(圖說：西雅圖富士山 Mt Rainier，國家公園裡頭 Reflection Lake。面對山，面對時間，我們人類在 post-AGI 可以幹嘛，至少好好做人是底線吧，你們這些魔族。我不確定我能不能住在鄉下，但我喜歡旁邊沒有雜騖。然後我發現「心無旁騖」也是馬年吉祥話？！圖片來源：Ernest。）
✳️ 這個人從十一月開始，沒有改過一行程式碼
最近聽了 Lenny’s Podcast 訪問 Boris Cherny，他是 Anthropic Claude Code 的負責人，曾是 Instagram 產出最高的工程師之一。他從去年十一月開始 100% 程式碼由 Claude Code 撰寫，沒有手動改過一行，每天出 10 到 30 個 PR，錄影同時有 5 個 agent 在跑。Anthropic 工程團隊成長 4 倍，每位工程師產出提升 200%。他的前一份工作在 Meta 負責全公司 code quality，當年幾百個工程師投入一整年通常只能提升幾個百分點的生產力，如今是幾百個百分點，完全不同量級。
Boris 的時間軸值得參考：二月 Claude Code 發佈時寫他 20% 的程式碼（昨天才剛看到 Claude Code 一歲生日的照片），五月 30%，十一月 100%。GitHub 上 4% 的 commits 來自 Claude Code，年底目標 20%。
我們團隊去年的時間軸也是類似，只是摸索了從 Cline 到 Cursor 到 Claude Code 再到 Kiro。初期用 Cline 還停留在「把 AI 當聰明的 autocomplete」，到 Claude Code 才意識到自己不是在用工具，是在跟一個（一群）能動腦的 agent 一起工作（各種燒腦與反直覺）。
✳️ 印刷術問世的時候，抄寫員是興奮的
Boris 用印刷術做歷史類比。印刷術前歐洲識字率不到 1%，發明後 50 年內產出的印刷品比過去一千年的總和還多，成本降了 100 倍。但識字率到 70% 花了 200 年。寫程式這件事才 60 年的歷史，從打孔卡到軟體，Boris 的祖父是蘇聯第一批程式設計師，用打孔卡寫程式，也沒趕上軟體轉型。這個領域從來就不是靜態的。
最反直覺的細節：15 世紀有位抄寫員被問到對印刷術的感受，竟然很興奮。「我最不喜歡的就是抄來抄去，我喜歡的是畫插圖和裝訂書」。Boris 說自己完全一樣，寫程式一直以來就是瑣碎的部分，弄 git、搞各種工具，那不是好玩的事。好玩的是想做什麼、跟使用者對話、思考大系統怎麼設計。他說從來沒有像現在這樣享受 coding。
Boris 提到 80% 任務用 Plan Mode 開始，本質上就是「請你先不要寫 code」，計劃確認後直接 auto accept，用 Opus 4.6 幾乎每次一次到位。這跟我們在去年 Q1 時期的做法相似，我們當年（唉，才一年前）在 prompt 裡特別寫「請你不要先寫 code，請你先跟我討論」、或是出動 multishots。原來，慢下來才能更快，不只是我的年度回顧主題，Boris 也在實踐同一件事。
✳️ 當所有人都會寫程式
Boris 預測 software engineer 頭銜會消失，被 builder 取代。年底每個人都是 PM，每個人都寫程式。Claude Code 團隊裡 PM 寫程式、設計師寫程式、工程管理者寫程式、data scientist 寫程式、連財務的人也寫程式。
他的建議是做通才，最強的工程師同時有產品思維和設計感，最強的 PM 也能直接改程式。他看到最有效率的人不只是 AI native，而是帶著好奇心跨越學科邊界，能把工程、產品、設計、商業邏輯融合在一起的人。（taskforce 型態的組織在這一輪可能有較佳反應能力，跟我們在 PAFERS 成立的 Product & Technology Integration 團隊概念近似，職能可以分工到多個部門，但增加的通訊 latency 時間要如何調校就是隱性關鍵。我從高中思考組織分工到現在，儘管實作過幾次，還是不敢把解法寫下來，但大略是離 disintegrate 或 collapse 不遠，需要社會安全網幫忙接，扯遠了回來回來…）
真正關鍵的競爭力在 Domain Insight、Systems Thinking、Verification 的交集處，不是程式碼寫多快，是能不能拆解需求、定義清楚、驗證正確，參考去年我在台灣產品年會分享的投影片「知道、說到、做到」。
話說 Boris 加入 Anthropic 前住在日本鄉下，跟鄰居交換自己做的味噌（可惡想要！！），一批白味噌要等三個月，紅味噌要兩三年，逼你用完全不同的時間尺度思考。問他 post-AGI 要幹嘛，他說大概去做味噌（我也是！我想進廚房玩食物）。
這集 1.5 小時少數資訊密度偏高的 podcast，推薦可聽。
✳️ 延伸閱讀
- Anthropic 內部如何使用 Claude Code
- 從 Vibe Coding 到 Agentic Coding：把話講清楚才是新戰場
- Coding Is Just Typing，然後呢？從 Explicit Programming 到 Implicit Programming
✳️ 知識圖譜
(更多關於知識圖譜…)
General Concept
Three Layers of AI Safety
✳️ 逐字稿
AI 對程式開發生產力的影響
- 100% of my code is written by Claude Code.
我 100% 的程式碼都是由 Claude Code 撰寫的。 - I have not edited a single line by hand since November.
自從十一月以來，我沒有手動編輯過任何一行程式碼。 - Every day I ship 10, 20, 30 p requests.
我每天送出 10、20、30 個 pull request。 - So at the moment I have like five agents running while we’re recording this.
所以在我們錄製的當下，我大約有五個 agent 在同時執行。 - » Yeah.
- Yeah.
- Do you miss writing code?
你會懷念寫程式嗎？ - » I have never enjoyed coding as much as I do today because I don’t have to deal with all the minutia.
» 我從來沒有像今天這樣享受寫程式，因為我不用再處理那些瑣碎的細節了。 - Productivity per engineer has increased 200%.
每位工程師的生產力提升了 200%。 - » There’s always this question, should I learn to code?
» 總是有人問這個問題：我應該學寫程式嗎？ - In a year or two, it’s not going to matter.
再過一兩年，這個問題就不重要了。
軟體開發角色的未來
- Coding is largely solved.
寫程式這件事基本上已經被解決了。 - I imagine a world where everyone is able to program.
我想像一個每個人都能寫程式的世界。 - Anyone can just build software anytime.
任何人都可以隨時建構軟體。 - What’s the next big shift to how software is written?
軟體開發方式的下一個重大轉變是什麼？ - » Claude is starting to come up with ideas.
» Claude 已經開始自己產生想法了。 - It’s looking through feedback.
它會檢視使用者回饋。 - It’s looking at bug reports.
它會查看錯誤回報。 - It’s looking at telemetry for bug fixes and things to ship a little more like a co-orker or something like that.
它會分析遙測資料來修復錯誤，並且更像一個同事一樣交付成果。 - » A lot of people listening to this are product managers and they’re probably sweating.
» 很多正在聽這集的人是產品經理，他們大概正在冒冷汗。 - I think by the end of the year, everyone’s going to be a product manager and everyone codes.
我認為到今年年底，每個人都會是產品經理，每個人都會寫程式。 - The title software engineer is going to start to go away.
