[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-2026-ai-roadmap-repo-ml-agentic-ai-zh":3,"article-related-2026-ai-roadmap-repo-ml-agentic-ai-zh":30,"series-tools-e6f5e2d9-233e-4103-83e2-2c9e09883ec7":88},{"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":11,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":11},"e6f5e2d9-233e-4103-83e2-2c9e09883ec7","2026-ai-roadmap-repo-ml-agentic-ai-zh","2026 AI 路線圖：從 ML 到 Agent","\u003Cp>一個 GitHub repo 只有 \u003Cstrong>1 顆星\u003C\u002Fstrong>，卻想做很大。它把從數學基礎到 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fkodigitaccount\u002F2026-ROADMAP-FOR-ADVANCE-ML-AI-GENERATIVE-AI-AGENTIC-AI\" target=\"_blank\" rel=\"noopener\">agentic AI\u003C\u002Fa> 的路，直接畫成一張學習地圖。\u003C\u002Fp>\u003Cp>這個專案叫 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fkodigitaccount\u002F2026-ROADMAP-FOR-ADVANCE-ML-AI-GENERATIVE-AI-AGENTIC-AI\" target=\"_blank\" rel=\"noopener\">2026-ROADMAP-FOR-ADVANCE-ML-AI-GENERATIVE-AI-AGENTIC-AI\u003C\u002Fa>。講白了，它在回答一個很實際的問題：2026 年想做 ML、GenAI，或 agent 工作，該先學什麼。\u003C\u002Fp>\u003Cp>我覺得這種 repo 很容易被小看。因為網路上「30 天學 \u003Ca href=\"\u002Fnews\u002Fspec-driven-ai-turns-mcp-into-workflow-engine-zh\">AI\u003C\u002Fa>」太多了。真的能用的，通常不是最炫的那個，而是把路徑排清楚的那個。\u003C\u002Fp>\u003Ch2>這份路線圖到底寫了什麼\u003C\u002Fh2>\u003Cp>這份 README 把 2026 年切成幾段。從 1 月到 3 月打基礎，最後一路走到 10 月到 12 月的 agentic AI。這種編排很務實，因為 AI 工作本來就不是一口氣跳到最前面。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775199980318-nz4a.png\" alt=\"2026 AI 路線圖：從 ML 到 Agent\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>很多人一開始就衝 prompt engineering。結果碰到資料清理、模型評估、部署延遲，就整個卡住。說真的，這很常見。\u003C\u002Fp>\u003Cp>這個 roadmap 先放的是硬底子。像是線性代數、機率、最佳化、Python、SQL、Git、Docker，還有基本系統設計。接著才進到監督式學習、非監督式學習、時間序列和部署。\u003C\u002Fp>\u003Cp>後面才是深度學習、生成式 AI，最後才碰 agent 框架。像是 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flanggraph\" target=\"_blank\" rel=\"noopener\">LangGraph\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen\" target=\"_blank\" rel=\"noopener\">AutoGen\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FcrewAIInc\u002FcrewAI\" target=\"_blank\" rel=\"noopener\">CrewAI\u003C\u002Fa>，還有 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fsmolagents\" target=\"_blank\" rel=\"noopener\">smolagents\u003C\u002Fa>。\u003C\u002Fp>\u003Cul>\u003Cli>基礎：線代、機率、最佳化、Python、SQL、Git、Docker\u003C\u002Fli>\u003Cli>ML：回歸、分類、分群、降維、時間序列\u003C\u002Fli>\u003Cli>深度學習：CNN、RNN、attention、Transformer、PyTorch\u003C\u002Fli>\u003Cli>GenAI：LLM、diffusion、fine-tuning、RLHF、prompt pattern\u003C\u002Fli>\u003Cli>Agentic AI：工具使用、記憶、規劃、人機協作流程\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這個順序其實很像真實職場。招 ML 工程師時，主管通常先看資料處理、評估、部署。不是先問你會不會背模型名單。\u003C\u002Fp>\u003Cp>做 GenAI 的人也一樣。你得懂 retrieval、推論成本、延遲，還有\u003Ca href=\"\u002Fnews\u002Fmcp-explained-from-prompts-to-production-zh\">怎麼把\u003C\u002Fa>模型接到外部工具。只會聊天，通常沒辦法上線。\u003C\u002Fp>\u003Cp>這份 repo 也列了一串工具。像是 \u003Ca href=\"https:\u002F\u002Fmlflow.org\" target=\"_blank\" rel=\"noopener\">MLflow\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fdvc.org\" target=\"_blank\" rel=\"noopener\">DVC\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers\" target=\"_blank\" rel=\"noopener\">Hugging Face Transformers\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fhuggingface.co\" target=\"_blank\" rel=\"noopener\">Hugging Face Hub\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Ffastapi.tiangolo.com\" target=\"_blank\" rel=\"noopener\">FastAPI\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fstreamlit.io\" target=\"_blank\" rel=\"noopener\">Streamlit\u003C\u002Fa>。