[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-github-last30days-skill-ai-research-model-zh":3,"article-related-github-last30days-skill-ai-research-model-zh":30,"series-ai-agent-becd8e42-53b8-4250-8989-d3c8f60ea909":75},{"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":22,"views":26,"created_at":27,"published_at":28,"topic_cluster_id":29},"becd8e42-53b8-4250-8989-d3c8f60ea909","github-last30days-skill-ai-research-model-zh","GitHub 的 last30days skill 才是 AI 研究的正確模型","\u003Cp data-speakable=\"summary\">\u003Ca href=\"\u002Ftag\u002Fgithub\">GitHub\u003C\u002Fa> 的 last30days skill 是 AI 研究更好的模型，因為它優先抓取即時的人類訊號，而不是過時的網頁內容。\u003C\u002Fp>\u003Cp>GitHub 的 \u003Ccode>last30days-skill\u003C\u002Fcode> 應該成為 AI 研究工具的預設範式：先抓活的訊號，再把它們整理成有根據的摘要。這個倉庫的賭注很明確，它從 Reddit、X、YouTube、Hacker News、Polymarket、GitHub 和一般網頁蒐集資料，再依照互動和金錢訊號排序，而不是看頁面權重或編輯包裝。這很重要，因為 AI、產品、文化這些變動最快的\u003Ca href=\"\u002Fnews\u002F5-ai-agent-papers-worth-tracking-zh\">主題\u003C\u002Fa>，已經不能只靠搜尋結果理解。真正有用的是最近一個月的留言、討論串、逐字稿、提交紀錄和預測市場機率。\u003C\u002Fp>\u003Ch2>第一個論點\u003C\u002Fh2>\u003Cp>它比傳統搜尋更快抓到現實。\u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> 和靜態網頁搜尋是用來找頁面，不是用來量測當下的共識。這個倉庫的 README 直接點出差異：Google 在聚合編輯與索引，而 \u003Ccode>\u002Flast30days\u003C\u002Fcode> 在搜尋人。這不是修辭差別，而是方法論差別。一篇部落格可以很快被收錄，卻完全說不出現在使用者在抱怨什麼、開發者這週在\u003Ca href=\"\u002Fnews\u002Fopenai-partner-network-delivery-strategy-zh\">交付\u003C\u002Fa>什麼、社群已經否決了什麼。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781752672625-a8f4.png\" alt=\"GitHub 的 last30days skill 才是 AI 研究的正確模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>它自己的使用案例更能說明問題。README 描述了一個會前研究流程：一般 Google 搜尋會先回來 2023 年的 LinkedIn 個人頁，但 \u003Ccode>\u002Flast30days\u003C\u002Fcode> 會抓到最近的推文、Podcast 逐字稿、GitHub PR 和 Reddit 爭論。這就是「靜態身份」和「活躍訊號」的差別。對創辦人、PM、工程師來說，真正要問的從來不是「這個人以前說過什麼」，而是「他現在在做什麼、現在相信什麼、以及市場現在怎麼回應」。一個月的視窗夠新鮮，也夠長，足以看出趨勢。\u003C\u002Fp>\u003Ch2>第二個論點\u003C\u002Fh2>\u003Cp>它使用的排序訊號，比 SEO 更接近真實重要性。這個工具的核心主張是，互動比關鍵字優化更能代表相關性。Reddit 的 upvote、HN 的 points、X 的 likes、YouTube 的觀看數、Polymarket 的交易量，都是不同形式的人類投入。一篇 1,500 upvote 的 Reddit 討論、一支 360 萬觀看的短片、或一個有 6.6 萬美元交易量的市場，傳達的\u003Ca href=\"\u002Fnews\u002Ffable-5-drew-rare-praise-ai-voices-zh\">關注\u003C\u002Fa>程度都比一篇打磨精美的公司部落格更有價值。README 沒把這些訊號當噪音，而是當證據，因為它們來自真實注意力與真實分歧。\u003C\u002Fp>\u003Cp>Polymarket 是最能說明這件事的例子。倉庫明確對比的是「意見」和「機率」：不是某人說會發生什麼，而是人們願意拿錢押什麼。這比單純做情緒分析更強，也正好對症當前 \u003Ca href=\"\u002Ftag\u002Fai-工具\">AI 工具\u003C\u002Fa>的老問題，因為很多摘要會把炒作、信念和價格混成一團。當系統依照使用者實際互動來加權時，產出的綜述更難被操縱，也更容易信任。SEO 可以被設計出來，但注意力和金錢要大規模偽造，難得多。\u003C\u002Fp>\u003Ch2>反方可能怎麼說\u003C\u002Fh2>\u003Cp>最強的反對意見是，這種方法會過度迎合「聲量」。Reddit 不是世界，X 上充滿表演式確定感，YouTube 獎勵冗長輸出，預測市場在很多類別都太薄，GitHub stars 也不是品質保證。若系統把互動當成核心信號，就可能把受歡迎誤認成真實，把吵鬧誤認成重要，尤其當受眾本來就偏向早期採用者、開發者或高度線上化社群時。若把它拿來做所有決策，你會只替最吵的人設計產品。