軟體工程師這個職稱將會開始消失。 - It’s just going to be replaced by builder and it’s going to be painful for a lot of people.
它會被「builder」所取代，而這對很多人來說會很痛苦。
來賓與主題介紹
- Today my guest is Boris Churnney, head of Claude Code at Anthropic.
今天的來賓是 Boris Cherny，Anthropic 的 Claude Code 負責人。 - It is hard to describe the impact that Claude Code has had on the world.
很難用言語形容 Claude Code 對這個世界造成的影響。 - Around the time this episode comes out will be the one-year anniversary of Claude Code.
大約在這集節目上線的時候，就是 Claude Code 的一週年。 - And in that short time, it has completely transformed the job of a software engineer and it is now starting to transform the jobs of many other functions in tech which we talk about.
在這麼短的時間內，它已經徹底改變了軟體工程師的工作，而且現在正開始改變科技業中許多其他職能的工作，我們在節目中會談到這些。 - Claude Code itself is also a massive driver of Anthropic overall growth over the past year.
Claude Code 本身也是 Anthropic 過去一年整體成長的重要推動力。 - They just raised a round at over $350 billion.
他們剛以超過 3,500 億美元的估值完成了一輪募資。 - And as Boris mentions, the growth of Claude Code itself is still accelerating.
正如 Boris 提到的，Claude Code 本身的成長仍在加速。 - Just in the past month, their daily active users has doubled.
光是在過去一個月，他們的每日活躍使用者就翻了一倍。 - Boris is also just a really interesting, thoughtful, deepinking human.
Boris 也是一個非常有趣、深思熟慮的人。 - And during this conversation, we discover we were born in the same city in Ukraine.
在這次對話中，我們發現我們出生在烏克蘭的同一座城市。 - That is so funny.
這真的太巧了。 - I had no idea.
我完全不知道。 - A huge thank you to Ben Man, Jenny Wen, and Mike Griger for suggesting topics for this conversation.
非常感謝 Ben Man、Jenny Wen 和 Mike Griger 為這次對話提供了建議的主題。 - Don’t forget to check out lennisprodpass.com for an incredible set of deals available exclusively to Lenny’s newsletter subscribers.
別忘了查看 lennisprodpass.com，那裡有專屬於 Lenny 電子報訂閱者的超棒優惠。 - Let’s get into it after a short word from our wonderful sponsors.
在聽完贊助商的簡短訊息後，我們就開始吧。
贊助商：DX 開發者智慧平台
- Today’s episode is brought to you by DX, the developer intelligence platform designed by leading researchers.
本集節目由 DX 贊助，DX 是由頂尖研究人員設計的開發者智慧平台。 - To thrive in the AI era, organizations need to adapt quickly.
要在 AI 時代蓬勃發展，組織需要快速適應。 - But many organization leaders struggle to answer pressing questions like which tools are working?
但許多組織的領導者難以回答這些迫切的問題，例如哪些工具有效？ - How are they being used?
它們是如何被使用的？ - What’s actually driving value?
什麼才是真正驅動價值的？ - DX provides the data and insights that leaders need to navigate this shift.
DX 提供領導者所需的資料和洞見，以應對這場轉變。 - With DX, companies like Dropbox, Booking.com, Adion, and Intercom get a deep understanding of how AI is providing value to their developers and what impact AI is having on engineering productivity.
透過 DX，像 Dropbox、Booking.com、Adion 和 Intercom 這樣的公司，能夠深入了解 AI 如何為開發者創造價值，以及 AI 對工程生產力產生了什麼影響。 - To learn more, visit DX’s website at getdx.com/lenny.
想了解更多，請造訪 DX 的網站 getdx.com/lenny。 - That’s getdx.com/lenny.
網址是 getdx.com/lenny。
贊助商：Sentry 應用程式監控
- Applications break in all kinds of ways.
應用程式會以各種方式出問題。 - Crashes, slowdowns, regressions, and the stuff that you only see once real users show up.
當機、變慢、功能退化，還有那些只有真實使用者上線後才會出現的問題。 - Sentry catches it all.
Sentry 能捕捉到所有這些問題。 - See what happened where, and why, down to the commit that introduced the error, the developer who shipped it, and the exact line of code all in one connected view.
你可以看到發生了什麼、在哪裡發生、為什麼發生，精確到引入錯誤的那個 commit、部署它的開發者，以及確切的程式碼行數，全部在一個關聯的視圖中呈現。 - I’ve definitely tried the five tabs and Slack thread approach to debugging.
我確實試過同時開五個分頁加上 Slack 討論串的除錯方式。 - This is better.
Sentry 更好用。 - Sentry shows you how the request moved, what ran, what slowed down, and what users saw.
Sentry 會展示請求如何流動、執行了什麼、什麼變慢了，以及使用者看到了什麼。 - Seir, Sentry’s AI debugging agent, takes it from there.
Seer 是 Sentry 的 AI 除錯 agent，它會從這裡接手。 - It uses all of that Sentry context to tell you the root cause, suggest a fix, and even opens a PR for you.
它利用所有 Sentry 的上下文資訊來告訴你根本原因、建議修復方式，甚至幫你開一個 PR。 - It also reviews your PR and flags any breaking changes with fixes ready to go.
它還會審查你的 PR，標記任何破壞性變更，並準備好修復方案。 - Try Sentry and SER for free at centry.io/lenny and use code Lenny for $100 in Sentry credits.
在 sentry.io/lenny 免費試用 Sentry 和 Seer，使用代碼 Lenny 可獲得 100 美元的 Sentry 點數。 - That’s s n t r y.io/lenny.
網址是 s-e-n-t-r-y.io/lenny。
Boris 回歸 Anthropic 與使命聚焦
- Boris, thank you so much for being here and welcome to the podcast.
Boris，非常感謝你來到這裡，歡迎來到這個 podcast。 - » Yeah, thanks for having me on.
» 謝謝你邀請我。 - » I want to start with a a spicy question.
» 我想先從一個辛辣的問題開始。 - About 6 months ago, I don’t know if people even remember this, you actually left Anthropic.
大約 6 個月前，我不知道大家是否還記得這件事，你其實離開了 Anthropic。 - You joined Curser and then two weeks later, you went back to Anthropic.
你加入了 Cursor，然後兩週後，你又回到了 Anthropic。 - What happened there?
到底發生了什麼事？ - I don’t think I’ve ever heard the actual story.
我想我從來沒有聽過真正的故事。 - It’s the fastest job change that I’ve ever had.
這是我有過最快的一次工作轉換。 - Um, I joined Cursor because I’m a big fan of the product and honestly I met the team and I was just really impressed.
我加入 Cursor 是因為我是這個產品的忠實粉絲，而且老實說我見到了團隊，我真的非常印象深刻。 - Uh, they’re an awesome team.
他們是一個很棒的團隊。 - Uh, I still I still think they’re awesome and they’re just building really cool stuff and kind of they they saw where AI coding was going I think before a lot of people did.
我現在仍然覺得他們很棒，他們正在做很酷的東西，而且我覺得他們比很多人更早看到 AI 程式設計的未來走向。 - So the idea of building good product was just very exciting for me.
所以打造好產品的想法對我來說非常令人興奮。 - I think as soon as I got there, what I started to realize is what I really missed about Ant was the mission.
我覺得我一到那裡，就開始意識到我真正懷念 Anthropic 的是使命。 - And that’s actually what originally drove me to Ant also cuz uh but before I joined Anthropic, I was, you know, I was working in big tech and then I was at some point I wanted to work at a at a lab to just help shape the future of this crazy thing that that we’re building in some way.