再加上 AWS、GCP、Azure 的部署思維。\u003C\u002Fp>\u003Cp>這代表作者不是只想寫筆記，而是真的有在看 production。這點我覺得加分很多。\u003C\u002Fp>\u003Ch2>為什麼它把 2026 放進標題\u003C\u002Fh2>\u003Cp>標題寫 2026，但重點其實是現在就開始變的工作型態。AI 已經從單一模型 demo，走向會規劃、會呼叫工具、會保留狀態的系統。\u003C\u002Fp>\u003Cp>這差很多。聊天機器人只會回答問題。\u003Ca href=\"\u002Fnews\u002Famazon-bedrock-agents-multi-agent-workflows-zh\">Agen\u003C\u002Fa>t 則可能幫你查 repo、整理內容、發起動作，甚至串起一整段工作流。\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fopenai.com\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa> 在官方文章裡寫過一句話：\u003Cblockquote>“We believe that there is a lot of value in AI systems that can take actions on behalf of users.”\u003C\u002Fblockquote>這句話很直白。意思就是，AI 不只要會講，還要會做。\u003C\u002Fp>\u003Cp>路線圖裡也提到 \u003Ca href=\"https:\u002F\u002Fai.google\" target=\"_blank\" rel=\"noopener\">Google Gemini\u003C\u002Fa>，以及 \u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fai-data-science\u002Fproducts\u002Fnim-microservices\u002F\" target=\"_blank\" rel=\"noopener\">NVIDIA NIM\u003C\u002Fa>。這透露一個現實：現在的 AI stack 變得更模組化了。\u003C\u002Fp>\u003Cp>很多團隊不再只押單一大模型。它們會混用 hosted API、open-weight model、retrieval layer，還有 workflow engine。這種做法很務實，也很像真正的工程世界。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fopenai.com\" target=\"_blank\" rel=\"noopener\">OpenAI\u003C\u002Fa>：API 與 agent 方向\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fai.google\" target=\"_blank\" rel=\"noopener\">Google Gemini\u003C\u002Fa>：多模態模型與 API\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fhuggingface.co\" target=\"_blank\" rel=\"noopener\">Hugging Face\u003C\u002Fa>：模型、資料集、部署工具\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fwww.nvidia.com\u002Fen-us\u002Fai-data-science\u002Fproducts\u002Fnim-microservices\u002F\" target=\"_blank\" rel=\"noopener\">NVIDIA NIM\u003C\u002Fa>：企業用推論服務\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flanggraph\" target=\"_blank\" rel=\"noopener\">LangGraph\u003C\u002Fa>：agent 流程編排\u003C\u002Fli>\u003C\u002Ful>\u003Cp>對學習者來說，這件事很重要。因為真正能長期吃飯的技能，不是只綁某一家產品。\u003C\u002Fp>\u003Cp>Python、評估方法、資料處理、部署、系統思維，這些才是通用貨幣。今天你用 GPT API，明天換 Llama，後天換內部模型，底層思路還是差不多。\u003C\u002Fp>\u003Ch2>它跟真實工作有多像\u003C\u002Fh2>\u003Cp>這份 repo 最強的地方，是它沒有停在模型理論。它列了很多實作題目。像是 churn prediction、loan default modeling、image classification、neural machine translation、text-to-image generation，還有 code review agent。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775199980223-ndyw.png\" alt=\"2026 AI 路線圖：從 ML 到 Agent\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這些題目很像真實團隊會做的事。不是那種只會在簡報裡發光的題目，而是會碰到資料髒、指標亂、上線後要維運的題目。\u003C\u002Fp>\u003Cp>你看職場就知道了。資料科學家可能花一週清資料，再花一週跟 PM 解釋指標。ML 工程師常常一半時間在管 CI\u002FCD、feature pipeline。GenAI 工程師則可能卡在 retrieval 品質、prompt 測試、或推論延遲。