at": "2026-06-22T15:34:00Z", "score": 8.0, "summary_zh": "JetBrains Copilot 新增 Claude 代理预览、消息排队与调试日志", "category": "产品动态", "tags": "[\"GitHub C\", \"Claude\", \"JetBrain\", \"对比\"]", "why_zh": "Claude 进入 Copilot 生态，为 Codex 用户提供替代选项，适合做差异对比内容", "idea_zh": "出图文对比：Codex vs Claude 在 Copilot JetBrains 上的体验，突出各自适合哪类任务，引导用户尝试，可布置「请尝试切换模型完成一个任务」的社群作业", "title_zh": ""}, {"url": "https://hustvl.github.io/Moebius/", "source": "Hacker News", "title": "Moebius: 0.2B image inpainting model with 10B-level performance", "excerpt": "Article URL: https://hustvl.github.io/Moebius/ Comments URL: https://news.ycombinator.com/item?id=48630171 Points: 4 # Comments: 1", "published_at": "2026-06-22T13:53:02Z", "score": 9.0, "summary_zh": "Moebius发布0.2B图像修复模型，性能比肩10B，轻量高效。", "category": "模型发布", "tags": "[\"AI生图\", \"图像修复\", \"轻量模型\", \"实测新工具\"]", "why_zh": "可实测演示轻量修图模型，展示对内容创作者的一人团队提效价值。", "idea_zh": "实测Moebius图片去水印、物体消除效果，对比传统PS和重级模型。小红书图文出修复前后对比，视频口播教用，落点“修图不用设计师，这个AI一分钟搞定”。", "title_zh": ""}, {"url": "https://www.artificialintelligence-news.com/news/loreal-maybelline-virtual-try-on-chatgpt/", "source": "AI News", "title": "L’Oréal brings Maybelline virtual try-on to ChatGPT", "excerpt": "L’Oréal has announced a collaboration with OpenAI that will bring Maybelline New York’s virtual makeup try-on feature into ChatGPT. The announcement was made at VivaTech 2026. The partnership covers consumer-facing shopping tools, product discovery, advertising pilots, research, and internal content production. The collaboration also covers L’Oréal’s internal use of AI in research, formulation, content production, […] The post L’Oréal brings Maybelline virtual try-on to ChatGPT appeared fi", "published_at": "2026-06-22T10:00:00Z", "score": 8.0, "summary_zh": "欧莱雅集团与OpenAI合作，将美宝莲虚拟试妆功能引入ChatGPT", "category": "产品动态", "tags": "[\"AI虚拟试妆\", \"ChatGPT新\", \"美妆科技\"]", "why_zh": "ChatGPT接入电商场景，可实测演示AI如何提升消费决策效率，适合内容创作", "idea_zh": "实测ChatGPT的美宝莲虚拟试妆功能，制作“用AI试口红省下专柜钱”选题，小红书图文对比试色，突出省时省钱提效，吸引美妆个人IP关注", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3864063849927939?f=rss", "source": "36氪", "title": "微软再推AI智能助手Copilot强制安装策略", "excerpt": "有用户表示，如今即使用户不主动使用，Microsoft365 Copilot组件也会悄悄地重新出现在用户的办公套件中。微软在其管理后台的更新中确认，Microsoft365 Copilot应用将自动安装到符合条件的、已安装Microsoft365桌面应用的Windows电脑上。该公司表示，如果管理员不希望Microsoft365 Copilot自动出现在其组织内的设备上，则需要“选择退出”。（财联社）", "published_at": "2026-06-22T08:15:02Z", "score": 7.0, "summary_zh": "微软Microsoft365 Copilot将自动安装到Windows电脑，管理员需手动退出。", "category": "产品动态", "tags": "[\"微软\", \"Copilot\", \"自动安装\"]", "why_zh": "强制安装影响广泛，可做关闭教程或讨论AI助手对办公的侵入。", "idea_zh": "视频口播/图文：《微软强迫你用Copilot？教你一键关闭》，演示退出步骤，讨论对一人公司的利弊。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/437317.html", "source": "量子位", "title": "阿里发布视频生成模型HappyHorse 1.1：五大维度全面升级", "excerpt": "", "published_at": "2026-06-22T07:30:07Z", "score": 9.0, "summary_zh": "阿里发布视频生成模型 HappyHorse 1.1，宣称在五大维度全面升级。", "category": "模型发布", "tags": "[\"视频生成\", \"阿里\", \"HappyHor\", \"AI生视频\"]", "why_zh": "视频生成赛道新模型，可直接实测对比，适合展示一人公司用 AI 高效创作视频。", "idea_zh": "制作 HappyHorse 1.1 实测视频：用它生成同一脚本的片段，与可灵、Sora 对比生成速度、画面一致性、运镜控制，以口播+屏幕录制发布视频号，突出“一个人搞定宣传片”的提效场景。", "title_zh": ""}, {"url": "https://techstackups.com/comparisons/glm-5.2-vs-opus/", "source": "Hacker News", "title": "GLM 5.2 vs. Opus", "excerpt": "Article URL: https://techstackups.com/comparisons/glm-5.2-vs-opus/ Comments URL: https://news.ycombinator.com/item?id=48626866 Points: 17 # Comments: 4", "published_at": "2026-06-22T07:22:03Z", "score": 7.0, "summary_zh": "GLM 5.2 与 Opus 模型对比评测，涵盖能力、性能实测。", "category": "模型发布", "tags": "[\"模型对比\", \"GLM 5.2\", \"Opus\", \"实测\"]", "why_zh": "大模型横向对比，可直接转化为实测选题，吸引技术受众。", "idea_zh": "制作小红书图文或视频号口播，实测 GLM 5.2 与 Opus 在文案生成、商业分析等场景的表现，用数据展示降本增效潜力。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3863966325838857?f=rss", "source": "36氪", "title": "阿里巴巴发布视频生成模型HappyHorse 1.1", "excerpt": "36氪获悉，6月22日，阿里巴巴发布视频生成模型HappyHorse 1.1，较1.0版本，在动态表现力、主体一致性、指令遵循、视觉质感和音频能力等维度系统性升级。HappyHorse官网、阿里云百炼和千问云均已接入最新版本。", "published_at": "2026-06-22T06:59:15Z", "score": 9.0, "summary_zh": "阿里发布视频生成模型HappyHorse 1.1，大幅提升动态、一致性和音频能力。", "category": "模型发布", "tags": "[\"视频生成\", \"阿里\", \"模型升级\"]", "why_zh": "最新视频生成模型发布，可做实测对比，直接服务创作者降本增效。", "idea_zh": "实测HappyHorse 1.1：用同一指令对比旧版/其他工具，展示动态与一致性提升。小红书图文出效果对比图+参数，视频号口播讲解亮点与用法，引导关注AI视频创作提效。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/437235.html", "source": "量子位", "title": "入选ECCV 2026！清华开源空间模型打败Gemini：真正的空间智能是在世界变化中持续学习", "excerpt": "120分钟长视频一边看一边记", "published_at": "2026-06-22T03:48:38Z", "score": 9.0, "summary_zh": "清华开源空间模型入选ECCV 2026，可处理120分钟长视频，性能优于Gemini。", "category": "模型发布", "tags": "[\"空间智能\", \"开源模型\", \"长视频理解\", \"ECCV2026\"]", "why_zh": "开源模型可直接用于长视频分析，为创作者提供新的AI提效工具，实测选题价值极高。", "idea_zh": "第一时间实测该模型：用一条120分钟课程或直播录像，对比传统分段理解与全量处理效果，展示如何快速生成摘要和剪辑点，落点在一人高效处理长素材。适合视频号口播和小红书图文拆解。", "title_zh": ""}, {"url": "https://sakana.ai/fugu/", "source": "Hacker News", "title": "Sakana Fugu", "excerpt": "Article URL: https://sakana.ai/fugu/ Comments URL: https://news.ycombinator.com/item?id=48624782 Points: 12 # Comments: 0", "published_at": "2026-06-22T02:08:18Z", "score": 8.0, "summary_zh": "Sakana发布Fugu新AI模型，可能为新一代大模型或进化算法产品", "category": "模型发布", "tags": "[\"Sakana\", \"Fugu\", \"新模型\", \"AI智能体\"]", "why_zh": "Sakana新模型发布，适合实测对比，为创作者提供降本增效工具测评。", "idea_zh": "小红书图文/视频号口播：第一时间实测Fugu模型，与Claude/GPT等对比，展示其在内容创作、工作流等方面的能力，突出提效优势。", "title_zh": ""}, {"url": "https://notes.ansonbiggs.com/youre-probably-using-agent-skills-wrong/", "source": "Hacker News", "title": "You're probably using Agent Skills wrong", "excerpt": "Article URL: https://notes.ansonbiggs.com/youre-probably-using-agent-skills-wrong/ Comments URL: https://news.ycombinator.com/item?id=48624327 Points: 4 # Comments: 1", "published_at": "2026-06-22T01:02:16Z", "score": 8.0, "summary_zh": "多数人错误使用Agent Skills，文章提供正确方法。", "category": "教程观点", "tags": "[\"AI智能体\", \"Agent Sk\", \"工作流\", \"使用技巧\"]", "why_zh": "直接切中Agent