這其實也是最初驅使我加入 Anthropic 的原因，因為在加入 Anthropic 之前，我在大型科技公司工作，然後到了某個時間點，我想要在一個實驗室工作，以某種方式幫助塑造我們正在打造的這個瘋狂事物的未來。 - And the thing that drew me to Anthropic was the mission.
而吸引我到 Anthropic 的就是使命。 - And it was, you know, it’s all about safety.
而這一切都是關於安全性。 - And when you talk to people at Anthropic, just like find someone in the hallway, if you ask them why they’re here, the answer is always going to be safety.
當你和 Anthropic 的人聊天，隨便在走廊上找個人，如果你問他們為什麼在這裡，答案永遠都是安全性。 - Um, and so this kind of like missiondrivenness just really really resonated with me.
所以這種使命驅動的精神真的非常非常地引起我的共鳴。 - And I just know personally it’s something I need in order to be happy.
而且我個人很清楚，這是我需要的東西，才能感到快樂。 - Um, and I that’s just a thing that I really missed.
而這正是我真正懷念的。 - And I found that, you know, whatever the work might be, no matter how exciting, even if it’s building a really cool product, it’s just not really a substitute for that.
我發現不管工作內容是什麼，不管多麼令人興奮，即使是在打造一個很酷的產品，也無法真正替代那種感覺。 - Um, so for me it was actually u it was pretty obvious that that I was missing that pretty quick.
所以對我來說，其實很快就很明顯地感受到我缺少了這個。
回顧 Claude Code 的影響
- » Okay.
- So let me follow the thread of just coming back to Anthropic and the work you’ve done there.
讓我接著聊你回到 Anthropic 以及你在那裡所做的工作。 - This podcast is going to come out around the year anniversary of launching Claude Code.
這集 podcast 上線的時候大約是 Claude Code 發布的一週年。 - So I’m going to spend a little time just reflecting on the impact that you’ve had.
所以我想花一點時間回顧一下你所帶來的影響。 - There’s um this report that recently came out that I’m sure you saw by semi analysis that showed that 4% of all GitHub commits are authored by Claude Code now.
最近有一份由 Semi Analysis 發布的報告，我相信你已經看過了，顯示現在所有 GitHub commit 中有 4% 是由 Claude Code 撰寫的。 - and they predicted it’ll be a fifth of all code commits on GitHub by the end of the year.
而且他們預測到年底會達到 GitHub 上所有 code commit 的五分之一。 - The way they put it is while we blinked, AI consumed all software development.
他們的說法是：在我們眨眼的瞬間，AI 就吞噬了所有的軟體開發。 - The day that we’re recording this, Spotify just put out this uh headline that their best developers haven’t written a line of code since December thanks to AI.
就在我們錄製這集的這天，Spotify 剛發布了一則標題，說他們最優秀的開發者從去年 12 月起就沒有再寫過一行程式碼了，都是靠 AI。 - More and more of the most advanced senior engineers, including you, are sharing the fact that you don’t write code anymore, that it’s all AI generated.
越來越多最頂尖的資深工程師，包括你在內，都在分享一個事實：你們不再寫程式了，全部都是 AI 產生的。 - and many aren’t even looking at code anymore is how far we’ve gotten in large part thanks to this little project that you started and that your team has scaled over the past year.
而且很多人甚至不再看程式碼了，這就是我們已經走了多遠，這在很大程度上要歸功於你啟動的這個小專案，以及你的團隊在過去一年裡將它擴展起來。 - I’m curious just to hear your reflections on on this past year and the impact that your work has had.
我很好奇想聽聽你對過去這一年以及你的工作所產生的影響的反思。
快速成長與核心概念
- These numbers are just totally crazy, right?
這些數字真的太瘋狂了，對吧？ - Like four 4% of all commits in the world is just way more than I imagined and like like you said, it still feels like the starting point.
全世界所有 commit 的 4%，這遠超過我的想像，而且就像你說的，這還感覺只是個起點。 - Um these are also just public commits.
這些還只是公開的 commit。 - So we actually think if you look at private repositories, it’s quite a bit higher than that.
所以我們實際上認為如果看私有儲存庫的話，比例會高出不少。 - And I I think the craziest thing for me isn’t even the number that we’re at right now, but the pace at which we’re growing because if you look at Claude Code’s growth rate kind of across any metric, it’s continuing to accelerate.
而我覺得對我來說最瘋狂的甚至不是我們目前的數字，而是我們成長的速度，因為如果你看 Claude Code 在任何指標上的成長率，它持續在加速。 - Um so it’s not just going up, it’s going up faster and faster.
所以它不只是在上升，而是越來越快地上升。 - When I first started Claude Code, it was just going to be a like it was just supposed to be a little hack.
當我最初開始做 Claude Code 的時候，它只是一個小小的 hack。 - Um you know we we broadly knew at Anthropic that we wanted to get a we wanted to ship some kind of coding product and you know for Anthropic for a long time we were building the models in this way that kind of fit our mental model of the way that we build safe hi where the model starts by being really good at coding then it gets really good at tool use then it gets really good at computer use roughly this is like the trajectory uh and you know we’ve been working on this for a long time and
我們在 Anthropic 大致上知道我們想要推出某種程式設計產品，而且在 Anthropic 很長一段時間裡，我們一直在用一種符合我們心智模型的方式來建構模型，這個模型先是在程式設計方面變得非常出色，然後在工具使用方面變得非常出色，接著在電腦使用方面變得非常出色，大致上這就是發展軌跡，而且我們已經在這方面投入了很長的時間。 - when you look at the team that I started on it was called the Anthropic labs team uh and actually Mike Kger and you know Ben man they just kicked this team off again uh for kind of round two the team built some pretty cool stuff so we built Claude Code we built MCP we built the desktop app so you can kind of see the seeds of this idea you know like it’s coding then it’s tool use then it’s computer use and the reason this matters for Anthropic is uh because of safety it’s kind of again just back
我最初加入的團隊叫做 Anthropic Labs 團隊，實際上 Mike Krieger 和 Ben Mann 他們剛剛又重新啟動了這個團隊，算是第二輪。這個團隊做了一些很酷的東西，我們做了 Claude Code、MCP、桌面應用程式，所以你可以看到這個想法的種子，先是程式設計，然後是工具使用，接著是電腦使用。而這對 Anthropic 之所以重要，是因為安全性，又回到了那個核心。 - to that AI is getting more and more powerful it’s getting more and more capable the thing that’s happened in the last year is that for at least For engineers, the AI doesn’t just write the code.
AI 越來越強大、越來越有能力。過去一年發生的事情是，至少對工程師來說，AI 不只是寫程式碼。 - It it’s not just a conversation partner, but it actually uses tools.
它不只是一個對話夥伴，而是它實際上會使用工具。 - It acts in the world.
它在世界中採取行動。 - Um, and I think now with Cowork, we’re starting to see the transition for non-technical folks also.
而我認為現在有了 Cowork，我們也開始看到非技術人員的轉變。 - Um, for a lot of people that use conversational AI, this might be the first time that they’re using the thing that actually acts.
對於很多使用對話式 AI 的人來說，這可能是他們第一次使用真正會採取行動的東西。 - It can actually use your Gmail, it can use your Slack, it can do all these things for you and it’s quite good at it.