\u003C\u002Fp>\u003Cp>這份清單也提到 \u003Ca href=\"https:\u002F\u002Fwandb.ai\" target=\"_blank\" rel=\"noopener\">Weights & Biases\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.docker.com\" target=\"_blank\" rel=\"noopener\">Docker\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fkubernetes.io\" target=\"_blank\" rel=\"noopener\">Kubernetes\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.snowflake.com\" target=\"_blank\" rel=\"noopener\">Snowflake\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.databricks.com\" target=\"_blank\" rel=\"noopener\">Databricks\u003C\u002Fa>，還有 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Ffaiss\" target=\"_blank\" rel=\"noopener\">FAISS\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fwww.pinecone.io\" target=\"_blank\" rel=\"noopener\">Pinecone\u003C\u002Fa> 這類向量資料庫。\u003C\u002Fp>\u003Cp>這些工具放在一起看，就知道 2026 的 AI 工作長什麼樣。模型訓練、實驗追蹤、部署、檢索、監控，全部都要顧。\u003C\u002Fp>\u003Cp>我整理一個很直接的對比：\u003C\u002Fp>\u003Cul>\u003Cli>只看 notebook：上手快，但不太會部署\u003C\u002Fli>\u003Cli>做專案：慢一點，但更接近面試需求\u003C\u002Fli>\u003Cli>照 roadmap 學：前提是你真的有產出\u003C\u002Fli>\u003Cli>做 agent：要處理 API、狀態、錯誤、評估\u003C\u002Fli>\u003C\u002Ful>\u003Cp>repo 也列出幾個職稱。像是 ML Engineer、AI Engineer、GenAI Engineer、Agentic AI Engineer、LLM Fine-Tuning Specialist、MLOps Engineer、AI-focused Data Scientist。\u003C\u002Fp>\u003Cp>這些名稱現在已經滿地都是了。但我覺得最後贏的人，不會是最會背名詞的人，而是最會講 tradeoff 的人。\u003C\u002Fp>\u003Cblockquote>“We believe that there is a lot of value in AI systems that can take actions on behalf of users.” — Sam Altman, OpenAI\u003C\u002Fblockquote>\u003Ch2>這份路線圖值不值得跟\u003C\u002Fh2>\u003Cp>值，但有一個前提。你不能只把它當成閱讀清單。你要把它變成產出清單。\u003C\u002Fp>\u003Cp>如果你花 6 個月只收集概念，卻沒有做出一個分類器、一個 retrieval app、一個 agent workflow，那你還是會覺得自己什麼都不會。這點很現實。\u003C\u002Fp>\u003Cp>這份 repo 的好處，是它幫你把混亂的領域排出順序。先學數學，再學工具，再做專案，最後才把模型接到真實任務上。\u003C\u002Fp>\u003Cp>如果你現在要開始，我會建議這樣做：先做一個 tabular ML 專案，再做一個 PyTorch 實驗，接著做一個帶 retrieval 的 GenAI app，最後做一個會用工具的 agent。\u003C\u002Fp>\u003Cp>每個專案都要留數字。像是 accuracy、latency、cost、token 用量，或失敗率。沒有數字，很多 AI 專案都只是聊天而已。\u003C\u002Fp>\u003Cp>我自己的判斷很簡單：2026 年最吃香的人，不是追著新模型跑的人，而是能把模型變成產品的人。你如果這個月只能做一件事，那就選一個題目，真的把它做完。\u003C\u002Fp>\u003Cp>然後問自己一句：我能不能不用空話，把一個模型接成能跑的服務？\u003C\u002Fp>","一個只有 1 顆星的 GitHub repo，卻把 2026 年從 ML 基礎、GenAI 到 agentic AI 的學習路線排得很完整。","github.com","https:\u002F\u002Fgithub.com\u002Fkodigitaccount\u002F2026-ROADMAP-FOR-ADVANCE-ML-AI-GENERATIVE-AI-AGENTIC-AI",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775199980318-nz4a.png","tools","zh","ff851a57-4d5f-4e42-962e-ff8aca654710",[17,18,19,20,21,22,23,24,25,26],"AI 路線圖","ML","GenAI","agentic AI","GitHub repo","2026","PyTorch","MLOps","LLM","FastAPI",4,"2026-04-01T09:33:31.750053+00:00","2026-04-03T07:04:12.877378+00:00",{"tags":31,"relatedLang":47,"relatedPosts":51},[32,33,35,37,39,41,43,45],{"name":22,"slug":22},{"name":18,"slug":34},"ml",{"name":17,"slug":36},"ai-路線圖",{"name":24,"slug":38},"mlops",{"name":25,"slug":40},"llm",{"name":19,"slug":42},"genai",{"name":21,"slug":44},"github-repo",{"name":23,"slug":46},"pytorch",{"id":15,"slug":48,"title":49,"language":50},"2026-ai-roadmap-repo-ml-agentic-ai-en","2026 AI roadmap repo maps ML to agentic AI","en",[52,58,64,70,76,82],{"id":53,"slug":54,"title":55,"cover_image":56,"image_url":56,"created_at":57,"category":13},"8520cd4f-2531-4808-a95d-26f590239d7a","500-ai-agent-projects-show-where-agents-work-now-zh","500 個 AI agent 專案，現在能做什麼","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781033591132-c0nh.png","2026-06-09T19:32:37.03924+00:00",{"id":59,"slug":60,"title":61,"cover_image":62,"image_url":62,"created_at":63,"category":13},"c557ef1c-7fde-4c86-918e-4fb9680ee9df","chocolatey-go-package-policy-installs-zh","Chocolatey 的 