Skills使用痛点，可做实测教程。", "idea_zh": "将文中错误用法与正确方法做对比实测，演示如何提升Agent效率，制作小红书图文/视频，落点一人公司提效。", "title_zh": ""}, {"url": "https://status.claude.com/incidents/lv35v0q9nsj2", "source": "Hacker News", "title": "Claude: Elevated Error Rates for Opus 4.8, Opus 4.7, Opus 4.6, and Sonnet 4.6", "excerpt": "Article URL: https://status.claude.com/incidents/lv35v0q9nsj2 Comments URL: https://news.ycombinator.com/item?id=48624153 Points: 6 # Comments: 1", "published_at": "2026-06-22T00:38:50Z", "score": 8.0, "summary_zh": "Claude 状态页显示 Opus 4.8/4.7/4.6 及 Sonnet 4.6 出现错误率升高。", "category": "产品动态", "tags": "[\"Claude\", \"模型故障\", \"AI可用性\"]", "why_zh": "Claude 模型突发高错误率，直接影响用户使用，是实时发生的事件，可做一手体验或避坑指南。", "idea_zh": "快速发布小红书图文/视频号口播，实录 Claude 报错场景，分析可能原因与替代方案，强化社群用户对模型稳定性的关注与应对策略。", "title_zh": ""}, {"url": "https://openai.com/index/codex-maxxing-long-running-work", "source": "OpenAI", "title": "Codex-maxxing for long-running work", "excerpt": "Learn how Jason Liu uses Codex to preserve context, manage complex projects, and help work continue beyond a single prompt.", "published_at": "2026-06-22T00:00:00Z", "score": 7.0, "summary_zh": "OpenAI分享Jason Liu用Codex保持长任务上下文、管理复杂项目的方法。", "category": "教程观点", "tags": "[\"Codex\", \"AI工作流\", \"上下文管理\"]", "why_zh": "Codex长任务管理技巧可转化为一人公司提效的实操教程。", "idea_zh": "口播短视频演示用Codex管理跨日项目，突出“AI当项目经理”的降本增效玩法。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3863099932464386?f=rss", "source": "36氪", "title": "中信建投：国产模型加速迭代，算力景气度持续", "excerpt": "36氪获悉，中信建投研报称，国内模型持续迭代，GLM-5.2、Kimi K2.7 Code强化1M上下文、长程Agent、Agentic Coding和真实工程交付能力，推动国产模型从通用问答转向开发者工具和企业级工作流。Kimi补强国际化运营能力，DeepSeek融资强化头部模型产业化预期，微信AI灰度测试则显示AI入口正从独立App走向超级应用生态，有望提升模型调用频次和推理token消耗。海外方面，Anthropic收入、大客户数和融资规模继续上修，企业工作流和Agentic Coding需求仍处扩张通道。算力侧，AI云厂商backlog增长，GPU租赁价格上行，CPU、DRAM/NAND供给约束扩散，算力、存储和租赁服务景气度有望延续。", "published_at": "2026-06-21T23:40:19Z", "score": 8.0, "summary_zh": "国产模型GLM-5.2与Kimi K2.7 Code强化Agent能力与长上下文", "category": "模型发布", "tags": "[\"国产大模型\", \"Agent\", \"代码能力\", \"长上下文\"]", "why_zh": "国产模型在Agent和长上下文能力提升显著，可实测用于内容提效", "idea_zh": "实测Kimi K2.7 Code的Agentic Coding，演示如何用它自动生成小红书文案或剪辑脚本，拍摄口播视频展示流程。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/21/the-7-types-of-agent-memory-a-technical-guide-for-ai-engineers/", "source": "MarkTechPost", "title": "The 7 Types of Agent Memory: A Technical Guide for AI Engineers", "excerpt": "LLMs are stateless by default. Agent memory fixes that. This guide breaks down all 7 types — working, semantic, episodic, procedural, retrieval, parametric, and prospective. It covers what each stores, where it lives, and when to build it. Includes a comparison table and working Python code. The post The 7 Types of Agent Memory: A Technical Guide for AI Engineers appeared first on MarkTechPost .", "published_at": "2026-06-21T23:12:31Z", "score": 8.0, "summary_zh": "技术指南详解AI Agent七种记忆类型：工作、情景、语义等，附Python代码。", "category": "教程观点", "tags": "[\"AI智能体\", \"记忆管理\", \"Agent开发\"]", "why_zh": "直接命中Agent实战赛道，可产出面向非技术用户的“Agent记忆”通俗教程。", "idea_zh": "制作小红书图文或口播视频“AI Agent记不住你？7种记忆让助手变聪明”，图解每种记忆适合的业务场景，并演示如何给Agent配置记忆，突出“一人公司提效”。", "title_zh": ""}, {"url": "https://www.teachmecoolstuff.com/viewarticle/fine-tuning-a-local-llm-to-categorize-questions", "source": "Hacker News", "title": "Good results fine tuning a local LLM like Qwen 3:0.6B to categorize questions", "excerpt": "Article URL: https://www.teachmecoolstuff.com/viewarticle/fine-tuning-a-local-llm-to-categorize-questions Comments URL: https://news.ycombinator.com/item?id=48623434 Points: 8 # Comments: 0", "published_at": "2026-06-21T22:55:23Z", "score": 9.0, "summary_zh": "用Qwen3:0.6B微调实现问题分类，本地运行效果好，适合轻量自动化任务。", "category": "教程观点", "tags": "[\"本地模型微调\", \"Qwen3\", \"问题分类\", \"AI工作流\"]", "why_zh": "展示本地小模型微调实战，可帮创作者实现内容自动归类，提效显著。", "idea_zh": "实操选题：手把手用Qwen3:0.6B微调分类模型，演示数据准备、训练步骤，落地为小红书选题自动打标签或评论意图识别，降低人工分类成本。", "title_zh": ""}, {"url": "https://www.marble.onl/posts/cancel_claude.html", "source": "Hacker News", "title": "There is minimal downside to switching to open models", "excerpt": "Article URL: https://www.marble.onl/posts/cancel_claude.html Comments URL: https://news.ycombinator.com/item?id=48622518 Points: 5 # Comments: 0", "published_at": "2026-06-21T20:56:01Z", "score": 7.0, "summary_zh": "文章认为切换到开源AI模型几乎没有坏处", "category": "教程观点", "tags": "[\"开源模型\", \"成本\", \"替代\"]", "why_zh": "开源模型可作为降本方案，适合预算有限的一人公司", "idea_zh": "制作对比视频：用开源模型（如Qwen3）替代Claude生成小红书图文，对比质量与成本，突出省钱提效。", "title_zh": ""}, {"url": "https://twitter.com/MattZirwas/status/2068365802491834541", "source": "Hacker News", "title": "I was wrong about the Midjourney ultra-sound scanner", "excerpt": "Article URL: https://twitter.com/MattZirwas/status/2068365802491834541 Comments URL: https://news.ycombinator.com/item?id=48618536 Points: 9 # Comments: 1", "published_at": "2026-06-21T12:55:08Z", "score": 8.0, "summary_zh": "作者反思Midjourney超声扫描工具看法，或涉图像扫描新功能。", "category": "产品动态", "tags": "[\"Midjourn\", \"AI生图\", \"工具实测\", \"反思\"]", "why_zh": "直接涉及Midjourney新功能，可做实测内容展示效果。", "idea_zh": "实测Midjourney超声扫描工具，小红书图文/视频演示，落点AI生图提效玩法。", "title_zh": ""}, {"url": "https://martinfowler.com/articles/reliable-llm-bayer.html", "source": "Hacker News", "title": "Building Reliable Agentic AI Systems", "excerpt": "Article URL: https://martinfowler.com/articles/reliable-llm-bayer.html Comments URL: https://news.ycombinator.com/item?id=48615680 Points: 6 # Comments: 0", "published_at": "2026-06-21T04:28:39Z", "score": 7.0, "summary_zh": "Martin Fowler文章探讨构建可靠AI智能体系统的原则与实践", "category": "教程观点", "tags": "[\"AI智能体\", \"可靠性\", \"系统设计\"]", "why_zh": "与智能体赛道高度相关，可提炼实操技巧", "idea_zh": "可做一期视频解读文章要点，结合实测演示如何提升Agent稳定性，适合视频号口播", "title_zh": ""}, {"url": "https://github.com/nikitadoudikov/claude-pulse", "source": "Hacker News", "title": "Show HN: Pulse – Dashboard for Claude Code, approve tool calls from your phone", "excerpt": "Hi everyone, I'm a student from Flanders and I like to use Claude Code for my purposes, ideas and also just for fun and also make solutions for problems in our world!) So that's why I built \"Pulse\", it's an local application that you can easily install to your device and easily follow what your claude agent is doing right now in your terminal session with an ambiance design and easily give permissions for your agent. For those who wants to see directly how much tokens you spent, and how much the", "published_at": "2026-06-20T20:46:50Z", "score": 8.0, "summary_zh": "Pulse仪表盘可通过手机远程审批Claude Code工具调用", "category": "产品动态", "tags": "[\"Claude\", \"Agent工具\", \"远程审批\"]", "why_zh": "一人公司用AI编程提效的新工具，可做手机控制实操演示", "idea_zh": "拍摄Pulse实测视频：从手机审批Claude Code的工具调用，展示如何用AI编程提效，落点一人公司高效开发。