它真的可以使用你的 Gmail、使用你的 Slack，幫你做所有這些事情，而且做得相當好。 - Um, and it’s only going to get better from here.
而且從現在開始只會越來越好。
原型開發與早期發展
- So I think for Anthropic for a long time there was this feeling that we wanted to build something but it wasn’t obvious what and so uh when I joined ant I spent one month kind of hacking and you know built a bunch of like weird prototypes most of them didn’t ship and you know weren’t even close to shipping it was just kind of understanding the boundaries of what the model can do then I spent a month doing post- training um so to understand kind of the research side of it and I think honestly
我覺得在 Anthropic 很長一段時間裡都有一種感覺，就是我們想要做些什麼，但不確定該做什麼。所以當我加入 Anthropic 的時候，我花了一個月的時間做各種嘗試，做了一堆奇怪的原型，大部分都沒有發布，甚至離發布還很遠，這只是在了解模型能力的邊界。然後我花了一個月做 post-training，以了解研究的那一面。老實說， - that’s just for me as an engineer I find that to do good work you really have to understand the layer under the layer at which you work.
作為一名工程師，我發現要做好工作，你真的必須理解你工作層級之下的那一層。 - And with traditional engineering work, you know, if you’re working on product, you want to understand the infrastructure, the runtime, the virtual machine, the language kind of whatever that is, the system that you’re building on.
在傳統的工程工作中，如果你在做產品，你會想要了解基礎設施、執行環境、虛擬機器、程式語言，不管那是什麼，就是你建構在其上的系統。 - But, uh, yeah, if you’re like if you’re working in AI, you just really have to understand the model to some degree to to do good work.
但如果你是在 AI 領域工作，你真的必須在某種程度上理解模型才能做好工作。 - So, I took a little detour to do that and then I came back and just started prototyping what eventually became Claude Code.
所以我繞了一小段路去做這件事，然後回來開始做原型，最終就成了 Claude Code。 - Uh, and the very first version of it, I I have like a there’s like a video recording of the summer because I recorded this demo and I posted it.
最初的版本，我有一段影片紀錄，因為我錄了這個 demo 然後發布出去了。 - It was called ClaudeCLI back then.
那時候它叫做 Claude CLI。 - And I just kind of showed off how it used a few tools and the shocking thing for me was that I gave it a batch tool and uh it just was able to use that to write code to tell me what music I’m listening to when I asked it like what music am I listening to?
我展示了它如何使用幾個工具，而讓我震驚的是，我給了它一個 bash 工具，當我問它「我現在在聽什麼音樂？」的時候，它居然能夠用那個工具寫程式來告訴我我正在聽什麼音樂。 - And this is the craziest thing, right?
這是最瘋狂的事情，對吧？ - cuz it’s like there’s no we I I didn’t instruct the model to say, you know, use, you know, this tool for this or kind of do whatever.
因為我沒有指示模型說要用這個工具做什麼，或者怎麼做。 - The model was given this tool and I figured out how to use it to answer this question that I had that I wasn’t even sure if it could answer.
模型被給了這個工具，然後它自己想出了如何使用它來回答我的問題，而我甚至不確定它能不能回答。 - What music am I listening to?
我在聽什麼音樂？ - And so I I I started prototyping this a little bit more.
所以我開始更進一步地做原型。 - Um I made a post about it and I announced it internally and it got two likes.
我寫了一篇關於它的文章，在內部宣布了，結果只得到了兩個讚。 - That’s the that was that was the extent of the reaction at the time because I think people internally you know like when you think of coding tools you think of like you think of IDE you think about kind of all these pretty sophisticated environments no one thought that this thing could be terminal based um that’s sort of a weird way to design it and that wasn’t really the intention but uh you know from the start I built it in a terminal because you know for the first couple months it was
那就是當時反應的全部了，因為我覺得內部的人，當你想到程式設計工具的時候，你會想到 IDE，你會想到各種相當精緻的環境，沒有人覺得這個東西可以是終端介面的，這算是一種奇怪的設計方式，而且那本來也不是這個意圖，但從一開始我就在終端裡建構它，因為在最初的幾個月裡，
終端介面的做法與初期反應
- just me so it was just the easiest way to build uh and for me this is actually a pretty important product lesson right is like you want to underresource things a little bit at the start.
只有我一個人，所以這就是最簡單的建構方式。而對我來說，這其實是一個很重要的產品教訓，就是你會想要在一開始的時候稍微少投入一些資源。 - Then we started thinking about what other form factors we should build and we actually decided to stick with the terminal for a while and the biggest reason was the model is improving so quickly.
然後我們開始思考應該做哪些其他的產品形式，而我們實際上決定先繼續使用終端一段時間，最大的原因是模型進步得太快了。 - We felt that there wasn’t really another form factor that could keep up with it.
我們覺得真的沒有其他產品形式能跟上它的速度。 - And honestly this was just me kind of like struggling with kind of like what should we build you know like for the last year Claude Code has just been all I think about.
老實說，這就是我一直在糾結的事情，就是我們到底應該做什麼。過去一年裡，Claude Code 就是我腦子裡唯一在想的事情。 - And so just like late at night, this is just something I was thinking about like, okay, the model is continuing to improve.
所以就像深夜裡，我一直在想的就是，好，模型持續在進步。 - What do we do?
我們該怎麼辦？ - How can we possibly keep up?
我們怎麼可能跟上？ - And the terminal was honestly just the only idea that I had.
而終端老實說就是我唯一想到的點子。 - And uh yeah, it ended up catching on after after I released it pretty quickly.
然後在我發布它之後，它很快就流行起來了。 - It became a hit at Anthropic and you know, the the daily active users just went vertical.
它在 Anthropic 內部大受歡迎，每日活躍使用者直接垂直上升。 - And really early on, actually before I launched it, Ben man uh nudged me to make a DAU chart and I was like, you know, it’s like kind of early maybe, you know, should we really do it right now?
其實很早期的時候，在我正式發布之前，Ben Mann 就催促我做一個 DAU 的圖表，而我當時覺得，可能還太早了吧，我們現在真的要做這個嗎？ - and he was like, “Yeah.”
然後他說：「要。」 - And so the the chart just went vertical pretty immediately.
然後那個圖表幾乎馬上就垂直上升了。 - Uh and then in February, we released it externally.
然後在二月，我們對外發布了它。
從使用者學習與產品演進
- Actually, something that people don’t really remember is Claude Code was not initially a hit when we released it.
其實有件事大家不太記得了，Claude Code 在我們發布的時候，一開始並不是大熱門。 - It it got a bunch of users.
它確實有了一批使用者。 - There was a lot of early adopters that got it immediately, but it actually took many months for everyone to really understand what this thing is.
有很多早期採用者馬上就理解了，但實際上花了好幾個月，大家才真正了解這個東西是什麼。 - Just again, it’s like it’s just so different.
再說一次，它就是太不一樣了。 - And when I think about it, kind of part of the reason Claude Code works is this idea of latent demand where we bring the tool to where people are and it makes existing workflows a little bit easier, but also because it’s it’s in a terminal.
當我想這件事的時候，Claude Code 之所以能成功，部分原因是這個潛在需求的概念，我們把工具帶到人們所在的地方，讓現有的工作流程稍微容易一點，但也因為它是在終端裡。 - It’s like a little surprising.
它有點令人驚訝。 - It’s a little alien in this way.
