[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-alphafold-breakup-turns-science-into-gemini-work-zh":3,"article-related-alphafold-breakup-turns-science-into-gemini-work-zh":29,"series-industry-32e46e55-e017-43f1-8ada-bd33c57cbf32":74},{"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":11},"32e46e55-e017-43f1-8ada-bd33c57cbf32","alphafold-breakup-turns-science-into-gemini-work-zh","AlphaFold 拆組成 Gemini 工程","\u003Cp data-speakable=\"summary\">以前 AlphaFold 是獨立科研隊，現在它的能力被拆進 \u003Ca href=\"\u002Ftag\u002Fgemini\">Gemini\u003C\u002Fa> 和藥物發現工作裡。\u003C\u002Fp>\u003Cp>我盯 \u003Ca href=\"\u002Ftag\u002Fgoogle-deepmind\">Google DeepMind\u003C\u002Fa> 很久了，這次消息一出來，我第一個反應不是震驚，是那種很熟的悶。AlphaFold 這種案子，難得不是拿來做展示櫃的 AI。它真的解了問題，真的有人在用，真的把研究工作往前推了一大步。也因為這樣，看到它的團隊被拆掉，我會覺得怪：不是因為團隊不能調動，而是因為這支隊伍曾經像是 Google 還能安靜做難題的證據。現在看起來，證據被搬去別的地方了。\u003C\u002Fp>\u003Cp>我用這篇 \u003Ca href=\"https:\u002F\u002Fwww.engadget.com\u002F2225849\u002Fgoogle-shuts-down-alphafold\u002F\">Engadget 報導\u003C\u002Fa>當切口，它引用了 \u003Ca href=\"https:\u002F\u002Fwww.ft.com\u002F\">Financial Times\u003C\u002Fa> 和 DeepMind 的說法。重點不是「關掉」兩個字有多戲劇化，而是人被怎麼重新分配：一部分去 Gemini，一部分去 \u003Ca href=\"https:\u002F\u002Fwww.isomorphiclabs.com\u002F\">Isomorphic Labs\u003C\u002Fa>，還有少數人已經離開。這種變動很值得拆，因為它講的是公司現在到底把什麼放在中心。\u003C\u002Fp>\u003Ch2>AlphaFold 沒死，是被吸收掉了\u003C\u002Fh2>\u003Cblockquote>“Google has reassigned some of the team's key members and the original authors of its papers to both Gemini-related projects and scientific research endeavors, while a few others have already left the company.”\u003C\u002Fblockquote>\u003Cp>翻譯一下就是：AlphaFold 這個獨立隊形沒了，但能力沒有消失。\u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa> 把人、經驗、還有那堆很難寫進簡報的技術記憶，塞進別的桶裡。這跟把技術砍掉是兩回事。大公司很常這樣處理成功的研究線：不直接說收掉，只說調整、整併、資源重配，聽起來比較不痛。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785524596007-62sw.png\" alt=\"AlphaFold 拆組成 Gemini 工程\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>我看過太多內部平台隊被這樣拆過。先是\u003Ca href=\"\u002Fnews\u002Fvibe-island-changelog-right-product-bets-zh\">證明\u003C\u002Fa>自己能打，接著高層說下一步不是保留這個隊，而是把這批人拆去更大的賭注裡。成果保住了，身份沒了。你如果待過那種「原本很像核心，後來\u003Ca href=\"\u002Fnews\u002Fopenai-newsroom-announcements-digest-zh\">變成\u003C\u002Fa>供應人力」的團隊，應該知道那個味道。\u003C\u002Fp>\u003Cp>AlphaFold 之所以特別，是因為它不是一個可有可無的 demo。DeepMind 在 2018 年啟動，2020 年解蛋白質折疊問題，2021 年在 \u003Ca href=\"https:\u002F\u002Fwww.nature.com\u002F\">Nature\u003C\u002Fa> 發方法和全人類蛋白質組預測，還把 \u003Ca href=\"https:\u002F\u002Falphafold.ebi.ac.uk\u002F\">AlphaFold Protein Structure Database\u003C\u002Fa> 開給研究者用，超過 2 億筆預測可查。這不是裝飾品，這是生物學基礎設施。\u003C\u002Fp>\u003Cp>所以我不會把這件事讀成失敗。我會讀成：公司覺得下一個中心故事，不在這裡了。從組織語言來看，AlphaFold 已經從主角變成供應主角的人。\u003C\u002Fp>\u003Cp>實操上，如果你在公司裡帶隊，看到自己的團隊開始被「借人」到別的核心專案，就要警覺了。那通常代表公司認為成果本身沒那麼重要，重要的是做成果的人。這時候最該做的不是抱怨，是把方法、決策脈絡、評估標準先寫死，免得 org chart 一轉，大家都假裝自己一直都知道怎麼做。\u003C\u002Fp>\u003Ch2>真正的主線是 Gemini 在吸資源\u003C\u002Fh2>\u003Cblockquote>“This represents Google's decision to put more and more of its resources into developing Gemini.”\u003C\u002Fblockquote>\u003Cp>這句話其實就把整件事講完了。AlphaFold 不再是公司級主線，Gemini 才是。你一旦接受這個前提，後面所有變動都合理了：人被調走、研究被拆散、敘事被收斂到旗艦模型家族。\u003C\u002Fp>\u003Cp>也就是說，Google 現在選的是集中火力，不是四處開花。