适合视频号口播+小红书图文教程。", "title_zh": ""}, {"url": "https://www.theverge.com/ai-artificial-intelligence/950844/vibe-coding-security-risks-apps", "source": "The Verge AI", "title": "Read this before you vibe-code another app", "excerpt": "Bob Starr was delighted with his vibe-coded website. \"Boomberg\" showed how much US tax money is going to tech companies, and Starr launched it online immediately after making it. It wasn't until months after the site went live that he realized there was a problem: a hidden SQL injection risk. It could've left the site […]", "published_at": "2026-06-20T16:31:42Z", "score": 8.0, "summary_zh": "Vibe coding网站上线数月后才发现SQL注入漏洞，警示非技术创业者防坑", "category": "教程观点", "tags": "[\"Vibe cod\", \"AI安全\", \"SQL注入\", \"非技术创业\"]", "why_zh": "用真实案例讲清AI写代码的安全盲区，直击不懂技术的创业者痛点", "idea_zh": "做一期「用AI写代码，这个致命坑九成人都不知道」口播，拆解SQL注入原理及简易检测方法，推荐无代码安全扫描工具，强调“提效不降安”", "title_zh": ""}, {"url": "https://ianbarber.blog/2026/06/19/llms-are-complicated-now/", "source": "Hacker News", "title": "LLMs Are Complicated Now", "excerpt": "Article URL: https://ianbarber.blog/2026/06/19/llms-are-complicated-now/ Comments URL: https://news.ycombinator.com/item?id=48605355 Points: 17 # Comments: 1", "published_at": "2026-06-20T01:25:14Z", "score": 7.0, "summary_zh": "深度分析大语言模型变得复杂的原因及应对策略，涉及提示词与使用技巧。", "category": "教程观点", "tags": "[\"LLM\", \"提示工程\", \"使用技巧\", \"复杂性\"]", "why_zh": "深度解析LLM使用中的复杂性，可直接转化为提示词优化和实操避坑内容。", "idea_zh": "可做“AI工具越来越难用？”选题，实测Claude/GPT等新功能，对比简洁用法与复杂设置的效果，给出降本增效的实用提示词模板，口播或图文。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/19/sean-lynch/#atom-everything", "source": "Simon Willison", "title": "Quoting Sean Lynch", "excerpt": "The real valuable capability MCP offers over skills/CLI is isolating the auth flow outside of the agent’s context window, and potentially out of the harness completely. [...] Maybe the idealized form of MCP is just an auth gateway for the API and nothing else. That’d still be a win. — Sean Lynch , comment on Hacker News Tags: model-context-protocol , llms , ai , generative-ai , skills", "published_at": "2026-06-19T22:45:49Z", "score": 7.0, "summary_zh": "MCP的真正价值在于隔离认证流程，使AI代理安全调用外部工具。", "category": "教程观点", "tags": "[\"MCP\", \"AI代理\", \"认证隔离\", \"工作流\"]", "why_zh": "揭示MCP核心价值，对AI智能体工作流设计有启发。", "idea_zh": "制作视频或图文解释MCP如何让AI代理安全连接外部工具，结合案例展示认证隔离好处，落脚到'一人公司用AI整合服务'。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/19/vibethinker-3b-a-3b-dense-reasoning-model-built-on-qwen2-5-coder-3b-with-the-spectrum-to-signal-post-training-pipeline/", "source": "MarkTechPost", "title": "VibeThinker-3B: A 3B Dense Reasoning Model Built on Qwen2.5-Coder-3B With the Spectrum-to-Signal Post-Training Pipeline", "excerpt": "VibeThinker-3B, a 3B MIT-licensed reasoning model matching DeepSeek V3.2 and Kimi K2.5 on verifiable benchmarks. The post VibeThinker-3B: A 3B Dense Reasoning Model Built on Qwen2.5-Coder-3B With the Spectrum-to-Signal Post-Training Pipeline appeared first on MarkTechPost .", "published_at": "2026-06-19T22:06:31Z", "score": 8.0, "summary_zh": "VibeThinker-3B: 3B参数推理模型开源，性能比肩DeepSeek V3.2和Kimi K2.5。", "category": "模型发布", "tags": "[\"小模型\", \"推理模型\", \"VibeThin\", \"开源\"]", "why_zh": "小模型性能强劲，适合一人公司低成本部署，可做实测内容。", "idea_zh": "实测VibeThinker-3B与DeepSeek等大模型在推理任务上的表现，突出“3B战胜大模型”的降本增效，适合小红书图文或视频号。", "title_zh": ""}, {"url": "https://arrowtsx.dev/bigger-models/", "source": "Hacker News", "title": "GPT-5.5 hallucinates 3x more than MIT-licensed GLM-5.2", "excerpt": "Article URL: https://arrowtsx.dev/bigger-models/ Comments URL: https://news.ycombinator.com/item?id=48600167 Points: 62 # Comments: 14", "published_at": "2026-06-19T16:11:25Z", "score": 7.0, "summary_zh": "GPT-5.5幻觉率是GLM-5.2的三倍，引发对模型可靠性的讨论。", "category": "教程观点", "tags": "[\"模型幻觉\", \"模型对比\", \"提示词技巧\"]", "why_zh": "直观对比幻觉数据，为创作者选用AI工具与优化提示词提供参考。", "idea_zh": "可制作一期视频，实测几个主流模型的幻觉率，分享降低幻觉的提示词技巧，突出对内容创作的影响。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436864.html", "source": "量子位", "title": "我把昨晚的梦输入AI，它居然直接把我拉进去玩儿了一把？！", "excerpt": "创作者的终极玩具来了", "published_at": "2026-06-19T08:42:04Z", "score": 8.0, "summary_zh": "新工具可将梦境描述输入AI，直接生成可进入的互动3D世界，堪称“创作者终极玩具”。", "category": "产品动态", "tags": "[\"AI生图\", \"AI视频\", \"梦境生成\", \"创作者工具\"]", "why_zh": "极具视觉冲击与实操演示价值，适合推出“用AI走入梦境”的实测视频。", "idea_zh": "用该工具将梦境描述转化为可探索的场景，拍摄“AI让我走进自己昨晚的梦”实测视频，展示文字到沉浸式世界的转换过程，强调一人即可创作大片级内容，发布于小红书和视频号。", "title_zh": ""}, {"url": "https://vettedconsumer.com/glm-5-2-the-most-powerful-open-weight-model-yet-and-the-brutal-reality-of-running-it-locally/", "source": "Hacker News", "title": "GLM-5.2: The Most Powerful Open Model yet and the Brutal Reality of Running It", "excerpt": "Article URL: https://vettedconsumer.com/glm-5-2-the-most-powerful-open-weight-model-yet-and-the-brutal-reality-of-running-it-locally/ Comments URL: https://news.ycombinator.com/item?id=48594012 Points: 24 # Comments: 2", "published_at": "2026-06-19T01:58:00Z", "score": 9.0, "summary_zh": "GLM-5.2号称最强开源模型，但本地运行挑战重重", "category": "模型发布", "tags": "[\"大模型\", \"开源\", \"GLM-5.2\", \"实测\"]", "why_zh": "可做实测对比与本地部署教程，契合AI智能体赛道", "idea_zh": "一期视频/图文：实测GLM-5.2性能，对比其他开源模型，并给出本地运行配置建议，落点「个人/小团队如何低成本用上最强开源模型」", "title_zh": ""}, {"url": "https://www.intheweights.com/", "source": "Hacker News", "title": "Show HN: Are You in the Weights?", "excerpt": "With more traffic moving off-web and into LLMs, I got curious about what traces we leave \"in the weights\". My design partner and I built a site in the past few weeks that checks recognition across frontier and small models. It queries many of them in parallel, clusters the responses, and tells you how strongly they recognize you. Happy to answer any questions here! Comments URL: https://news.ycombinator.com/item?id=48591348 Points: 8 # Comments: 0", "published_at": "2026-06-18T20:49:03Z", "score": 8.0, "summary_zh": "新工具可检查个人/品牌名在多个AI模型中的识别度，评估你在大模型中的存在感。", "category": "产品动态", "tags": "[\"AI识别\", \"个人IP\", \"AI搜索\", \"曝光度\"]", "why_zh": "直接关联创作者IP和公域获客，AI是否认识你影响未来流量，实测性强。", "idea_zh": "制作对比实测：用该工具搜索你与3个同行，分析AI识别差异的原因，给出如何通过内容布局提升AI可见度的建议，适合小红书图文或视频号口播。