在這方面它有點陌生。 - So you have to you have to kind of be open-minded and you had to learn to use it.
所以你必須保持開放的心態，而且你必須學習如何使用它。 - And of course now you know Claude Code is available you know in the iOS and Android Claude app.
當然現在 Claude Code 已經可以在 iOS 和 Android 的 Claude 應用程式中使用了。 - It’s available in the desktop app.
在桌面應用程式中也可以使用。 - It’s available on the website.
在網站上也可以使用。 - It’s available as IDE extensions in Slack and GitHub.
也有 IDE 擴充套件，在 Slack 和 GitHub 上也可以使用。 - you know all these places where engineers are it’s a little more familiar but that wasn’t the starting point so yeah I mean at the beginning it was kind of a surprise that this thing was even useful and uh you know as the team grew as the product grew as it started to become more and more useful to people just people around the world from you know small startups to the biggest fang companies started using it and they started giving feedback and I think just reflecting back it’s been such a
所有這些工程師所在的地方，它就更熟悉了一些，但那不是起點。所以在一開始，這個東西居然有用本身就令人驚訝。隨著團隊成長、產品成長，當它開始對人們越來越有用的時候，全世界的人，從小型新創公司到最大的 FAANG 公司，都開始使用它，然後開始給回饋。回顧過去，這真的是一個非常 - humbling experience cuz we just we keep learning from our users and just the most exciting thing is like you know none of us really know what we’re doing.
令人謙卑的經驗，因為我們一直在從使用者身上學習，而最令人興奮的是，我們其實都不太知道自己在做什麼。 - Um and we’re just trying to figure out along with everyone else and the single best signal for that is just feedback from users.
我們只是在和所有人一起摸索，而最好的訊號就是來自使用者的回饋。 - Um so that’s just been the best I’ I’ve been surprised so many times.
所以那真的是最棒的，我已經被驚訝了太多次了。 - It’s incredible how fast something can change in today’s world.
在今天的世界裡，事情能改變得多快，真的令人難以置信。
軟體工程的劇烈變化
- You launched this a year ago and it wasn’t the first time people could use AI to code but uh in a year the entire profession of software engineering has dramatically changed like there’s all these predictions oh AI is going to be written 100% AI’s code is going to be written by AI everyone’s like no that’s crazy what are you talking about now it’s like » of course it’s happening exactly as they said it’s just so things move so fast and change so fast now » yeah it’s really fast back at uh
你一年前發布了這個，那不是人們第一次可以用 AI 寫程式，但在一年內，整個軟體工程的職業已經發生了巨大的變化。之前有所有這些預測說 AI 將會寫 100% 的程式碼，大家都說不可能，你在說什麼，現在卻變成了「當然正在發生，完全如他們所說的」，事情變化得太快了。 » 是的，真的很快。回到 - back at code with Claude back in May that was like our first uh you know like developer conference that we did as Anthropic.
回到五月的 Code with Claude，那是我們作為 Anthropic 舉辦的第一個開發者大會。 - Um I did a short talk and in the Q&A after the talk people were asking what are your predictions for the end of the year and my prediction back in May of 2025 was by the end of the year you might not need an ID to code anymore and we’re going to start to see engineers not doing this and I remember the room like audibly gasped.
我做了一個簡短的演講，在演講後的 Q&A 環節，人們問我對年底有什麼預測，我在 2025 年五月的預測是，到年底你可能不再需要 IDE 來寫程式了，而且我們會開始看到工程師不再這樣做了。我記得全場發出了驚嘆聲。 - It was such a crazy prediction but I think like at Anthropic like this is just the way the way we think about things is exponentials and this is like very deep in the DNA.
這是一個如此瘋狂的預測，但我覺得在 Anthropic，我們思考事情的方式就是指數成長，這深深刻在我們的 DNA 裡。 - Like if you look at our co-founders like three of them were the first three authors on the scaling laws paper.
如果你看我們的共同創辦人，其中三位是 scaling laws 論文的前三位作者。 - Um so we really just think in exponentials and if you kind of look at the exponential of the percent of code that was written by Claude at that point if you just trace the line it’s pretty obvious we’re going to cross 100% by the end of the year even if it just does not match intuition at all.
所以我們真的就是用指數的方式思考，如果你看當時由 Claude 寫的程式碼百分比的指數曲線，你只要延伸那條線，就很明顯到年底我們會突破 100%，即使這完全不符合直覺。 - Um, and so all I did was trace the line and yeah, in November that, you know, that happened for me personally and that’s been the case since and we’re starting to see that for a lot of different customers too.
所以我做的就只是延伸那條線，然後在十一月，對我個人來說那就發生了，從那之後一直如此，而且我們開始在很多不同的客戶身上也看到這種情況。 - I thought was really interesting what you just shared there about kind of the journey is this kind of idea of just playing around and seeing what happens.
我覺得你剛才分享的那段旅程非常有趣，就是這種隨意嘗試、看看會發生什麼的想法。
透過實驗驅動創新
- This came up comes up with open claw a lot just like Peter was playing around and just like a thing happen.
這在 OpenClaw 中也經常出現，就像 Peter 在玩耍然後就有事情發生了。 - And it feels like that’s a central kind of ingredient to a lot of the biggest innovations in AI is people just sitting around trying stuff to pushing the models further than most other people.
而且感覺這是許多 AI 最大創新的核心要素，就是人們坐在那裡嘗試各種東西，把模型推得比大多數人更遠。 - » I mean this the thing about innovation right like you can’t uh you can’t force it.
» 我的意思是，這就是創新的本質，你無法強迫它發生。 - There’s no road map for innovation.
創新沒有路線圖。 - Um you just have to give people space.
你只需要給人們空間。 - You have to give them maybe the word is like safety.
你必須給他們，也許這個詞是安全感。 - So it’s like psychological safety that it’s okay to fail.
就像是心理安全感，讓人覺得失敗是可以的。 - It’s okay if 80% of the ideas are bad.
即使 80% 的想法是糟糕的也沒關係。 - Um you also have to hold them accountable a bit.
你也必須讓他們有一點責任感。 - So if the idea is bad, you you know you cut your losses, move on to the next idea instead of investing more.
所以如果這個想法不好，你就止損，繼續下一個想法，而不是投入更多。 - Uh in the early days of Claude Code, I had no idea that this thing would be useful at all.
在 Claude Code 的早期，我完全不知道這個東西會有用。 - Cuz even in February when we released it, it was writing maybe I don’t know like 20% of my code, not more.
因為即使在二月我們發布它的時候，它大概只寫了我 20% 的程式碼，不會更多。 - And even in May, it was writing maybe 30%.
即使到了五月，它大概也只寫了 30%。 - I was still using you know curtzer for most of my code.
我的大部分程式碼還是用 Cursor 寫的。 - And it only crossed 100% in November.
它到十一月才突破 100%。 - So it took a while.
所以花了一段時間。 - But even from the earliest day, it just felt like I was on to something.
但即使從最早的時候，我就覺得我發現了什麼重要的東西。 - And I was just spending like every night, every weekend hockey on this.
我就是每天晚上、每個週末都投入在這上面。 - And luckily my, you know, my wife was very supportive.
幸運的是我太太非常支持我。 - Um, but it it just felt like it was on to something.
但就是感覺它有某種潛力。 - It wasn’t obvious what.
只是不明顯是什麼。 - And and sometimes, you know, you find a thread, you just have to pull on it.
