[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"tag-ai-工作流":3},{"tag":4,"articles":11,"peer_article_count":59},{"id":5,"name":6,"slug":7,"article_count":8,"description_zh":9,"description_en":10},"7ce353aa-84e6-4b2a-96e2-f341690a3ff9","AI 工作流","ai-工作流",3,"AI 工作流關注的是把模型從單次問答，接到可重複執行的任務鏈：資料讀取、工具呼叫、人工確認、格式驗證與錯誤回復。從 MCP、長任務代理到提示規格化，重點是讓 AI 進入正式開發與營運流程。","AI workflows focus on turning model output into repeatable task chains: data access, tool calls, human approval, format checks, and failure recovery. Topics here include MCP, long-running agents, and prompt specs that fit real development and operations pipelines.",[12,21,29,36,44,51],{"id":13,"slug":14,"title":15,"summary":16,"category":17,"image_url":18,"cover_image":18,"language":19,"created_at":20},"254f12ec-ff29-4308-a158-3e1c2bc193e0","codex-chatgpt-work-code-assistant-zh","Codex把 ChatGPT 拉進工作流","OpenAI 的 Codex 把 ChatGPT 拉進工作和寫程式流程，主打 KPI、簡報、客戶摘要、bug triage 與 prototype 草稿。","ai-agent","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781115477227-mis0.png","zh","2026-06-10T18:17:26.326284+00:00",{"id":22,"slug":23,"title":24,"summary":25,"category":26,"image_url":27,"cover_image":27,"language":19,"created_at":28},"2e4d54e9-b481-44c9-834a-a4c8bcf85228","5-codex-new-features-openai-direction-zh","5 個 Codex 新功能看 OpenAI 方向","5 個 Codex 更新加上 PowerPoint 測試，顯示 OpenAI 正把 AI 變成能長時間執行工作的夥伴。","industry","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1779822353841-dnuw.png","2026-05-26T19:05:23.852841+00:00",{"id":30,"slug":31,"title":32,"summary":33,"category":26,"image_url":34,"cover_image":34,"language":19,"created_at":35},"c0b7cf36-670a-499e-9680-45ffa7531323","why-anthropics-small-business-push-is-right-zh","為什麼 Anthropic 押注小型企業是對的 AI 賭局","Anthropic 押注小型企業是對的，因為 SMB 工作流才是下一個真正能規模化的 AI 市場。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778883044236-877k.png","2026-05-15T22:10:25.72013+00:00",{"id":37,"slug":38,"title":39,"summary":40,"category":41,"image_url":42,"cover_image":42,"language":19,"created_at":43},"97f9c411-2849-4a70-9d1e-3bba0dde23bf","claude-opus-4-7-release-workflow-vision-zh","Claude Opus 4.7 上線：更會做事了","Anthropic 推出 Claude Opus 4.7，強化長任務、視覺理解與程式工作流，但 Token 消耗也更高。","model-release","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1776859436510-gf5y.png","2026-04-22T12:03:36.654584+00:00",{"id":45,"slug":46,"title":47,"summary":48,"category":17,"image_url":49,"cover_image":49,"language":19,"created_at":50},"96d8e8c8-1edd-475d-9145-b1e7a1b02b65","mcp-explained-from-prompts-to-production-zh","MCP 怎麼把提示詞變工作流","MCP 讓 AI 能用標準介面讀資料、叫工具、要確認。Zoho 用 elicitation 拆解它怎麼把助手帶進正式工作流程。","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775050818756-mope.png","2026-04-01T09:24:39.321274+00:00",{"id":52,"slug":53,"title":54,"summary":55,"category":56,"image_url":57,"cover_image":57,"language":19,"created_at":58},"af9c46c3-7a28-410b-9f04-32b3de30a68c","prompting-in-2026-what-actually-works-zh","2026 提示工程，真正有用的是什麼","2026 年的提示工程更吃模型差異。資料顯示，38.5% 對話要靠反覆修正才成功。真正有效的方法不是花式 wording，而是把提示寫成精簡規格，配合限制條件、格式要求與驗證流程。","tools","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1774517388287-tt7a.png","2026-03-26T08:08:12.453028+00:00",0]