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781752668508-99rm.png\" alt=\"GitHub 的 last30days skill 才是 AI 研究的正確模型\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>這個批評成立，而且這個倉庫其實也給了正確答案：它從來沒說單一來源就是真理。它主張的是跨來源綜合，會比任何單一管道更好。這個工具最強的地方，在於它能做交叉驗證。Reddit 討論、HN 辯論、YouTube 逐字稿、GitHub release、Polymarket 走勢放在一起，就能看出哪些是雜訊、哪些是共識、哪些是信念。限制不是互動訊號沒用，而是它們本來就只是局部真相。把它們合成後，可靠度遠高於搜尋排序或編輯摘要。\u003C\u002Fp>\u003Cp>所以反方只打中一半：這不是人氣投票，而是結構化的交叉檢查。它只有在你只看單一來源時才危險；而它的設計正是要避免這件事，因為它把多個激勵不同的系統拉進來，再逼摘要層去對齊。這不是缺點，反而是必要約束。\u003C\u002Fp>\u003Ch2>你能做什麼\u003C\u002Fh2>\u003Cp>如果你是工程師、PM 或創辦人，研究流程就該圍繞即時、多來源訊號，而不是泛用網頁搜尋。任何會前準備、發版前判斷、招募、產品下注，都先看最近 30 天：使用者在抱怨什麼、建造者在交付什麼、社群在吵什麼、哪些人願意付錢或押注。然後把綜合摘要當起點，不是當神諭。真正有優勢的習慣不是讀更多，而是先讀對月份、對來源，搶在大家追上之前看見現實。\u003C\u002Fp>","GitHub 的 last30days skill 是 AI 研究更好的模型，因為它優先抓取即時的人類訊號，而不是過時的網頁內容。","github.com","https:\u002F\u002Fgithub.com\u002Fmvanhorn\u002Flast30days-skill",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781752672625-a8f4.png","ai-agent","zh","e60c0f75-2fb3-4038-b0ab-4b0012007c73",[17,18,19,20,21],"GitHub","last30days skill","AI research","live signals","prediction markets",[23,24,25],"AI 研究應優先看最近 30 天的活訊號，而不是過時網頁。","多來源加權比單一搜尋排名更能反映當下共識與分歧。","工程師、PM、創辦人都應把即時訊號納入研究與決策流程。",0,"2026-06-18T03:17:22.68358+00:00","2026-06-18T03:17:22.677+00:00","e3b68196-9e64-4c18-a3b6-a73e73bfb367",{"tags":31,"relatedLang":34,"relatedPosts":38},[32],{"name":17,"slug":33},"github",{"id":15,"slug":35,"title":36,"language":37},"github-last30days-skill-ai-research-model-en","GitHub’s last30days skill is the right model for AI research","en",[39,45,51,57,63,69],{"id":40,"slug":41,"title":42,"cover_image":43,"image_url":43,"created_at":44,"category":13},"1459a665-b180-487b-b15b-65c046c6392c","tcs-anthropic-enterprise-ai-partnership-zh","TCS 和 Anthropic 企業 AI 合作成形","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781713081888-ob2e.png","2026-06-17T16:17:35.294583+00:00",{"id":46,"slug":47,"title":48,"cover_image":49,"image_url":49,"created_at":50,"category":13},"98a0d6a4-e485-46c0-b69a-8c25cef0a7d9","minimax-m3-real-edge-agentic-work-not-broad-excellence-zh","MiniMax M3 的真正優勢是 agentic 工作，不是全面稱王","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781611375147-mhkz.png","2026-06-16T12:02:21.598128+00:00",{"id":52,"slug":53,"title":54,"cover_image":55,"image_url":55,"created_at":56,"category":13},"45c6e678-8ac7-4881-8096-34703d7db136","yong-langgraph-zuo-chu-dai-li-shi-rag-xi-tong-zh","用 LangGraph 做出代理式 RAG 系統","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781485382723-k7xk.png","2026-06-15T01:02:29.343467+00:00",{"id":58,"slug":59,"title":60,"cover_image":61,"image_url":61,"created_at":62,"category":13},"5478bdd3-1241-4185-858f-345b365b24a8","manus-ai-proves-agents-are-ready-for-real-work-zh","Manus AI 