有時候你找到一條線索，你就必須順著它拉下去。
使用 Claude Code 的個人編碼工作流程
- » So at this point, 100% of your code is written by Claude Code.
» 所以到目前為止，你 100% 的程式碼都是由 Claude Code 寫的。 - Is that is that kind of the current state of your coding?
這就是你目前寫程式的狀態嗎？ - » Yeah.
- So 100% of my code is written by Claude Code.
是的，我 100% 的程式碼都是由 Claude Code 寫的。 - Um, I am a fairly prolific coder.
我是一個相當多產的程式設計師。 - Um, and this has been the case even when I worked back at Instagram.
這在我之前在 Instagram 工作時就是這樣了。 - I was like one of the top few most productive engineers.
我是最高產的幾位工程師之一。 - Um and that’s actually that’s still the case uh here at Anthropic.
而這在 Anthropic 這裡其實仍然如此。 - » Wow.
- Even as head of head of the team.
即使身為團隊負責人。 - » Yeah.
- Yeah.
- Do still do a lot of coding.
確實還是寫很多程式。 - Um and so every you know every day I ship like 10 20 30 p requests something like that » every day.
所以每天我大概會送出 10、20、30 個 pull request，大概這樣 » 每天。 - » Every day.
- Yeah.
- » Good god.
- » Uh 100% written by Claude Code.
» 100% 由 Claude Code 撰寫。 - I have not edited a single line by hand since uh November.
自從十一月以來，我沒有手動編輯過任何一行程式碼。 - And yeah, that that’s been it.
是的，就是這樣。 - I do look at the code.
我確實會看程式碼。 - So I I don’t think we’re kind of at the point yet where you can be totally hands-off, especially when there’s a lot of people, you know, like running the program.
所以我不認為我們已經到了可以完全不管的程度，特別是當有很多人在執行這個程式的時候。 - You have to make sure that it’s correct.
你必須確保它是正確的。 - You have to make sure it’s safe and so on.
你必須確保它是安全的，諸如此類。 - Um, and then we also have Claude doing automatic code review for everything.
而且我們也讓 Claude 對所有東西做自動化的程式碼審查。 - Um, so here at Anthropic, Claude reviews 100% of poll requests.
所以在 Anthropic 這裡，Claude 審查 100% 的 pull request。 - Um, there’s still layer of like human review after it, but you kind of like you still do want some of these checkpoints like you still want a human looking at the code.
在那之後仍然有人類審查的層級，但你確實還是需要這些檢查點，你還是希望有人類在看程式碼。 - um unless it’s like pure prototype code that you know it’s not going to run it’s not going to run anywhere it’s just a prototype.
除非是純粹的原型程式碼，你知道它不會在任何地方執行，它只是一個原型。
AI 驅動的構想產生
- » What’s kind of the next frontier?
» 下一個前沿是什麼？ - So at this point 100% of your code is being written by AI.
到目前為止你 100% 的程式碼都是由 AI 撰寫的。 - This is clearly where everyone is going in software engineering.
這顯然是軟體工程中每個人都在走的方向。 - That felt like a crazy milestone.
那感覺像是一個瘋狂的里程碑。 - Now it’s just like of course this is the world now.
現在就只是「當然，這就是現在的世界了」。 - What’s what’s kind of the next big shift to how software is written that either your team’s already operating in or you think will head towards?
你的團隊已經在實踐的，或者你認為會朝向的，下一個軟體撰寫方式的重大轉變是什麼？ - I think something that’s happening right now is Claude is starting to come up with ideas.
我覺得現在正在發生的一件事是 Claude 開始提出想法了。 - Um so Claude is looking through feedback.
Claude 在瀏覽回饋意見。 - It’s uh looking at bug reports.
它在查看 bug 回報。 - It’s looking at um you know like telemetry and and things like this and it’s starting to come up with ideas for bug fixes and things to ship.
它在看遙測資料之類的東西，然後它開始提出 bug 修復和要交付的功能的想法。 - So it’s just starting to get a little more um you know like a little more like a co-orker or something like that.
所以它開始變得更像是一個同事之類的。 - I think the second thing is we’re starting to branch out of coding a little bit.
我覺得第二件事是我們開始稍微跳脫出寫程式的範疇了。 - So I think at this point it’s safe to say that coding is largely solved.
所以我認為在這個時間點可以說寫程式基本上已經被解決了。 - At least for the kind of programming that I do, it’s just a solved problem because Claude can do it.
至少對於我做的那種程式設計來說，這就是一個已解決的問題，因為 Claude 可以做到。
更廣泛的任務自動化
- And so now we’re starting to think about okay like what’s next?
所以現在我們開始思考，好的，下一步是什麼？ - What’s beyond this?
超越這個的是什麼？ - There’s a lot of things that are kind of adjacent to coding.
有很多事情是跟寫程式相鄰的。 - Um and I think this is going to be coming.
我覺得這些都會到來。 - But also just you know general tasks, you know, like I use Cowork every day now to do all sorts of things that are just not related to coding at all and just to do it automatically.
但也包括一般性的任務，像是我現在每天都用 Cowork 來做各種完全跟寫程式無關的事情，而且是自動完成的。 - Like for example, I had to pay a parking ticket the other day.
比如說，我前幾天要繳一張停車罰單。 - I just had Cowork do it.
我就讓 Cowork 去做了。 - um all of my project management for the team uh Cowork does all of it.
我團隊所有的專案管理，Cowork 全部包辦。 - It’s like syncing stuff between spreadsheets and messaging people on Slack and email and all this kind of stuff.
就像在試算表之間同步資料、在 Slack 上傳訊息給人、寄 email，所有這類的事情。 - So I think the frontier is something like this and I I don’t think it’s coding because I think coding is you know it’s pretty much solved and over the next few months I think what we’re going to see is just across the industry it’s going to become increasingly solved you know for every kind of codebase every tech stack that people work on.
所以我認為前沿是這類的事情，我不認為是寫程式，因為我覺得寫程式基本上已經被解決了，在接下來幾個月我們會看到整個產業都會越來越被解決，無論是什麼樣的程式碼庫、什麼樣的技術堆疊。 - » This idea of helping you come up with what to work on is so interesting.
» 這個幫你想出要做什麼的概念真的很有趣。
利用 AI 處理產品回饋
- A lot of people listening to this are product managers and they’re probably sweating.
很多聽這個的人是產品經理，他們可能正在冒汗。 - How do you use Claude for this?
你怎麼用 Claude 來做這件事？ - Do you just talk to it?
你就直接跟它說話嗎？ - Is there anything clever you’ve come up with to help you use it to come up with what to build?
你有想出什麼聰明的方法來幫助你用它來決定要做什麼嗎？ - » Honestly, the simplest thing is like open Claude Code or Cowork and point it at a Slack thread.
» 老實說，最簡單的方法就是打開 Claude Code 或 Cowork，然後把它指向一個 Slack 討論串。 - Um, you know, like for us, we have this channel that that’s all the internal feedback about Claude Code.
對我們來說，我們有一個頻道是所有關於 Claude Code 的內部回饋。 - Since we first released it, even in like 2024 internally, it’s just been this fire hose of feedback.
自從我們第一次發布它，即使是 2024 年在內部，它就一直是大量回饋不斷湧入。 - Um, and it’s the best.
而且這是最棒的。 - And like in the early days, what I would do is anytime that someone sends feedback, I would just go in and I would fix every single thing as fast as I possibly could.