少一點英雄式支線，多一點資源壓在它認為能撐住 AI 敘事的模型線上。站在董事會角度，這很正常。站在做過專門研究的人角度，這就有點硬。因為很多有價值的工作，本來就不是聊天機器人那種容易講成故事的東西。\u003C\u002Fp>\u003Cp>DeepMind 研究副總裁 Pushmeet Kohli 說過，過去九年的策略是聚焦「grand challenges」，每個專案都有一個具體目標，然後他也補了一句「The strategy has evolved.」這種話我看太多了。意思很清楚：有 pivot，但希望你把它聽成自然演化。老實說，我可以接受策略變，但我不接受公司假裝沒有代價。\u003C\u002Fp>\u003Cp>代價就是焦點。Gemini 拿到更多焦點，AlphaFold 被併進更大的機器。如果你在做 AI 產品，這很值得盯。很多團隊最後都會走到同一種狀態：某個模型家族變成戰略中心，其他東西就慢慢變成支援單位，即使它原本是皇冠上的那顆。\u003C\u002Fp>\u003Cul>\u003Cli>Gemini 變成總敘事。\u003C\u002Fli>\u003Cli>專門研究被導去更產品化的工作。\u003C\u002Fli>\u003Cli>科學成果變成平台能力的證明。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>實操上，如果你手上有多條 AI 線，請直接列出：哪條拿最多 executive airtime、哪條拿最多招人預算、哪條最常被拿來當「我們應該圍繞它來建」的理由。那才是公司真正的策略，不是簡報封面。\u003C\u002Fp>\u003Ch2>AlphaFold 從來不只是模型\u003C\u002Fh2>\u003Cblockquote>“AlphaFold is an AI program that can accurately predict three-dimensional structures of proteins from their amino acid sequences in minutes instead of years.”\u003C\u002Fblockquote>\u003Cp>這句話就是 AlphaFold 之所以重要的原因。它不是一個貼上科學標籤的\u003Ca href=\"\u002Fnews\u002Fbuild-small-language-model-deepmind-zh\">語言模型\u003C\u002Fa>，而是一個真的能改變研究流程的領域工具。蛋白質結構預測從幾年縮到幾分鐘，直接影響藥物發現、疫苗設計、疾病機制研究，像阿茲海默症、帕金森氏症這些題目都會被它碰到。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785524589596-j4x3.png\" alt=\"AlphaFold 拆組成 Gemini 工程\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>我覺得很多 AI 報導都偷懶，什麼都叫 AI，然後用同一套話術講到底。其實完全不是。能幫生物學家更快做事的模型，跟幫你寫信的模型，不是同一種東西。AlphaFold 的名聲是自己做出來的，因為它真的解了科學界卡很久的難題。\u003C\u002Fp>\u003Cp>DeepMind 把 \u003Ca href=\"https:\u002F\u002Falphafold.ebi.ac.uk\u002F\">database\u003C\u002Fa> 開出去這件事也很重要。2 億筆預測免費給人用，等於把門檻往下打。沒有大算力、沒有完整 ML 團隊的實驗室，也能直接拿來做事。這種外溢效果，比模型本身還值得記一筆。\u003C\u002Fp>\u003Cp>2024 年 Demis Hassabis 和 John Jumper 拿到諾貝爾化學獎，這件事也把 AlphaFold 的地位釘死了。軟體團隊能拿到這種級別的認可，幾乎就是永久蓋章。可就算這樣，組織還是照拆不誤。獎項不保組織形狀，只會讓重組看起來更像是刻意的。\u003C\u002Fp>\u003Cp>實操上，如果你在做垂直領域 AI，不要只盯模型分數。你要看的是：這個工具有沒有改變誰能參與、速度有沒有變快、下游工作有沒有被打開。這才是應用研究真正的護城河。\u003C\u002Fp>\u003Ch2>人一搬，知識也跟著搬\u003C\u002Fh2>\u003Cblockquote>“DeepMind has confirmed to the Times that it also moved staff members internally to Gemini-focused projects.”\u003C\u002Fblockquote>\u003Cp>這段是最務實的一刀。公司很愛講 knowledge transfer，聽起來像有條有理，其實常常很虛。知識不住在文件裡，知識住在人腦裡，住在某個人記得的奇怪例外條件，住在 2 點半系統炸掉時誰知道先看哪裡。\u003C\u002Fp>\u003Cp>我以前在平台團隊就踩過這個坑。當時我們以為把資深的人調去新專案、文件留給其他人，就算交接完成。結果文件都對，事情卻不對。文件不會記得哪些假設早就被打臉，哪些捷徑只是暫時的，哪些邊角案例其實已經變成部落傳說。人會。\u003C\u002Fp>\u003Cp>所以 AlphaFold 這次的人事搬移，不只是 headcount 調整。它是把深技術記憶往更高優先級的賭注裡塞。這種做法可以成功，也可以把原本很特別的文化磨平。尤其當研究團隊被吸進更大的模型工程，原本的專門問題很容易開始跟通用目標搶資源。\u003C\u002Fp>\u003Cul>\u003Cli>把原始研究目標寫成一頁文件，別只留在口頭上。\u003C\u002Fli>\u003Cli>把評估指標鎖回領域問題，不要只看父模型指標。\u003C\u002Fli>\u003Cli>列出誰在搬走、誰留下、誰負責最後的知識交接。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>實操上，每次團隊重組，我都建議先問一句：這個 group 現在到底要解什麼問題？如果答案很虛，通常代表這團隊已經開始漂了。\u003C\u002Fp>\u003Ch2>Isomorphic Labs 是最誠實的旁白\u003C\u002Fh2>\u003Cblockquote>“Other former staff members were reassigned to Alphabet's drug-discovery company Isomorphic Labs.”\u003C\u002Fblockquote>\u003Cp>這個細節很容易滑過去，但我覺得它超關鍵。Isomorphic Labs 本來就是朝藥物發現走的公司，所以 Google 並不是把 AlphaFold 的人亂丟去產品線，而是想把科學能力放進一個比較好講商業故事的容器裡。\u003C\u002Fp>\u003Cp>這就是折衷。