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/18/perplexity-launches-brain/", "source": "MarkTechPost", "title": "Perplexity Launches Brain, a Self-Improving Memory System That Builds a Context Graph of an Agent’s Work and Learns Overnight", "excerpt": "Perplexity has launched Brain, a self-improving memory system for its Computer agent. Instead of remembering the user, Brain remembers the agent's work — what worked, what failed, and what corrections got made. It builds a traceable context graph, reviews it overnight, and reports early gains in correctness, recall, and cost. The post Perplexity Launches Brain, a Self-Improving Memory System That Builds a Context Graph of an Agent’s Work and Learns Overnight appeared first on MarkTechPost ", "published_at": "2026-06-18T20:26:20Z", "score": 8.0, "summary_zh": "Perplexity发布Brain记忆系统，为Computer agent构建工作上下文图谱，夜间学习优化，提升效率。", "category": "产品动态", "tags": "[\"AI Agent\", \"记忆系统\", \"Perplexi\", \"工作流优化\"]", "why_zh": "Perplexity新功能让AI Agent记忆工作过程并自我优化，直接关联AI智能体提效。", "idea_zh": "可实测Perplexity Brain，演示让AI记住工作步骤并自动迭代，制作对比教程，落地一人公司自动化提效（小红书/视频）。", "title_zh": ""}, {"url": "https://github.blog/changelog/2026-06-18-copilot-code-review-agents-md-support-and-ui-improvements", "source": "GitHub Changelog", "title": "Copilot code review: AGENTS.md support and UI improvements", "excerpt": "Copilot code review now supports repository-level AGENTS.md files, and it’s easier to request a review from Copilot on draft pull requests with the Request button. These changes are all generally… The post Copilot code review: AGENTS.md support and UI improvements appeared first on The GitHub Blog .", "published_at": "2026-06-18T19:11:51Z", "score": 7.0, "summary_zh": "Copilot 代码审查支持 AGENTS.md 文件及草稿 PR 请求按钮。", "category": "产品动态", "tags": "[\"Copilot\", \"Agent工作流\", \"AGENTS.m\"]", "why_zh": "AGENTS.md 可定义 AI 代理行为，与 Codex Skill 功能类似，可做对比教程。", "idea_zh": "实操演示配置 AGENTS.md，与 Codex Skill 对比。适合小红书图文，标题《AGENTS.md vs Skill 谁更听话？》。社群作业：为自己的项目写 AGENTS.md 并测试。", "title_zh": ""}, {"url": "https://github.com/openai/openai-cookbook/commit/432e0030de8952a4e2e37d81d3ec16ed23d78079", "source": "OpenAI Cookbook", "title": "Add interactive Workspace Agent API trigger cookbook", "excerpt": "Add interactive Workspace Agent API trigger cookbook", "published_at": "2026-06-18T15:02:43Z", "score": 8.0, "summary_zh": "OpenAI Cookbook 新增交互式 Workspace Agent API 触发器教程", "category": "教程观点", "tags": "[\"Agent工作流\", \"API触发器\", \"Codex\"]", "why_zh": "直接演示 Agent 工作流触发逻辑，可做成普通用户的实操指南", "idea_zh": "翻译为小红书图文：‘如何让 AI 自动执行工作流任务’，演示用 Codex 触发 Workspace Agent，对比手动操作，适合社群作业", "title_zh": ""}, {"url": "https://tester.army", "source": "Hacker News", "title": "Launch HN: TesterArmy (YC P26) – Agents that test web and mobile apps", "excerpt": "Hey HN - we’re Oskar, Szymon, and Piotr, and we’re building TesterArmy ( https://tester.army ). TesterArmy is an agentic testing platform that runs end-to-end checks before deployment and in production. Instead of wasting hours on manual testing or maintaining static scripts, we let you specify your tests in natural language and handle everything in between. We've built the platform fully around agents. Our agent will reliably execute the tests, but your coding agent can manage everything in our", "published_at": "2026-06-18T14:49:44Z", "score": 7.0, "summary_zh": "TesterArmy发布：AI代理平台可自动执行Web和移动应用端到端测试。", "category": "产品动态", "tags": "[\"AI代理\", \"自动化测试\", \"SaaS\"]", "why_zh": "Agent在实际场景的应用案例，能体现AI替代人工的提效逻辑。", "idea_zh": "选题：《AI代理帮你测APP？实测TesterArmy，省下10倍测试时间》。形式：小红书图文（测试流程对比截图+省时数据）；落点：一人公司用AI自动化测试，把精力留给核心业务。", "title_zh": ""}, {"url": "https://www.theverge.com/tech/952099/adobe-ai-assistants-photoshop-premiere-illustrator-beta-launch", "source": "The Verge AI", "title": "Photoshop and Premiere now have AI assistants", "excerpt": "Adobe's plan to stick AI assistants into all of its Creative Cloud suite is now fully underway, with new chatbots now rolling out to its biggest editing and design apps. As part of a public beta launching today, Photoshop, Premiere, Illustrator, InDesign, and Frame.io now each have a bespoke AI Assistant that can be used […]", "published_at": "2026-06-18T12:08:31Z", "score": 8.0, "summary_zh": "Adobe为Photoshop、Premiere等五款核心应用推出AI助手，可问答和指导操作。", "category": "产品动态", "tags": "[\"Adobe\", \"AI助手\", \"图像编辑\", \"视频剪辑\"]", "why_zh": "直接契合创作者剪辑与设计工作流，可实测AI助手提效效果。", "idea_zh": "制作Premiere AI助手实测短视频：用AI快速查功能、学剪辑技巧，对比传统搜索方式，突出‘一人团队’效率提升，挂剪辑服务引导。", "title_zh": ""}, {"url": "https://www.theverge.com/tech/952104/adobe-firefly-ai-agent-elements-projects-update", "source": "The Verge AI", "title": "Adobe’s redesigned AI studio remembers what your creations look like", "excerpt": "Adobe is introducing some new capabilities for its Firefly AI assistant, alongside a \"reimagined\" AI studio that lets you edit and generate new designs from a single interface. The new Firefly experience launching today in private beta is designed to give you \"persistent context, reusable assets, and