在早期，我做的就是每當有人送出回饋，我就會盡快去修復每一件事情。 - So like within a minute, within 5 minutes or whatever.
像是在一分鐘內、五分鐘內之類的。 - And this just really fast feedback cycle, it encourages people to give more and more feedback.
這種非常快速的回饋循環，它鼓勵人們給出越來越多的回饋。 - It’s just so important cuz it makes them feel heard cuz you know like usually when you use a product, you give feedback, it just goes into a black hole somewhere and then you don’t give feedback again.
這非常重要，因為它讓人們感覺被聽見了。通常當你使用一個產品，你給了回饋，它就進了某個黑洞，然後你就不再給回饋了。 - So if you make people feel heard, then they want to contribute and they want to help make the thing better.
所以如果你讓人們感覺被聽見了，他們就會想要貢獻，想要幫助把東西做得更好。 - Um, and so now I kind of do the same thing, but Claude honestly does a lot of the work.
所以現在我做的是差不多的事情，但老實說 Claude 做了很多工作。 - So I pointed at the channel and it’s like, “Okay, here’s a few things that I can do.
所以我把它指向那個頻道，然後它就說「好的，這裡有幾件我可以做的事情。 - I just put up a couple PRs.
我剛送了幾個 PR。 - Want to take a look at that one?”
要看一下嗎？」 - I’m like, “Yeah.”
我就說「好。」 - Have you noticed that it is getting much better at this?
你有注意到它在這方面變得越來越好嗎？
突破程式碼審查瓶頸
- Because this is kind of the holy grail, right?
因為這某種程度上是聖杯，對吧？ - Now it’s like, “Cool, building solved.”
現在就像是「酷，建構的問題解決了。」 - Code review became kind of the next bottleneck.
程式碼審查變成了下一個瓶頸。 - All these PRs, who’s going to review them all?
這些所有的 PR，誰來審查它們？ - The next big open question is just like, okay, now we need to now now humans are necessary for figuring out what to build, what to prioritize.
下一個重大的開放問題就是，好的，現在人類有必要去決定要建構什麼、要優先做什麼。 - And you’re saying that that’s where Claude Code is starting to help you.
而你說的是 Claude Code 開始在這方面幫助你了。 - Has it has it gotten a lot better with like say Opus 46 or what’s been the trajectory there?
它有沒有隨著像是 Opus 4.6 變得更好很多？那裡的發展軌跡是什麼？
前所未有的生產力提升
- » Yeah.
- Yeah, it’s improved a lot.
是的，它改進了很多。 - Um I think some of it is kind of like training that we do specific to coding.
我覺得其中一部分是我們針對寫程式所做的特定訓練。 - Um so you know obviously you know best coding model in the world and you know it’s getting better and better like 4.6 is just incredible but also actually a lot of the training that we do outside of coding translates pretty well too.
顯然是世界上最好的寫程式模型，而且它越來越好，像是 4.6 簡直不可思議，但實際上我們在寫程式以外所做的很多訓練也轉移得很好。 - So there is this kind of like transfer where you teach the model to do you know X and it kind of gets better at Y. Um yeah and the the gains have just been insane like at Anthropic over the last year like since we introduced Claude Code we probably I don’t know the exact number we probably like 4x the engineering team or something like this but productivity per engineer has increased 200%.
所以有一種遷移效果，你教模型做 X，它在 Y 上也會變得更好。在 Anthropic 過去一年的增長簡直瘋狂，自從我們引入 Claude Code 以來，我不知道確切的數字，我們大概把工程團隊擴大了 4 倍之類的，但每位工程師的生產力增加了 200%。 - in terms of like pull requests and like this number is just crazy for anyone that actually works in the space and works on deaf productivity because back in a previous life I was at Meta and you know one of my responsibilities was code quality for the company.
以 pull request 來衡量的話，這個數字對任何真正在這個領域工作、致力於開發者生產力的人來說都很瘋狂。因為在前一份工作中我在 Meta，我的一個職責是公司的程式碼品質。 - So this is like the all of our code bases that was my responsibility like Facebook, Instagram, WhatsApp all this stuff.
所以這就像我們所有的程式碼庫都是我的責任，像 Facebook、Instagram、WhatsApp 所有這些東西。 - Um and a lot of that was about productivity because if you make the code higher quality then engineers are more productive and things that we saw is you know in a year with hundreds of engineers working on it you would see a gain of like a few percentage points of productivity something like this.
其中很多是關於生產力的，因為如果你讓程式碼品質更高，工程師就會更有生產力。而我們看到的是，一年裡有數百名工程師在努力，你會看到大概幾個百分點的生產力提升，大概就是這樣。 - Um and so nowadays seeing these gains of just hundreds of percentage points it’s is just absolutely insane.
所以現在看到這些幾百個百分點的增長，簡直是瘋了。 - What’s also insane is just how normalized this has all been like we hear these numbers like of course AI is doing this to us.
同樣瘋狂的是這一切已經變得多麼正常化，我們聽到這些數字就像「當然 AI 正在對我們做這些事」。 - It’s just it’s so unprecedented the amount of change that is happening to software development to building products to just this the world of tech.
正在發生在軟體開發、產品建構、整個科技世界的變化量是如此史無前例。 - It’s just like so easy to get used to it.
就是很容易就習慣了。 - But it’s important to recognize this is crazy.
但重要的是要認識到這是瘋狂的。 - This is something like I have to remind myself once in a while.
這是我偶爾必須提醒自己的事情。 - There’s sort of like a downside of this because the model changes so well there’s actually like there’s many kind of downsides that that we could talk about but I think one of them on a personal level is the model changes so often that I sometimes get stuck in this like old way of of thinking about it and I even find that like new people on the team or even new grads that join do stuff in a more kind of like AGI forward way than I do.
這有某種缺點，因為模型變化得太快，實際上有很多種缺點我們可以討論，但我覺得其中一個在個人層面上是模型變化得太頻繁，我有時候會卡在這種舊的思維方式裡，我甚至發現團隊裡的新人或甚至剛加入的應屆畢業生，他們做事的方式比我更加 AGI 前瞻。
AI 開發的思維轉變
- So like sometimes for example I I I had this case like a couple months ago where there was a memory leak and so like what this is is you know like Claude Code the memory usage is going up and at some point it crashes.
比如有時候，像是幾個月前我遇到一個記憶體洩漏的情況，就是 Claude Code 的記憶體用量一直上升，到某個時候就當掉了。 - This is like a very common kind of engineering problem that you know every engineer has debugged a thousand times and traditionally the way that you do it is you take a heap snapshot you put it into a special debugger you kind of figure out what’s going on you know use these special tools to see what’s happening.
這是一個非常常見的工程問題，每個工程師都除錯過上千次，傳統的做法是你拍一個 heap snapshot，把它放進特殊的除錯器裡，搞清楚發生了什麼事，用這些特殊工具來看發生了什麼。 - Um, and I was doing this and I was kind of like looking through these traces and trying to figure out what was going on.
我當時正在做這件事，翻看這些追蹤資料，試著搞清楚發生了什麼。 - And the engineer that was newer on the team just uh had Claude Code do it and was like, “Hey Claude, it seems like there’s a leak.
然後團隊裡比較新的那個工程師就讓 Claude Code 去做，就說「嘿 Claude，好像有一個洩漏 (leak)。 - Can you figure it out?”
你能找出來嗎？」 - And so like Claude Code did exactly the same thing that I was doing.