Gemini 當總敘事，Isomorphic Labs 當生物應用敘事，AlphaFold 的獨立團隊則被拆散，但人才還留在能派上用場的地方。站在管理者角度，這很整齊。站在研究文化角度，這有點可惜。\u003C\u002Fp>\u003Cp>我不覺得這代表 Google 不做科學了。我覺得它代表：科學現在得經過公司主線的審核。能接 Gemini，最好。能接藥物發現 spin-off，也行。接不上主敘事的工作，通常就比較難拿到空間。\u003C\u002Fp>\u003Cp>這件事對任何在大公司裡帶技術團隊的人都很有參考價值。公司不是不愛你的工作，它只是更愛一個更容易講給外界聽的故事。很煩，但現實常常就是這樣。\u003C\u002Fp>\u003Cp>實操上，如果你想保住一支專門團隊，請準備兩版說法：一版給公司敘事，一版給技術使命。你只講前者，最後團隊大概率會被改造成更方便的樣子。\u003C\u002Fp>\u003Ch2>可抄的模板\u003C\u002Fh2>\u003Cpre>\u003Ccode># Team Reallocation Note: From Research Unit to Strategic Platform Work\n\n## What changed\nWe are moving members of [original team] into [new strategic area] and [adjacent specialized group]. The original standalone team will no longer operate as a separate unit.\n\n## Why we are doing this\nThe company is concentrating resources on [primary strategic bet]. The expertise built in [original team] is still valuable, but it now serves a broader set of priorities.\n\n## What stays the same\n- The core technical knowledge remains in the company.\n- Existing research artifacts, docs, and benchmarks should be preserved.\n- Ongoing domain-specific work continues under [new owner\u002Fteam].\n\n## What changes for the team\n- Members will report into [new org\u002Fteam].\n- Success metrics will be updated to match the new mission.\n- Any domain-specific roadmap items must be revalidated under the new ownership.\n\n## Questions we need answered\n1. Which problem is this group now responsible for?\n2. Which metrics still matter from the old team?\n3. What gets deprecated, and who decides?\n4. What knowledge needs explicit transfer before the reorg completes?\n5. What artifacts should be archived so the original work doesn’t get lost?\n\n## Copy-ready transition checklist\n- [ ] Name the new strategic owner.\n- [ ] List the domain-specific outcomes that must survive.\n- [ ] Identify the people whose knowledge is most critical.\n- [ ] Preserve evaluation data and benchmark history.\n- [ ] Publish a short FAQ for affected teams.\n- [ ] Set a review date to confirm the new structure is working.\n\n## One-sentence summary\nWe are not deleting the work; we are moving the expertise into the places where the company has decided it matters most.\u003C\u002Fcode>\u003C\u002Fpre>\u003Cp>如果我自己要把這件事變成內部 memo，我會寫得很白：改了什麼、人去哪裡、現在誰負責什麼、哪些東西要先留下。不要寫什麼「新篇章」或「令人興奮的轉型」。那種字眼通常只是在幫組織拖延說實話。\u003C\u002Fp>\u003Cp>而且最常被漏掉的，就是那個「先留下什麼」的清單。少了這段，六個月後大家又會裝作很驚訝，說怎麼專門知識不見了。\u003C\u002Fp>\u003Cp>來源：\u003Ca href=\"https:\u002F\u002Fwww.engadget.com\u002F2225849\u002Fgoogle-shuts-down-alphafold\u002F\">Engadget\u003C\u002Fa>，其內容引用了 \u003Ca href=\"https:\u002F\u002Fwww.ft.com\u002F\">Financial Times\u003C\u002Fa> 與 DeepMind 的說法。