organized workflows\" across your projects, according to Adobe, […]", "published_at": "2026-06-18T11:59:54Z", "score": 8.0, "summary_zh": "Adobe Firefly推出重新设计的AI工作室，支持持久上下文与可重用资源，内测中。", "category": "产品动态", "tags": "[\"Adobe Fi\", \"AI生图\", \"AI设计\"]", "why_zh": "Firefly新功能可能革新设计工作流，适合实测其为品牌IP提效。", "idea_zh": "图文对比Firefly新旧版，展示如何用持久上下文批量生成风格统一的素材，解决老板经常‘风格不一致’痛点，收尾引导AI社群。", "title_zh": ""}, {"url": "https://www.elastic.co/search-labs/blog/agent-memory-elasticsearch", "source": "Hacker News", "title": "We built a persistent agent memory layer on Elasticsearch with 0.89 recall", "excerpt": "Article URL: https://www.elastic.co/search-labs/blog/agent-memory-elasticsearch Comments URL: https://news.ycombinator.com/item?id=48583703 Points: 10 # Comments: 0", "published_at": "2026-06-18T11:20:18Z", "score": 7.0, "summary_zh": "基于Elasticsearch构建的AI Agent持久记忆层，召回率达0.89。", "category": "产品动态", "tags": "[\"AI Agent\", \"记忆\", \"Elastics\", \"工作流\"]", "why_zh": "提供Agent记忆具体实现方案，可转化为给AI助手赋予记忆的教程，吸引创业者。", "idea_zh": "小红书图文：实测用Elasticsearch为AI Agent添加长期记忆，演示配置步骤与效果对比，引导至社群。落点：一人公司用AI记忆功能处理客户常见问题。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3858471555142913?f=rss", "source": "36氪", "title": "小米正式发布“全屋智能AI开源方案”Xiaomi Miloco 2.0", "excerpt": "36氪获悉，6月18日，小米正式发布面向未来的“全屋智能AI开源方案”Xiaomi Miloco 2.0，该方案以小米自研 MiMo 大模型为智能核心，在Miloco 1.0的基础上实现交互方式、产品功能、记忆系统等方面的全面升级。同时Miloco 2.0使用门槛相比前代大幅降低，以Agent插件形式接入OpenClaw，支持一键安装和部署，支持macOS、Linux、Windows系统。隐私保护全面升级，支持数据隔离和本地数据一键清空。", "published_at": "2026-06-18T09:28:26Z", "score": 8.0, "summary_zh": "小米发布基于自研MiMo大模型的全屋智能AI开源方案Miloco 2.0，支持Agent插件一键部署。", "category": "产品动态", "tags": "[\"AI Agent\", \"全屋智能\", \"开源方案\", \"小米\"]", "why_zh": "小米AI Agent开源方案适合实测，可关联一人公司智能家居提效场景。", "idea_zh": "实测Miloco 2.0的Agent功能，演示在macOS/Windows上一键安装与OpenClaw联动，拍摄「一人公司用AI Agent控制全屋设备」的短视频，落点AI降本增效。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436518.html", "source": "量子位", "title": "全员本科生！何恺明组新作：文生图，258M参数就够了", "excerpt": "整篇论文一共六位作者。除了何恺明之外，其余五位都还是本科生。", "published_at": "2026-06-18T09:12:09Z", "score": 9.0, "summary_zh": "何恺明团队推出仅258M参数的文生图模型，作者多为本科生，参数极小。", "category": "论文研究", "tags": "[\"文生图\", \"轻量模型\", \"何恺明\"]", "why_zh": "极小模型文生图，可能大幅降低AI生图门槛和成本，非常适合一人公司创作者。", "idea_zh": "待模型开源后，实测对比SDXL，展示在小红书封面生成上的速度与质量优势，突出低成本提效。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436517.html", "source": "量子位", "title": "18年前不让你苦修PS的美图，这次不让你苦修AI了", "excerpt": "用八款产品，讲一个变化", "published_at": "2026-06-18T09:07:51Z", "score": 8.0, "summary_zh": "美图推出/整合多项AI功能，覆盖图片处理、设计、视频等，降低创作门槛。", "category": "产品动态", "tags": "[\"美图\", \"AI修图\", \"内容创作提效\"]", "why_zh": "美图AI工具易上手，适合不懂技术的创作者，可大幅提升封面、海报制作效率。", "idea_zh": "实测美图AI的修图、设计功能，与PS等对比，演示一键生成小红书封面，强调一人全流程搞定。", "title_zh": ""}, {"url": "https://chat.deepseek.com/", "source": "Hacker News", "title": "DeepSeek Introduces Vision", "excerpt": "Article URL: https://chat.deepseek.com/ Comments URL: https://news.ycombinator.com/item?id=48581458 Points: 9 # Comments: 0", "published_at": "2026-06-18T06:17:40Z", "score": 8.0, "summary_zh": "DeepSeek 模型新增视觉能力，支持图像输入。", "category": "模型发布", "tags": "[\"DeepSeek\", \"多模态\", \"视觉模型\", \"AI实测\"]", "why_zh": "DeepSeek 作为热门国产模型，新增视觉功能，是实测演示的绝佳选题。", "idea_zh": "制作视频/图文实测 DeepSeek 视觉功能：上传截图、图表或手写笔记，测试其理解与分析能力，可与 GPT-4V 对比，落点“用 AI 省掉设计/数据分析外包”。", "title_zh": ""}, {"url": "https://mindgard.ai/blog/chatgpt-spontaneously-generated-violent-images-from-a-viral-prompt", "source": "Hacker News", "title": "ChatGPT Spontaneously Generates Sexual Violence and Hardcore Snuff Imagery", "excerpt": "Article URL: https://mindgard.ai/blog/chatgpt-spontaneously-generated-violent-images-from-a-viral-prompt Comments URL: https://news.ycombinator.com/item?id=48578894 Points: 9 # Comments: 2", "published_at": "2026-06-18T00:24:04Z", "score": 7.0, "summary_zh": "ChatGPT可被诱导生成暴力色情图像，安全研究披露漏洞。", "category": "产品动态", "tags": "[\"AI安全\", \"ChatGPT\", \"图像生成\", \"漏洞\"]", "why_zh": "AI生图赛道可探讨内容安全过滤机制与创作者避坑指南。", "idea_zh": "可制作短视频或图文，解析该漏洞原理，提醒创作者注意AI平台内容政策，避免账号被封；落点在合规使用工具。", "title_zh": ""}, {"url": "https://huggingface.co/blog/is-it-agentic-enough", "source": "Hugging Face", "title": "Is it agentic enough? Benchmarking open models on your own tooling", "excerpt": "", "published_at": "2026-06-18T00:00:00Z", "score": 8.0, "summary_zh": "Hugging Face文章探讨在自己的工具链上评测开放模型的智能体能力。", "category": "教程观点", "tags": "[\"AI智能体\", \"基准测试\", \"开源模型\"]", "why_zh": "直接命中Agent评测，创作者可借鉴评测方法或对比模型，为社区提供实操指南。", "idea_zh": "可做一期视频/图文，演示如何用开源工具评测GPT、Claude、国产模型的Agent能力，给出评分和选择建议，落点一人公司如何选对模型提效。", "title_zh": ""}, {"url": "https://simonwillison.net/2026/Jun/17/glm-52/#atom-everything", "source": "Simon Willison", "title": "GLM-5.2 is probably the most powerful text-only open weights LLM", "excerpt": "Chinese AI lab Z.ai released GLM-5.2 to their coding plan subscribers on June 13th, and then yesterday (June 16th) released the full open weights under an MIT license. Similar in size to their previous GLM-5 and GLM-5.1 releases, this is 753B parameter, 1.51TB monster - with 40 active parameters (Mixture of Experts). GLM-5.2 is a text input only model - Z.ai have a separate vision family most recently represented by GLM-5V-Turbo , but that one isn't open weights. GLM-5.2 has a 1 million token co", "published_at": "2026-06-17T23:58:39Z", "score": 9.0, "summary_zh": "Z.ai发布开源大模型GLM-5.2，753B参数，1.51TB，采用MIT许可。", "category": "模型发布", "tags": "[\"GLM-5.2\", \"开源\", \"大模型\", \"Z.ai\"]", "why_zh": "国产开源巨型模型，实测对比潜力大，契合AI Agent赛道的内容选题。", "idea_zh": "选题：GLM-5.2 vs Claude/GPT 实测，聚焦文案生成、策划提效，小红书图文+视频号演示，突出“免费调用顶级模型降本增效”。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3857877083378692?f=rss", "source": "36氪", "title": "谷歌开始销售搭载Gemini的Google Home音箱", "excerpt": "谷歌周三表示，其为Gemini打造的Google Home音箱将从下周开始销售。谷歌称，该设备是谷歌首款为该AI模型打造的音频设备，售价为99.99美元。