[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-entire-agent-git-network-ai-code-trust-zh":3,"article-related-entire-agent-git-network-ai-code-trust-zh":32,"series-industry-0254d671-ea12-452e-9d96-681f8cc68e92":79},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":24,"views":28,"created_at":29,"published_at":30,"topic_cluster_id":31},"0254d671-ea12-452e-9d96-681f8cc68e92","entire-agent-git-network-ai-code-trust-zh","Entire 的 Git 網路把 AI 程式風險收緊了","\u003Cp>Entire 的 Git 式 \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa> 網路，真的能讓 AI 寫程式更可信嗎？\u003C\u002Fp>\u003Cp data-speakable=\"summary\">Entire 正在做一個像 Git 的 agent 網路，結合隔離執行與驗證流程，讓 AI \u003Ca href=\"\u002Fnews\u002Fopenai-54-token-efficiency-ai-coding-battleground-zh\">寫碼\u003C\u002Fa>更容易審核。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>規格 A\u003C\u002Fth>\u003Cth>規格 B\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>Git 式 agent 網路\u003C\u002Ftd>\u003Ctd>可追蹤、可回放的動作紀錄\u003C\u002Ftd>\u003Ctd>方便審計與比對變更\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>WebAssembly 執行環境\u003C\u002Ftd>\u003Ctd>沙箱化執行\u003C\u002Ftd>\u003Ctd>降低未受信任程式的風險\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>驗證層\u003C\u002Ftd>\u003Ctd>合併或使用前先檢查\u003C\u002Ftd>\u003Ctd>提早攔下錯誤輸出\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>多 agent 協作模型\u003C\u002Ftd>\u003Ctd>共享執行與審查流程\u003C\u002Ftd>\u003Ctd>適合團隊式開發\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>1. Git 式 agent 網路\u003C\u002Fh2>\u003Cp>Entire 想把 agent 的工作方式，從黑盒子改成像 \u003Ca href=\"https:\u002F\u002Fgit-scm.com\u002F\">Git\u003C\u002Fa> 一樣可追蹤的流程。重點不是讓 AI 只會「產生答案」，而是讓它每一步動作都有紀錄，能被檢視、比較和回放。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783645361520-n3d6.png\" alt=\"Entire 的 Git 網路把 AI 程式風險收緊了\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這對 agent 寫碼很重要，因為一旦 AI 能改檔案、跑命令或呼叫服務，團隊就會立刻想知道三件事：改了什麼、誰改的、能不能信。\u003C\u002Fp>\u003Cul>\u003Cli>把 agent 動作拆成可追蹤事件\u003C\u002Fli>\u003Cli>支援回放與審查\u003C\u002Fli>\u003Cli>更貼近程式團隊的工作流\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>2. WebAssembly 隔離執行\u003C\u002Fh2>\u003Cp>這套架構的另一個核心，是用 \u003Ca href=\"https:\u002F\u002Fwebassembly.org\u002F\">WebAssembly\u003C\u002Fa> 把執行環境關\u003Ca href=\"\u002Fnews\u002Ftesla-model-y-l-fills-the-model-x-gap-zh\">起來\u003C\u002Fa>。當 agent 生成的程式碼或工具不一定可信時，沙箱就成了必要的安全邊界，而不是附加功能。\u003C\u002Fp>\u003Cp>這也是它比一般 workflow 方案更像基礎設施的地方。只要 agent 會跑程式，runtime 就是安全的一部分；Wasm 讓平台團隊能更清楚地限制它能碰到什麼。\u003C\u002Fp>\u003Cul>\u003Cli>限制對主機資源的存取\u003C\u002Fli>\u003Cli>建立更小的執行邊界\u003C\u002Fli>\u003Cli>可用在 agent 生成的程式與工具上\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>3. 先驗證，再放行\u003C\u002Fh2>\u003Cp>Entire 的設計也把驗證放在前面。這回應了 \u003Ca href=\"\u002Ftag\u002Fllm\">LLM\u003C\u002Fa> 寫碼常見的問題：輸出看起來合理，不代表真的可以直接用。先檢查、再接受，才是比較穩的做法。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783645363891-h2d7.png\" alt=\"Entire 的 Git 網路把 AI 程式風險收緊了\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>實作上，它更接近既有的 \u003Ca href=\"\u002Ftag\u002Fcode-review\">code review\u003C\u002Fa>、測試和 CI，只是現在參與者從人變成了 agent。對工程團隊來說，這種流程不陌生，差別只是多了一層自動化與風險控管。