Go 安裝變成政策","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781029110289-xkbh.png","2026-06-09T18:18:05.078435+00:00",{"id":65,"slug":66,"title":67,"cover_image":68,"image_url":68,"created_at":69,"category":13},"90b2df54-df6e-417d-9e16-91e9ad2f53d7","go-support-policy-turns-releases-into-a-checklist-zh","Go 支援政策把發版變清單","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781028200122-3m4u.png","2026-06-09T18:02:49.50176+00:00",{"id":71,"slug":72,"title":73,"cover_image":74,"image_url":74,"created_at":75,"category":13},"119c23c6-8ae7-4c4e-820e-1eba0730d702","rustdesk-self-hosting-secure-remote-access-zh","RustDesk 自架遠端存取部署指南","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781017373324-g7et.png","2026-06-09T15:02:24.118819+00:00",{"id":77,"slug":78,"title":79,"cover_image":80,"image_url":80,"created_at":81,"category":13},"b84491ba-e4af-4581-8c04-1890df39a1ad","aider-open-source-coding-agent-repo-edits-zh","Aider 讓開源編碼變成 repo 編輯","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781013817997-l4il.png","2026-06-09T14:02:56.179093+00:00",{"id":83,"slug":84,"title":85,"cover_image":86,"image_url":86,"created_at":87,"category":13},"b6bc009f-238c-4466-b7ec-c7085c7fdbe8","wwdc-2026-rumors-siri-assistant-ios-27-zh","WWDC 2026 讓 Siri 變助手","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781007517876-bmuu.png","2026-06-09T12:18:03.608802+00:00",[89,94,99,104,109,114,119,124,129,134],{"id":90,"slug":91,"title":92,"created_at":93},"855cd52f-6fab-46cc-a7c1-42195e8a0de4","surepath-real-time-mcp-policy-controls-zh","SurePath 推出即時 MCP 政策控管","2026-03-26T07:57:40.77233+00:00",{"id":95,"slug":96,"title":97,"created_at":98},"9b19ab54-edef-4dbd-9ce4-a51e4bae4ebb","mcp-in-2026-the-ai-tool-layer-teams-use-zh","2026 年 MCP：團隊真的在用的 AI 工具層","2026-03-26T08:01:46.589694+00:00",{"id":100,"slug":101,"title":102,"created_at":103},"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","2026-03-26T08:08:12.453028+00:00",{"id":105,"slug":106,"title":107,"created_at":108},"05553086-6ed0-4758-81fd-6cab24b575e0","garry-tan-open-sources-claude-code-toolkit-zh","Garry Tan 開源 Claude Code 工具包","2026-03-26T08:26:20.068737+00:00",{"id":110,"slug":111,"title":112,"created_at":113},"042a73a2-18a2-433d-9e8f-9802b9559aac","github-ai-projects-to-watch-in-2026-zh","2026 必看 20 個 GitHub AI 專案","2026-03-26T08:28:09.619964+00:00",{"id":115,"slug":116,"title":117,"created_at":118},"a5f94120-ac0d-4483-9a8b-63590071ac6a","claude-code-vs-cursor-2026-zh","Claude Code 與 Cursor 深度對比：202…","2026-03-26T13:27:14.279193+00:00",{"id":120,"slug":121,"title":122,"created_at":123},"0975afa1-e0c7-4130-a20d-d890eaed995e","practical-github-guide-learning-ml-2026-zh","2026 機器學習入門 GitHub 實用指南","2026-03-27T01:16:49.712576+00:00",{"id":125,"slug":126,"title":127,"created_at":128},"bfdb467a-290f-4a80-b3a9-6f081afb6dff","aiml-2026-student-ai-ml-lab-repo-review-zh","AIML-2026：像課綱的學生實驗 Repo","2026-03-27T01:21:51.467798+00:00",{"id":130,"slug":131,"title":132,"created_at":133},"80cabc3e-09fc-4ff5-8f07-b8d68f5ae545","ai-trending-github-repos-and-research-feeds-zh","AI Trending：把 AI 資源收成一張表","2026-03-27T01:31:35.262183+00:00",{"id":135,"slug":136,"title":137,"created_at":138},"3ce6e6e2-bac5-463e-9f8d-45caabcc61f7","awesome-ai-for-science-research-tools-map-zh","AI 科研工具清單，開始像地圖了","2026-03-27T01:46:50.521945+00:00"]