（新浪财经）", "published_at": "2026-06-17T23:21:50Z", "score": 7.0, "summary_zh": "谷歌搭载Gemini的Google Home音箱下周开售，售价99.99美元，首款为Gemini打造的音频设备。", "category": "产品动态", "tags": "[\"Gemini\", \"Google H\", \"AI音箱\", \"硬件发布\"]", "why_zh": "首款Gemini音频硬件，可实测AI语音助手在日常办公与内容创作中的提效场景。", "idea_zh": "做一期开箱实测短视频：用Google Home语音操控安排日程、快速查询信息、生成灵感笔记，对比手机端Gemini体验，突出「一人公司AI助理」的实用技巧，发布在视频号或小红书。", "title_zh": ""}, {"url": "https://openrouter.ai/blog/insights/royale-last-agent-standing/", "source": "Hacker News", "title": "A Robot Is Sprinting Towards You: Do You Want It Running on Claude or Grok?", "excerpt": "Article URL: https://openrouter.ai/blog/insights/royale-last-agent-standing/ Comments URL: https://news.ycombinator.com/item?id=48576824 Points: 4 # Comments: 0", "published_at": "2026-06-17T21:00:07Z", "score": 7.0, "summary_zh": "OpenRouter 让机器人选择运行 Claude 或 Grok，比较模型在 Agent 任务中的决策能力。", "category": "产品动态", "tags": "[\"Claude\", \"Grok\", \"Agent测试\", \"模型对比\"]", "why_zh": "趣味比较 Claude 与 Grok 的 Agent 决策能力，适合做实测类内容。", "idea_zh": "选题：实测 Claude vs Grok 在‘机器人危机’场景下的指令遵循与幻觉率。做成小红书图文，用截图对比输出结果，落点强调‘AI Agent 可靠性如何影响一人公司关键任务’。", "title_zh": ""}, {"url": "https://blog.cloudflare.com/agents-platform-flue-sdk/", "source": "Cloudflare Blog", "title": "Bringing more agent harnesses and frameworks to Cloudflare, starting with Flue", "excerpt": "The Agents SDK is now a runtime any agent framework can build on. Today we're opening up the Agents SDK primitives, with Flue as a first framework targeting Agents SDK, and rolling out agents in the dashboard.", "published_at": "2026-06-17T19:35:00Z", "score": 8.0, "summary_zh": "Cloudflare开放Agents SDK，Flue成为首个适配框架。", "category": "产品动态", "tags": "[\"AI Agent\", \"Cloudfla\", \"Flue\", \"SDK\"]", "why_zh": "新的Agent开发工具出现，可关注如何用它搭建自动化工作流提效。", "idea_zh": "测评Flue在Cloudflare上构建简单AI助手，展示一人公司如何用Agent做客服或内容抓取，适合小红书图文教程。", "title_zh": ""}, {"url": "https://github.com/evanklem/polypore", "source": "Hacker News", "title": "Agentic coding deserves more than a chat box bolted onto VS Code", "excerpt": "Article URL: https://github.com/evanklem/polypore Comments URL: https://news.ycombinator.com/item?id=48571811 Points: 5 # Comments: 0", "published_at": "2026-06-17T15:26:29Z", "score": 7.0, "summary_zh": "GitHub项目Polypore提出，为Agentic编程设计专用界面，而非简单聊天框嵌入IDE。", "category": "产品动态", "tags": "[\"智能体\", \"编程工具\", \"Polypore\", \"工作流\"]", "why_zh": "Agentic编程新界面方案，可实测对比现有工具，吸引开发者群体。", "idea_zh": "视频实测Polypore vs Copilot Chat：演示项目搭建速度，展示界面优势，强调一人开发提效，适配独立开发者/创业者。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3857211421791239?f=rss", "source": "36氪", "title": "微软正考虑接入深度求索模型赋能Copilot协同办公工具", "excerpt": "据阿克西奥斯新闻网报道，微软正考虑引入中国人工智能企业深度求索推出的 V4 大模型，作为旗下Copilot协同办公智能助手的低成本备选技术方案。报道称，这家美国科技巨头正在测试由微软云本地部署运行的深度求索V4版本以及其他开源大模型，用来替代目前为Copilot协同办公提供技术支持的Anthropic、OpenAI大模型，以此压缩服务成本。微软计划在未来几周内面向市场推出这款低成本大模型服务。（新浪财经）", "published_at": "2026-06-17T12:04:06Z", "score": 8.0, "summary_zh": "微软正测试深度求索V4等开源模型作为Copilot低成本备选方案，以替代OpenAI和Anthropic模型", "category": "模型发布", "tags": "[\"微软\", \"深度求索\", \"Copilot\", \"大模型\"]", "why_zh": "微软拥抱国产开源模型，话题性强，可做低成本方案实测对比", "idea_zh": "实测深度求索V4在办公场景的能力，对比GPT-4o，突出‘省钱提效’，适合小红书图文/视频口播，标题如“微软都搞的国产模型，打工人用起来”", "title_zh": ""}, {"url": "https://techcrunch.com/2026/06/17/pinterest-launches-an-experimental-ai-shopping-app-called-ask-pinterest/", "source": "TechCrunch AI", "title": "Pinterest launches an experimental AI shopping app called ‘Ask Pinterest’", "excerpt": "Pinterest has launched 'Ask Pinterest,' an experimental AI-powered shopping app that lets users seek recommendations and inspiration through a conversational interface.", "published_at": "2026-06-17T11:00:00Z", "score": 7.0, "summary_zh": "Pinterest发布实验性AI购物应用“Ask Pinterest”，用对话界面提供个性化推荐。", "category": "产品动态", "tags": "[\"对话AI\", \"购物\", \"产品发布\"]", "why_zh": "对话式AI在消费场景的新应用，可实测对话推荐效果，为内容创作和带货提供灵感。", "idea_zh": "实测“Ask Pinterest”的推荐精准度和对话体验，对比其他AI购物助手，拆解其如何用AI生成个性化内容，适合视频号口播演示，落点在“一人公司如何用AI快速获取选品灵感”。", "title_zh": ""}, {"url": "https://cells2pixels.github.io/", "source": "Hacker News", "title": "Show HN: High-Res Neural Cellular Automata", "excerpt": "Neural CAs model self-organizing pattern formation. Now they can generate patterns at HD resolution in real-time, enabled by turning each CA cell into a Neural Field. Try 3 demos: grow a pattern from a seed (and damage it, it heals), synthesize PBR textures that can regenerate, or create 3D textures like clouds. Comments URL: https://news.ycombinator.com/item?id=48567877 Points: 8 # Comments: 0", "published_at": "2026-06-17T09:28:36Z", "score": 7.0, "summary_zh": "神经网络元胞自动机可实时生成高清图案、PBR和3D纹理，支持自愈。", "category": "论文研究", "tags": "[\"AI生图\", \"元胞自动机\", \"纹理生成\"]", "why_zh": "可实测的AI生图新工具，展示自我修复和纹理生成，适合制作新奇内容。", "idea_zh": "实测该工具，录制从种子生成图案、破坏后自动修复的过程，并演示导出纹理用于视频封面或设计素材，发布小红书图文或视频号口播，突出AI创意的降本增效。", "title_zh": ""}, {"url": "https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index", "source": "Hacker News", "title": "GLM-5.2 is the new leading open weights model on Artificial Analysis", "excerpt": "Article URL: https://artificialanalysis.ai/articles/glm-5-2-is-the-new-leading-open-weights-model-on-the-artificial-analysis-intelligence-index Comments URL: https://news.ycombinator.com/item?id=48567759 Points: 13 # Comments: 1", "published_at": "2026-06-17T09:12:00Z", "score": 9.0, "summary_zh": "GLM-5.2成为Artificial Analysis评测中领先的开源权重模型。", "category": "模型发布", "tags": "[\"大模型\", \"开源\", \"GLM-5.2\", \"评测\"]", "why_zh": "国产开源模型登顶评测榜单，强性价比和中文能力适合实测。", "idea_zh": "标题：GLM-5.2 开源模型实测，到底能不能打？内容：用GLM-5.2与Claude、GPT做写作、策划、编程等任务对比，突出性价比和中文优势，小红书图文/视频号口播，落点：一人公司选模型省钱提效。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3857032335283459?f=rss", "source": "36氪", "title": "百度搭子DuMate移动端上线苹果App Store", "excerpt": "36氪获悉，百度搭子DuMate移动端APP iOS版本正式登陆苹果App Store。此前，DuMate已在Create大会上发布移动端能力，并率先上线安卓版本。此次iOS端上线后，用户可在安卓和iOS设备上调用DuMate，处理信息检索、资料整理、内容生成、任务规划等场景，让AI协作不再局限于固定设备。", "published_at": "2026-06-17T09:02:07Z", "score": 7.0, "summary_zh": "百度AI助手DuMate App登陆苹果商店，支持跨设备信息处理与内容生成。", "category": "产品动态", "tags": "[\"DuMate\", \"百度AI\", \"移动端\", \"AI助手\"]", "why_zh": "可实测百度DuMate App，探索其作为个人助理的内容提效能力。", "idea_zh": "小红书图文或视频号：《百度DuMate实测：免费AI助手能帮创业者省多少事？》