\u003C\u002Fp>\u003Ccode>agent_output -> verify -> approve -> execute\u003C\u002Fcode>\u003Ch2>4. 多 agent 協作模型\u003C\u002Fh2>\u003Cp>Git 網路的意思，不只是單一 agent 工作，而是多個 agent 共享狀態、接力處理任務。這很適合一個 agent 負責寫草稿，另一個負責測試，第三個負責檢查政策或安全規則。\u003C\u002Fp>\u003Cp>對平台團隊來說，這種模型的吸引力在於：可以讓 AI 幫忙，但不把控制權交出去。當 agentic workflows 變成主流，真正的問題不再只是「它會不會寫」，而是「系統能不能管住它寫出來的東西」。\u003C\u002Fp>\u003Cul>\u003Cli>支援 agent 之間的交接\u003C\u002Fli>\u003Cli>保留共享審計軌跡\u003C\u002Fli>\u003Cli>比單人試玩更適合團隊導入\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>5. 現在為什麼這個缺口重要\u003C\u002Fh2>\u003Cp>這篇消息的重點其實很直接：\u003Ca href=\"\u002Ftag\u002Fai-agent\">AI agent\u003C\u002Fa> 一旦能執行程式，風險就會比聊天模型高很多，而多數現有架構還沒把這個風險關得夠緊。Entire 的主張是，用 WebAssembly 和驗證流程補上這個洞，同時不要把生產力犧牲掉。\u003C\u002Fp>\u003Cp>所以它不只是某家新創的產品敘事，也是在反映 \u003Ca href=\"\u002Ftag\u002Fai-\">AI 基礎設施\u003C\u002Fa>的方向轉變。大家開始問的，不再只是模型能說什麼，而是 agent 在\u003Ca href=\"\u002Fnews\u002Ftesla-model-y-l-family-ev-premium-real-zh\">真實\u003C\u002Fa>系統裡，到底能安全做什麼。\u003C\u002Fp>\u003Cul>\u003Cli>執行能力已經是新風險面\u003C\u002Fli>\u003Cli>沙箱化會越來越像預設要求\u003C\u002Fli>\u003Cli>可審計性正在變成 AI 基礎設施的一部分\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>哪種適合你\u003C\u002Fh2>\u003Cp>如果你在做給軟體團隊用的 agent 工具，Entire 這套思路最適合需要審計軌跡、受控執行、以及先審後放行的場景。它不是在比模型有多強，而是在補治理能力。\u003C\u002Fp>\u003Cp>如果你只在意模型本身的輸出品質，這不是主角；但如果你想讓 agent 能跑程式，又不想把 runtime 變成安全賭局，Git 加上 WebAssembly 的方向就值得盯緊。\u003C\u002Fp>","4 個設計看點看 Entire 的 Git 式 agent 網路，如何用 WebAssembly 隔離、驗證流程和共享審計軌跡降低 AI 寫碼風險。","thenewstack.io","https:\u002F\u002Fthenewstack.io\u002Fentire-git-for-agents\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783645361520-n3d6.png","industry","zh","f859ed0b-8852-4999-9400-c8af204f7491",[17,18,19,20,21,22,23],"Entire","Git","WebAssembly","AI agent","程式安全","驗證流程","多 agent 協作",[25,26,27],"Git 式紀錄讓 agent 動作可追蹤、可回放，降低黑盒風險。","WebAssembly 沙箱把執行邊界收緊，適合未受信任的 agent 程式。","先驗證再放行的流程，讓 AI 寫碼更接近可治理的工程工作流。",0,"2026-07-10T01:02:19.835515+00:00","2026-07-10T01:02:19.81+00:00","0bc2562a-0fb1-480f-8d3f-fa975d4ad3af",{"tags":33,"relatedLang":38,"relatedPosts":42},[34,36],{"name":19,"slug":35},"webassembly",{"name":20,"slug":37},"ai-agent",{"id":15,"slug":39,"title":40,"language":41},"entire-agent-git-network-ai-code-trust-en","Entire’s agent Git network fixes AI code trust","en",[43,49,55,61,67,73],{"id":44,"slug":45,"title":46,"cover_image":47,"image_url":47,"created_at":48,"category":13},"200777f4-9d68-4b6b-abb4-7bc15e7d6e78","openai-gov-partnerships-access-policy-zh","OpenAI把存取權做成政策","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783684998337-83za.png","2026-07-10T12:02:53.475221+00:00",{"id":50,"slug":51,"title":52,"cover_image":53,"image_url":53,"created_at":54,"category":13},"4e7f9fda-2104-4ed5-895c-5d21b78341fb","kubernetes-ai-assisted-maintainership-rules-zh","Kubernetes替AI維護劃規則","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783672388328-9fvp.png","2026-07-10T08:32:42.07423+00:00",{"id":56,"slug":57,"title":58,"cover_image":59,"image_url":59,"created_at":60,"category":13},"fe489684-cb2b-4bbf-8632-d38490fb7f48","byoa-vibe-coding-apps-only-path-zh","BYOA 才是 