證明代理人已能上工，但定價會決定它的命運","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781444874515-g9dg.png","2026-06-14T13:47:21.276926+00:00",{"id":64,"slug":65,"title":66,"cover_image":67,"image_url":67,"created_at":68,"category":13},"7ea0ef5b-d12c-4b18-b8fd-6ae3de67c296","coinbase-ai-agent-accounts-strict-limits-zh","Coinbase 讓 AI 代理代交易與代支付是對的，但前提是嚴格限權","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781409758550-mjql.png","2026-06-14T04:02:15.334232+00:00",{"id":70,"slug":71,"title":72,"cover_image":73,"image_url":73,"created_at":74,"category":13},"7315dc1e-d3c0-4888-8466-1328e8819be0","peft-llm-fine-tuning-without-full-retraining-zh","PEFT LoRA 微調 LLM 實作指南","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781403475967-xlpz.png","2026-06-14T02:17:26.268208+00:00",[76,81,86,91,96,101,106,111,116,121],{"id":77,"slug":78,"title":79,"created_at":80},"4ae1e197-1d3d-4233-8733-eafe9cb6438b","claude-now-uses-your-pc-to-finish-tasks-zh","Claude 開始幫你操作電腦","2026-03-26T07:20:48.457387+00:00",{"id":82,"slug":83,"title":84,"created_at":85},"5bede67f-e21c-413d-9ab8-54a3c3d26227","googles-2026-ai-agent-report-decoded-zh","Google 2026 AI Agent 報告解讀","2026-03-26T11:15:22.651956+00:00",{"id":87,"slug":88,"title":89,"created_at":90},"2987d097-563f-46c7-b76f-b558d8ef7c2b","kimi-k25-review-stronger-still-not-legend-zh","Kimi K2.5 評測：更強，但還不是神作","2026-03-27T07:15:55.277513+00:00",{"id":92,"slug":93,"title":94,"created_at":95},"95c9053b-e3f4-4cb5-aace-5c54f4c9e044","claude-code-controls-mac-desktop-zh","Claude Code 也能操控 Mac 了","2026-03-28T03:01:58.58121+00:00",{"id":97,"slug":98,"title":99,"created_at":100},"dc58e153-e3a8-4c06-9b96-1aa64eabbf5f","cloudflare-100x-faster-ai-agent-sandbox-zh","Cloudflare 的 AI 沙箱跑超快","2026-03-28T03:09:44.142236+00:00",{"id":102,"slug":103,"title":104,"created_at":105},"1c8afc56-253f-47a2-979f-1065ff072f2a","openai-backs-isara-agent-swarm-bet-zh","OpenAI 挺 Isara 的 agent swarm …","2026-03-28T03:15:27.513155+00:00",{"id":107,"slug":108,"title":109,"created_at":110},"7379b422-576e-45df-ad5a-d57a0d9dd467","openai-plan-automated-ai-researcher-zh","OpenAI 想做自動化 AI 研究員","2026-03-28T03:17:42.090548+00:00",{"id":112,"slug":113,"title":114,"created_at":115},"48c9889e-86df-450b-a356-e4a4b7c83c5b","harness-engineering-ai-agent-reliability-2026-zh","駕馭工程：從「馬具」到「作業系統」，AI Agent 可靠性的終極密碼","2026-03-31T06:42:53.556721+00:00",{"id":117,"slug":118,"title":119,"created_at":120},"96d8e8c8-1edd-475d-9145-b1e7a1b02b65","mcp-explained-from-prompts-to-production-zh","MCP 怎麼把提示詞變工作流","2026-04-01T09:24:39.321274+00:00",{"id":122,"slug":123,"title":124,"created_at":125},"f2ca7720-b471-4ce5-9336-2a9ac2a876fd","amazon-bedrock-agents-multi-agent-workflows-zh","Amazon Bedrock Agents 進入多代理工作流","2026-04-01T09:30:29.945429+00:00"]