聚焦任务规划、内容生成，对比其他助手，落点一人公司提效。", "title_zh": ""}, {"url": "https://huggingface.co/blog/zai-org/glm-52-blog", "source": "Hugging Face", "title": "GLM-5.2: Built for Long-Horizon Tasks", "excerpt": "", "published_at": "2026-06-17T09:01:25Z", "score": 9.0, "summary_zh": "智谱发布GLM-5.2模型，专为长周期任务打造，可能具备更强复杂推理与规划能力。", "category": "模型发布", "tags": "[\"GLM-5.2\", \"智谱\", \"长任务\", \"大模型\"]", "why_zh": "新模型主打长周期任务，可实测对比Claude等，切中一人公司复杂项目提效需求。", "idea_zh": "视频号口播+小红书图文：用GLM-5.2策划一场线上活动，对比其他模型，展示其长链条推理能力，标题如《新一代长任务AI来了！GLM-5.2能让一人公司效率翻倍吗？》", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856972887971075?f=rss", "source": "36氪", "title": "国家超算互联网上线GLM-5.2模型服务", "excerpt": "36氪获悉，智谱全新旗舰大模型GLM-5.2今日正式开源，当日国家超算互联网同步上线GLM-5.2模型API调用服务及模型文件。目前超算互联网AI社区已汇集1400余款开源大模型，GLM全系列、DeepSeek-V4、MiniMax-M3、Kimi-K2.6等国产模型API均已上线。", "published_at": "2026-06-17T08:02:09Z", "score": 9.0, "summary_zh": "智谱开源旗舰模型GLM-5.2，国家超算互联网同步上线API及模型文件。", "category": "模型发布", "tags": "[\"大模型开源\", \"GLM-5.2\", \"超算互联网\"]", "why_zh": "国产大模型开源可实测，适合创作测评对比内容。", "idea_zh": "实测GLM-5.2在文案、数据分析等任务上的表现，对比DeepSeek等同级模型，做成小红书图文或视频，强调免费开源、一人公司提效。", "title_zh": ""}, {"url": "https://www.marktechpost.com/2026/06/17/minimax-sparse-attention-msa-a-two-branch-block-sparse-attention-trained-on-a-109b-parameter-moe-with-a-3t-token-budget/", "source": "MarkTechPost", "title": "MiniMax Sparse Attention (MSA): a Two-Branch Block-Sparse Attention Trained on a 109B-Parameter MoE With a 3T-Token Budget", "excerpt": "MiniMax released MSA, a sparse attention built on Grouped Query Attention. A lightweight Index Branch selects Top-k key-value blocks per query and GQA group; the Main Branch attends only to those blocks. It matches GQA on downstream benchmarks while reducing per-token attention compute 28.4× at 1M context. The post MiniMax Sparse Attention (MSA): a Two-Branch Block-Sparse Attention Trained on a 109B-Parameter MoE With a 3T-Token Budget appeared first on MarkTechPost .", "published_at": "2026-06-17T07:44:54Z", "score": 7.0, "summary_zh": "MiniMax发布稀疏注意力MSA，在109B模型上减少注意力计算28.4倍。", "category": "论文研究", "tags": "[\"稀疏注意力\", \"MiniMax\", \"大模型效率\"]", "why_zh": "MiniMax新注意力技术可大幅降低大模型推理成本，影响未来AI应用。", "idea_zh": "用通俗比喻解读MSA原理，对比普通注意力与稀疏注意力的算力差异，结合“一人公司降本”落点，适合小红书图文，突出“更便宜的AI”。", "title_zh": ""}, {"url": "https://artificialanalysis.ai/models/glm-5-2", "source": "Hacker News", "title": "GLM 5.2 Performance Benchmarks", "excerpt": "Article URL: https://artificialanalysis.ai/models/glm-5-2 Comments URL: https://news.ycombinator.com/item?id=48567004 Points: 5 # Comments: 0", "published_at": "2026-06-17T07:30:51Z", "score": 8.0, "summary_zh": "GLM 5.2模型性能基准测试结果公布，展现最新国产大模型能力。", "category": "模型发布", "tags": "[\"GLM 5.2\", \"大模型\", \"基准测试\", \"国产模型\"]", "why_zh": "国产模型GLM 5.2实测数据新鲜，可直接作为AI智能体赛道素材，对比测评潜力大。", "idea_zh": "标题：GLM 5.2 vs GPT-4o 实测对比：中文任务谁更强？形式：小红书图文，展示写作、逻辑、编码等典型场景截图，结尾总结一人公司选用建议，强调降本增效。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856862891037699?f=rss", "source": "36氪", "title": "微信支付发布AI专属卡", "excerpt": "36氪获悉，6月17日，微信支付正式发布“AI专属卡”。据介绍，这是一款专为AI智能体支付场景设计、内置于微信零钱内的产品功能。同时，腾讯桌面办公效率智能体WorkBuddy宣布接入微信支付，并与美团进一步深化生态合作。美团为WorkBuddy开发专属能力支持购买团购，同时提供专属优惠券。", "published_at": "2026-06-17T06:16:04Z", "score": 7.0, "summary_zh": "微信支付发布AI专属卡，为AI智能体支付场景设计；WorkBuddy接入并与美团合作。", "category": "产品动态", "tags": "[\"微信支付\", \"AI智能体\", \"Agent\"]", "why_zh": "为AI智能体提供支付能力，与创作者关注的Agent落地和提效直接相关。", "idea_zh": "实测微信AI专属卡用法，演示WorkBuddy+美团点餐，说明AI智能体简化生活操作，做成小红书图文/短视频，落点「一人公司用智能体自动支付」。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436134.html", "source": "量子位", "title": "Grok 4.3现已在Amazon Bedrock上正式可用", "excerpt": "xAI正式成为Amazon Bedrock的模型供应商之一", "published_at": "2026-06-17T05:59:53Z", "score": 8.0, "summary_zh": "xAI的Grok 4.3模型上线Amazon Bedrock，用户可通过API调用。", "category": "模型发布", "tags": "[\"Grok\", \"模型发布\", \"Amazon B\"]", "why_zh": "新模型可用，可实测并对比主流模型，吸引关注AI工具提效的受众。", "idea_zh": "选题“Grok 4.3实测对比：又多一个强模型可选？”，录制视频展示在Bedrock调用Grok完成选题、脚本生成，与GPT-4o、Claude对比质量与风格，落点帮创作者选模型提效。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856820367480065?f=rss", "source": "36氪", "title": "摩尔线程完成智谱GLM-5.2 Day-0适配", "excerpt": "36氪获悉，6月17日，摩尔线程宣布在AI训推一体全功能GPU智算卡MTT S5000上，完成对智谱新一代开源旗舰模型GLM-5.2的Day-0极速适配。此次适配延续了摩尔线程在GLM-5.1长上下文Prefill与P/D异构分离推理场景中的优化积累，并面向GLM-5.2超长上下文与复杂推理负载，进一步释放MTT S5000在长输入Prefill阶段的高吞吐优势。", "published_at": "2026-06-17T05:30:02Z", "score": 7.0, "summary_zh": "摩尔线程宣布在MTT S5000上完成智谱GLM-5.2模型Day-0适配，优化长上下文推理。", "category": "模型发布", "tags": "[\"智谱GLM-5.\", \"国产GPU\", \"模型适配\", \"长上下文\"]", "why_zh": "智谱新开源旗舰模型发布，国产芯片快速适配，是测评国产AI生态的机会。", "idea_zh": "可做一期对比实测：GLM-5.2在不同平台（英伟达vs摩尔线程）上的推理速度与效果，突出国产方案降本潜力，适合小红书图文。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436110.html", "source": "量子位", "title": "天工3.1 重磅发布：上线 Skywork Design 与 Dynamic Workflows，给 AI 一张画布和一支军团", "excerpt": "天工超级智能体的收入实现了三倍增长", "published_at": "2026-06-17T04:32:09Z", "score": 8.0, "summary_zh": "天工3.1发布Skywork Design设计工具和Dynamic Workflows动态工作流。", "category": "产品动态", "tags": "[\"天工\", \"AI设计\", \"工作流\", \"智能体\"]", "why_zh": "新功能直击AI设计与自动化工作流，可实测降本增效。", "idea_zh": "实测天工3.1：用Skywork Design做封面、Dynamic Workflows串联选题到发布全流程，演示一人公司内容生产提效。适合小红书图文或视频号口播。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856752470529284?f=rss", "source": "36氪", "title": "小米携手金山办公正式发布MiMo Claw", "excerpt": "36氪获悉，6月16日晚，Xiaomi MiMo开放平台宣布，智能体产品MiMo Claw正式版全面上线。此次迭代更新联动金山办公，打造全链路文档办公闭环提效方案；新增多项 OpenClaw核心便捷能力；同时升级免费用户权益，并面向高频、专业用户上线 TokenPlan分层订阅，满足不同使用需求。", "published_at": "2026-06-17T04:30:02Z", "score": 8.0, "summary_zh": "小米与金山办公联合推出智能体MiMo Claw正式版，打通WPS文档办公全链路。", "category": "产品动态", "tags": "[\"智能体\", \"MiMo Cla\", \"小米\", \"金山办公\"]", "why_zh": "新发布的办公智能体，直接对应一人公司提效场景，实测价值高。", "idea_zh": "「小米×WPS智能体实测」：演示用MiMo Claw快速生成报告、处理表格，对比手动操作，突出一人团队10分钟完成过去1小时的工作。