這篇拆解是我自己整理的，但事實基礎、引述與人物動向都來自上述來源與官方連結。","我拆 Google DeepMind 的 AlphaFold 重組，順手給你一份可直接套用的團隊對齊模板。","www.engadget.com","https:\u002F\u002Fwww.engadget.com\u002F2225849\u002Fgoogle-shuts-down-alphafold\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785524596007-62sw.png","industry","zh","4d053fa6-6a9a-46a2-a1ba-4e5ab183c583",[17,18,19,20,21],"AlphaFold","Gemini","DeepMind","org reorg","research-to-product",[23,24,25],"AlphaFold 的變動重點是人和能力被重新分配，不是技術消失。","Google 目前把資源更集中到 Gemini，科學研究被納入更大的戰略敘事。","團隊重組時，最該保住的是問題定義、評估指標和知識交接。",0,"2026-07-31T19:02:46.126821+00:00","2026-07-31T19:02:46.116+00:00",{"tags":30,"relatedLang":33,"relatedPosts":37},[31],{"name":18,"slug":32},"gemini",{"id":15,"slug":34,"title":35,"language":36},"alphafold-breakup-turns-science-into-gemini-work-en","AlphaFold’s breakup turns science into Gemini work","en",[38,44,50,56,62,68],{"id":39,"slug":40,"title":41,"cover_image":42,"image_url":42,"created_at":43,"category":13},"0c487fac-7e5e-4ae0-ad18-59ca7c37cd01","rust-to-zig-rewrite-progress-update-zh","Rust 轉 Zig：重寫已過最難關","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785501166060-u7ke.png","2026-07-31T12:32:19.644785+00:00",{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"522d7bc1-3349-41a8-ae2c-6d7ef7e3b843","nvidia-open-ai-security-alliance-partners-zh","Nvidia 牽頭 AI 安全聯盟","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785486780858-e4b3.png","2026-07-31T08:32:30.687219+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"f71bb1e1-1ba8-40f5-8769-63e218a14caa","kimi-k3-pushes-open-weight-ai-default-zh","Kimi K3 把開放權重變預設","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785422004101-o2l3.png","2026-07-30T14:32:56.994914+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"2f84c385-b870-4ef8-90ee-f8b1c545392c","anthropic-open-model-fight-lonely-ai-stance-zh","Anthropic 對開放模型的孤立姿態","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785414769795-ommd.png","2026-07-30T12:32:24.061158+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"06ecd48d-e824-4918-9b91-1aa6cb7e9ddf","immich-docker-compose-setup-common-errors-zh","Immich Docker Compose 5 個常見錯誤修正","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785403975315-zugi.png","2026-07-30T09:32:28.534813+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"54073d4e-0838-4c5f-bb9b-8628a746c800","anthropic-buy-books-scan-destroy-training-zh","Anthropic買書掃描再銷毀，想守住訓練合法性","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1785396792157-cnvv.png","2026-07-30T07:32:42.993389+00:00",[75,80,85,90,95,100,105,110,115,120],{"id":76,"slug":77,"title":78,"created_at":79},"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":81,"slug":82,"title":83,"created_at":84},"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":86,"slug":87,"title":88,"created_at":89},"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":91,"slug":92,"title":93,"created_at":94},"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":96,"slug":97,"title":98,"created_at":99},"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":101,"slug":102,"title":103,"created_at":104},"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":106,"slug":107,"title":108,"created_at":109},"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":111,"slug":112,"title":113,"created_at":114},"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":116,"slug":117,"title":118,"created_at":119},"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":121,"slug":122,"title":123,"created_at":124},"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"]