vibe coding 的正解","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783670583486-kf5y.png","2026-07-10T08:02:30.385787+00:00",{"id":62,"slug":63,"title":64,"cover_image":65,"image_url":65,"created_at":66,"category":13},"c7de2080-60bd-4c7e-b7dd-841217286f81","model-companies-are-eating-software-industry-zh","模型公司正在吃掉软件行业，应用层护城河正在塌陷","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783668784221-1pzn.png","2026-07-10T07:32:36.195659+00:00",{"id":68,"slug":69,"title":70,"cover_image":71,"image_url":71,"created_at":72,"category":13},"03cdfc0d-bff7-42c8-b7de-bea35eb1b933","5-ge-ml-xue-xi-pai-hao-shun-xu-de-zi-yuan-zh","5 個把 ML 學習排好順序的資源","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783650769467-chf3.png","2026-07-10T02:32:23.014288+00:00",{"id":74,"slug":75,"title":76,"cover_image":77,"image_url":77,"created_at":78,"category":13},"b5112980-dea0-4e30-8fea-d5b49d64b637","openai-54-token-efficiency-ai-coding-battleground-zh","OpenAI 的 54% token 效率提升，才是 AI 寫碼真正戰場","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783643566365-if59.png","2026-07-10T00:32:20.466027+00:00",[80,85,90,95,100,105,110,115,120,125],{"id":81,"slug":82,"title":83,"created_at":84},"ee073da7-28b3-4752-a319-5a501459fb87","ai-in-2026-what-actually-matters-now-zh","2026 AI 真正重要的事","2026-03-26T07:09:12.008134+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"83bd1795-8548-44c9-9a7e-de50a0923f71","trump-ai-framework-power-speech-state-preemption-zh","川普 AI 框架瞄準電力、言論與州權","2026-03-26T07:12:18.695466+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"ea6be18b-c903-4e54-97b7-5f7447a612e0","nvidia-gtc-2026-big-ai-announcements-zh","NVIDIA GTC 2026 重點拆解","2026-03-26T07:14:26.62638+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"4bcec76f-4c36-4daa-909f-54cd702f7c93","claude-users-spreading-out-and-getting-better-zh","Claude 用戶更分散，也更會用","2026-03-26T07:22:52.325888+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"bd903b15-2473-4178-9789-b7557816e535","openclaw-raises-hard-question-for-ai-models-zh","OpenClaw 逼問 AI 模型價值","2026-03-26T07:24:54.707486+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"eeac6b9e-ad9d-4831-8eec-8bba3f9bca6a","gap-google-gemini-checkout-fashion-search-zh","Gap 把結帳搬進 Gemini","2026-03-26T07:28:23.937768+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"0740e53f-605d-4d57-8601-c10beb126f3c","google-pushes-gemini-transition-to-march-2026-zh","Google 把 Gemini 轉換延到 2026 年 3…","2026-03-26T07:30:12.825269+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"e660d801-2421-4529-8fa9-86b82b066990","metas-llama-4-benchmark-scandal-gets-worse-zh","Meta Llama 4 分數風波又擴大","2026-03-26T07:34:21.156421+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"183f9e7c-e143-40bb-a6d5-67ba84a3a8bc","accenture-mistral-ai-sovereign-enterprise-deal-zh","Accenture 攜手 Mistral AI 賣主權 AI","2026-03-26T07:38:14.818906+00:00",{"id":126,"slug":127,"title":128,"created_at":129},"191d9b1b-768a-478c-978c-dd7431a38149","mistral-ai-faces-its-hardest-year-yet-zh","Mistral AI 迎來最硬的一年","2026-03-26T07:40:23.716374+00:00"]