适合小红书图文拆解+视频号口播展示。", "title_zh": ""}, {"url": "https://www.qbitai.com/2026/06/436085.html", "source": "量子位", "title": "刚刚，Fable-5之下，智谱开源的GLM-5.2拿下AI编程第一！", "excerpt": "1M上下文", "published_at": "2026-06-17T02:42:10Z", "score": 9.0, "summary_zh": "智谱开源GLM-5.2模型，拥有1M上下文，在AI编程排行榜中夺冠。", "category": "模型发布", "tags": "[\"GLM-5.2\", \"开源模型\", \"AI编程\", \"大模型实测\"]", "why_zh": "国产开源大模型重大突破，编程能力登顶，适合做测评与提效教程。", "idea_zh": "实测对比GLM-5.2与Claude Sonnet在真实代码任务上的表现，制作视频口播或图文教程，突出其1M上下文在长代码维护中的降本增效场景。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856610761233669?f=rss", "source": "36氪", "title": "沐曦股份曦云C系列GPU Day 0 适配智谱GLM-5.2旗舰模型", "excerpt": "36氪获悉，6月17日，智谱AI全新旗舰大模型GLM-5.2正式开源发布。沐曦股份曦云C系列GPU率先完成该模型Day 0适配。", "published_at": "2026-06-17T01:58:03Z", "score": 8.0, "summary_zh": "智谱AI开源旗舰大模型GLM-5.2，沐曦曦云C系列GPU完成Day 0适配。", "category": "模型发布", "tags": "[\"智谱AI\", \"GLM-5.2\", \"开源大模型\", \"GPU适配\"]", "why_zh": "国产新旗舰开源模型，适合做实测对比内容，吸引关注AI提效的创业者和内容创作者。", "idea_zh": "录制GLM-5.2与DeepSeek/其他模型在文案生成、数据分析等场景的对比实测，突出效率与成本，发布小红书图文对比。", "title_zh": ""}, {"url": "https://36kr.com/newsflashes/3856507024872706?f=rss", "source": "36氪", "title": "竞争加剧+成本下降，Token迎降价潮", "excerpt": "今年以来一路走高的Token（词元）价格迎来反转。进入6月，全球多家大模型厂商开始重新调整Token价格，包括OpenAI在内，多家大厂先后被曝将下调Token价格，以回应企业用户对AI成本暴涨的不满。然而，相较此前声势浩大的推广潮，此次Token价格下调更像是行业新一轮残酷竞赛的序幕。多位市场人士分析称，随着OpenAI及Anthropic等一批头部AI厂商陆续推进IPO，市场对AI公司的估值模型正在调整，能否用更低成本持续获取更多用户，成为每家公司必须面对的考验。（上证报）", "published_at": "2026-06-17T00:27:11Z", "score": 7.0, "summary_zh": "多家大模型厂商下调Token价格，回应成本高涨不满，AI使用门槛降低。", "category": "行业新闻", "tags": "[\"Token降价\", \"AI成本\", \"大模型厂商\"]", "why_zh": "Token降价直接降低AI内容创作成本，利好一人公司提效，可做降本增效选题。", "idea_zh": "对比主流大模型降价前后成本，测算视频脚本、文案生成等场景的花费变化，制作图文/视频突出“一人公司如何用更少钱跑AI工作流”。", "title_zh": ""}, {"url": "https://writings.stephenwolfram.com/2026/06/launching-version-15-of-wolfram-language-mathematica-built-in-useful-ai-lots-of-new-core-functionality/", "source": "Hacker News", "title": "Wolfram Language and Mathematica Version 15, AI Assistant, Symbolic Music, More", "excerpt": "Article URL: https://writings.stephenwolfram.com/2026/06/launching-version-15-of-wolfram-language-mathematica-built-in-useful-ai-lots-of-new-core-functionality/ Comments URL: https://news.ycombinator.com/item?id=48563609 Points: 4 # Comments: 1", "published_at": "2026-06-16T23:15:44Z", "score": 8.0, "summary_zh": "Wolfram语言15版发布，内置AI助手及符号音乐等全新核心功能。", "category": "产品动态", "tags": "[\"Wolfram\", \"AI助手\", \"新版本\", \"符号计算\"]", "why_zh": "集成AI助手，可实测其辅助数据分析与内容生成的效率，适合演示给创作者。", "idea_zh": "选题：实测Wolfram AI助手：用自然语言提问生成数据图表，能否替代传统制表工具？形式：小红书图文，展示步骤和效果，落点：一人公司用AI快速产出专业可视化内容。", "title_zh": ""}, {"url": "https://www.theverge.com/ai-artificial-intelligence/951099/meta-ai-mode-search-hands-on", "source": "The Verge AI", "title": "AI search grounded in Facebook posts? What could go wrong?", "excerpt": "AI is pretty reliable at putting things on your calendar these days, but it hasn't quite cracked answering the related and all-important question of \"What should I do this weekend?\" Meta's new AI Mode in search could be a useful tool - if it ever learns to stop getting stuff wrong. AI Mode is a […]", "published_at": "2026-06-16T22:11:23Z", "score": 7.0, "summary_zh": "Meta推出AI Mode搜索功能，基于Facebook帖子回答生活化问题，但准确性仍有问题。", "category": "产品动态", "tags": "[\"AI搜索\", \"Meta\", \"AI Mode\"]", "why_zh": "Meta新AI搜索功能实测话题性强，可吸引关注AI与生活结合的受众。", "idea_zh": "可做小红书图文，标题「我用Meta AI搜索规划周末，结果…」实测AI Mode的回答，对比真实推荐，落点在一人公司如何用AI搜索获取本地内容灵感。", "title_zh": ""}, {"url": "https://arstechnica.com/ai/2026/06/anthropic-pauses-token-based-billing-for-its-claude-agent-sdk/", "source": "Ars Technica", "title": "Anthropic \"pauses\" token-based billing for its Claude Agent SDK", "excerpt": "Move originally planned for Monday would have heavily increased power users' costs.", "published_at": "2026-06-16T21:00:35Z", "score": 7.0, "summary_zh": "Anthropic暂停Claude Agent SDK基于token的计费，原计划周一执行。", "category": "产品动态", "tags": "[\"Claude A\", \"计费变更\", \"AI Agent\"]", "why_zh": "Agent计费变动直接影响开发者和创业者成本，有实测与解读空间。", "idea_zh": "解读暂停原因及对新老用户的影响，对比早期计费模式，给出成本优化建议，适合图文/短视频。", "title_zh": ""}, {"url": "https://blog.goodman.dev/blog/octopus-agent-architecture/", "source": "Hacker News", "title": "The octopus architecture for AI agents", "excerpt": "Article URL: https://blog.goodman.dev/blog/octopus-agent-architecture/ Comments URL: https://news.ycombinator.com/item?id=48558696 Points: 3 # Comments: 1", "published_at": "2026-06-16T17:24:45Z", "score": 8.0, "summary_zh": "探讨 Octopus 架构：一种 AI 智能体系统设计", "category": "教程观点", "tags": "[\"AI智能体\", \"架构\", \"系统设计\"]", "why_zh": "新颖的智能体架构设计，可直接转化为 Agent 搭建教程或测评素材", "idea_zh": "制作视频/图文拆解 Octopus 架构，演示如何用类似思路构建 AI 营销助手，突出一人公司提效", "title_zh": ""}, {"url": "https://tim.blog/2026/06/12/has-ai-already-killed-nonfiction/", "source": "Hacker News", "title": "Has AI already killed self-help nonfiction books?", "excerpt": "Article URL: https://tim.blog/2026/06/12/has-ai-already-killed-nonfiction/ Comments URL: https://news.ycombinator.com/item?id=48558489 Points: 20 # Comments: 17", "published_at": "2026-06-16T17:11:05Z", "score": 7.0, "summary_zh": "Tim Ferriss 探讨 AI 是否已取代自助类非虚构书籍的写作，引发对内容创作未来的思考。", "category": "教程观点", "tags": "[\"AI写作\", \"非虚构书籍\", \"内容创作\", \"AI影响\"]", "why_zh": "话题引发创作者对 AI 写作工具与原创内容价值的讨论，契合受众关注点。", "idea_zh": "可做一期视频探讨 AI 写作对非虚构内容的影响，结合实测 AI 写书效果，给出自己的观点，吸引创作者讨论。", "title_zh": ""}, {"url": "https://arstechnica.com/ai/2026/06/spacex-will-acquire-coding-tool-cursor-to-compete-with-anthropic-openai/", "source": "Ars Technica", "title": "SpaceX to acquire AI coding platform Cursor for $60 billion", "excerpt": "Separately, neither could compete. Now they hope they can.", "published_at": "2026-06-16T16:37:45Z", "score": 7.0, "summary_zh": "SpaceX将以600亿美元收购AI编程平台Cursor。", "category": "融资商业", "tags": "[\"Cursor\", \"AI编程\", \"收购\"]", "why_zh": "Cursor是热门AI编程工具，收购可能影响其发展，值得关注。", "idea_zh": "可制作视频探讨Cursor对一人团队开发提效的价值，并预测收购可能带来的变化，强调AI赋能独立开发者。平台：视频号口播。", "title_zh": ""}]}