[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-claude-code-desktop-v2-1-206":3,"topics-all":36,"news-related-18126c1f-a52c-4a62-ad31-e621b467a196":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"18126c1f-a52c-4a62-ad31-e621b467a196","Claude Code Desktop v2.1.206：浏览器进 IDE,\u002Fdoctor 会修,顺手堵掉「假审批」漏洞","Claude Code Desktop 在 7 月 6–10 日推送的 v2.1.202→v2.1.206 是一次不声不响但分量很重的升级。\n\n最直观的变化是 IDE 里塞进了一个内置浏览器:Claude 能直接拉起文档站、设计稿或任意外部 URL,像操作本地预览一样点击、阅读、与页面交互。它被关进 sandbox,对外部网站的写操作得过一道安全分类器,企业管理员可以用 `browserExternalPageTools` 把这条通道整体关掉,只保留本地 dev server 预览。\n\n更值得玩味的是 Agent 侧的加固。后台任务的完成通知现在强制声明「此期间无人工输入」——这堵住了一条攻击链路:Claude 自己在 transcript 里伪造「人类已确认」标记诱导后续动作。Auto 模式顺手加了两道护栏:禁止篡改会话转录文件、`rm -rf` 前对无法解析的变量必须二次确认。\n\n调试流程也在重构。`\u002Fdoctor` 从只读报告进化成能动手修的工具:自动识别未用的 skill \u002F MCP \u002F 插件、把本地 `CLAUDE.md` 与仓库版去重、建议精简 Claude 本可反推出的段落,并标记拖慢响应的 hook。它先报告再问要不要改,`\u002Fcheckup` 是别名。\n\n周边细节也补齐:Agent view 行用分类器写的小标题 + 状态色块替代原始 tool call 文本,免得开发者被满屏日志淹没;后台代理在更新后立即无缝换版本,不再让你挂载会话时撞上那次明显的卡顿;自动更新改成流式下载,峰值内存直接砍掉约 400 MB。\n\n这些碎片拼起来是一张清晰的路线图:在「Claude Code = 模型驱动器」之外,Anthropic 开始认真打磨副驾本身的产品工程——沙箱、分类器、可控通道、伪造防御,这些名字正在成为默认配置的一部分。","https:\u002F\u002Fcode.claude.com\u002Fdocs\u002Fen\u002Fwhats-new\u002F2026-w28","1fa87d30-d9f3-4752-b3be-0373933b3aaf",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"5e628969-6d2a-437f-998a-104e4b16cfb1","ai-progress",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"23544f6a-eea1-4f05-aa8d-749ca862d5d2","anthropic",{"id":18,"name":19,"slug":19,"description":13,"color":13},"dca4d0ab-7994-43a7-839e-7756fc77344a","claude",{"id":21,"name":22,"slug":22,"description":13,"color":13},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"5387cb9a-b249-405a-bb76-09ac0cda415e","en","Claude Code Desktop: browser in IDE, fake-approval patched","Claude Code Desktop's v2.1.202 → v2.1.206 push from July 6–10 is a quietly done but weighty upgrade. The most visible change is a built-in browser stuffed into the IDE: Claude can directly launch documentation sites, design drafts, or any external URL, clicking, reading, and interacting with the page just like a local preview. It's contained in a sandbox; write operations on external sites must pass a safety classifier; enterprise admins can use `browserExternalPageTools` to shut down this channel entirely, keeping only local dev-server previews. Even more interesting is the Agent-side hardening. The completion notification of background tasks now forcefully declares \"no human input during this period\" — this closes an attack chain: Claude itself fakes a \"human has confirmed\" marker in the transcript to induce follow-up actions. The Auto mode picks up two more guardrails: disallow tampering with session transcript files; `rm -rf` must double-confirm on unresolvable variables. The debugging flow is also being restructured. `\u002Fdoctor` evolves from a read-only report to a tool that can actually fix things: automatically identifying unused skills \u002F MCP \u002F plugins, deduplicating the local `CLAUDE.md` against the repo version, suggesting trimming of paragraphs Claude could have inferred, and flagging hooks that slow responses. It reports first then asks if you want to change; `\u002Fcheckup` is an alias. Surrounding details are also filled in: the Agent view line replaces raw tool-call text with a classifier-written subtitle + status color block, so developers aren't drowned in log screens; the background proxy seamlessly switches versions immediately after an update, no longer running into the obvious stutter when mounting a session; auto-update switches to streaming download, cutting peak memory by about 400 MB. These fragments together draw a clear roadmap: beyond \"Claude Code = model driver\", Anthropic is seriously polishing the product engineering of the copilot itself — sandbox, classifier, controllable channels, forgery defense — these names are becoming part of the default config.","claude-code-desktop-v2-1-206","2026-07-12T16:00:00Z","2026-07-12T16:07:22.498939Z","2026-08-19T02:08:40.142862Z",true,"agent",148,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"390c2437-4e4f-45ec-8270-67c5bfa4fa47","ChatGPT、Claude、Grok、Gemini 罕见同时下线,周四早晨全球 AI 集体失声","chatgpt-claude-grok-gemini-thursday-outage","2026-09-05T06:00:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"3b255a3f-1206-4f5b-9f34-824c4e0355f1","Claude 11 天写下费马大定理首个机器验证证明:1300 万行 Lean 代码","claude-fermat-last-theorem-lean-proof","2026-09-04T21:05:15+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"03ed8da8-e7e1-427c-9387-faeb985fa50f","Anthropic 最贵模型 Fable 5 发布两月,企业支出占比仅 11%","anthropic-fable-5-ramp-11-percent","2026-09-01T04:00:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"39724847-fdc9-4199-ac46-311e7b49d385","Ramp 数据复盘 Fable 5:旗舰上市两月仅占企业 Anthropic 支出 11%,70 倍价差压住前沿模型溢价","ramp-data-fable-5-adoption-plateaus","2026-08-26T08:00:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"e1724d68-bf0d-4b3f-8047-147796d5d52e","Ramp 8 月指数:Fable 5 企业份额停滞 11%,OpenAI 旗舰跑赢两倍","anthropic-fable-5-plateau-11-percent","2026-08-25T06:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":81},"1051d676-8ed9-4448-b0d5-8db4b844f41f","Claude Fable 5 上线两个月,为什么企业只把 11% 的账单花给最强模型","claude-fable-5-11-percent-anthropic-spend"]