然後 Claude Code 做了跟我一模一樣的事情。 - It it took the heap snapshot.
它拍了 heap snapshot。 - It wrote a little tool for itself so it can kind of like analyze it itself.
它為自己寫了一個小工具，這樣它就可以自己分析。 - Um, it was sort of like a just in time program.
這有點像是一個即時程式。 - Uh, and it found the issue and put up a pull request faster than I could.
然後它找到問題並提交了一個 pull request，比我還快。 - So it’s it’s something where like for those of us that have been using the model for a long time, you still have to kind of transport yourself to the current moment and not get stuck back in an old model because it’s not sonnet 3.5 anymore.
所以對於我們這些使用模型很久的人來說，你仍然必須把自己帶到當下，不要卡在舊模型的思維裡，因為它已經不是 Sonnet 3.5 了。 - The new models are just completely completely different.
新的模型完全完全不同。 - Uh and just this this mindset shift is is very different.
這個思維轉變是非常不同的。
團隊 AI 協作原則
- I hear you have these very specific principles that you’ve codified for your team that when people join you you kind of walk them through them.
我聽說你有這些非常具體的原則，你已經為你的團隊制定成文，當有人加入時你會帶他們過一遍。 - I believe one of them is what’s better than doing something having Claude do it.
我相信其中一個是「比自己做事更好的是什麼？讓 Claude 去做。」 - And it feels like that’s exactly what you describe with this memory leak is just like you almost forgot that principle of like okay let me see if Claude can solve this for me.
而且感覺這正是你描述的那個記憶體洩漏的情況，就是你幾乎忘了那個原則，像是「好，讓我看看 Claude 能不能幫我解決這個問題。」 - There’s this uh there’s this interesting thing that happens also when you um when you underfund everything a little bit uh because then people are kind of forced to clify and this is something that we see.
當你稍微給不足資源的時候，會發生一個有趣的事情，因為那樣人們就被迫要用 Claude，而這是我們看到的現象。 - So you know for work where sometimes we just put like one engineer on a project and the way that they’re able to ship really quickly because they want to ship quickly.
比如有時候我們就只派一個工程師在一個專案上，他們能夠很快交付，因為他們想要快速交付。 - This is like an intrinsic motivation that comes from within is just wanting to do a good job.
這是一種來自內在的動機，就是想要做好工作。 - One if you have a good idea you just really want to get it out there.
如果你有一個好想法，你就真的很想把它實現出來。 - No one has to force you to do that.
沒有人需要強迫你去做。 - That comes from you.
那來自於你自己。 - Um and and so if you have claude, you can just use that to automate a lot of work.
所以如果你有 Claude，你就可以用它來自動化很多工作。 - Uh and that that’s kind of what we see over and over.
而這就是我們一再看到的情況。
以資源不足激發創新
- So I think that’s kind of like one principle is underfunding things a little bit.
所以我認為其中一個原則就是稍微給不足資源。
以 AI 鼓勵速度
- I think another principle is just encouraging people to go faster.
我覺得另一個原則就是鼓勵人們更快。 - So if you can do something today, you should just do it today.
所以如果你今天就能做一件事，你就應該今天做。 - And this is something we we really really encourage on the team.
這是我們在團隊裡非常非常鼓勵的事情。 - Early on it was really important because it was just me and so our only advantage was speed.
早期這真的很重要，因為只有我一個人，所以我們唯一的優勢就是速度。 - that’s the only way that we could ship a product that would compete in this very crowded coding market.
那是我們能在這個非常擁擠的程式設計市場中推出有競爭力產品的唯一方式。 - But nowadays, it’s still very much a principle we have on the team.
但如今這仍然是我們團隊非常重視的一個原則。 - And if you want to go faster, a really good way to do that is to just have Claude do more stuff.
如果你想更快，一個很好的方法就是讓 Claude 做更多事情。 - Um, so he it just very much encourages that.
所以它就是非常鼓勵這樣做。 - This idea of underfunding, it’s so interesting because in general there’s this feeling like AI is going to allow you to not have as many employees, not have as many engineers.
這個資源不足的概念非常有趣，因為一般來說有一種感覺是 AI 會讓你不需要那麼多員工、不需要那麼多工程師。 - And so it’s not only you can be more productive.
所以不只是你可以更有生產力。 - What you’re saying is that you will actually do better if you underfund.
你說的是如果你給的資源不足，你實際上會做得更好。 - It’s not just that AI can make you faster.
不只是 AI 可以讓你更快。 - It’s you will get more out of the AI tooling if you have fewer people working on something.
而是如果你讓更少的人做一件事，你會從 AI 工具中得到更多。 - Yeah.
是的。 - If you if you hire great engineers, they’ll figure out how to do it.
如果你僱用優秀的工程師，他們會想出如何做到。 - And uh especially if you empower them to do it.
特別是如果你賦予他們權力去做。 - This is something I actually talk talk a lot about with uh you know with like CTO’s and kind of all sorts of companies.
這是我其實經常和各種公司的 CTO 們討論的事情。 - My advice generally is don’t try to optimize.
我的建議通常是不要試圖最佳化。 - Don’t don’t try to cost cut at the beginning.
不要在一開始就試圖削減成本。 - Start by just giving engineers as many tokens as possible.
先從給工程師盡可能多的 token 開始。 - And now now you’re starting to see companies like you know at Anthropic we have you know everyone can use a lot of tokens.
現在你開始看到像 Anthropic 這樣的公司，每個人都可以使用大量的 token。 - We’re starting to see this come up as like a perk at some companies.
我們開始看到這在一些公司成為一種福利。 - Like if you join you get unlimited tokens.
像是如果你加入，你就有無限的 token。 - This is a thing I very much encourage because um it makes people free to try these ideas that would have been too crazy and then if there’s an idea that works then you can figure out how to scale it and that’s the point to kind of optimize and to cost cut figure out like you know maybe you can do it with haiku or with sonnet instead of opus or whatever but at the beginning you just want to throw a lot of tokens at it and see if the idea works and give engineers the freedom to do that.
這是我非常鼓勵的一件事，因為它讓人們可以自由嘗試那些原本太瘋狂的想法，然後如果有一個想法可行，你再想辦法如何擴展它，那才是該最佳化和削減成本的時候，搞清楚像是也許你可以用 Haiku 或 Sonnet 而不是 Opus 之類的，但在一開始你就是想投入大量的 token 看看這個想法是否可行，並給工程師這樣做的自由。 - So the advice here is uh just be be loose with your tokens with this the cost on on using these models.
所以這裡的建議是對你的 token 寬鬆一點，對使用這些模型的成本不要太計較。 - People hearing this may be like of course he works at Anthropic.
聽到這個的人可能會想，當然他在 Anthropic 工作。 - He wants us to use as many tokens as possible.
他想讓我們用盡可能多的 token。 - But you’re what you’re saying here is the the most interesting innovative ideas will come out of someone just kind of taking it to the max and seeing what’s possible.
但你這裡要說的是，最有趣的創新想法會來自於某個人把它推到極限，看看什麼是可能的。 - » Yeah.
- And I and I think the reality is like at small scale like you know you’re not going to get like a giant bill or anything like this.
» 而且我覺得現實是，在小規模的時候，你不會收到什麼巨額帳單之類的。 - Like if it’s an individual engineer experimenting, it’s the token cost is still probably relatively low relative to their salary or you know other costs of running the business.
如果是一個工程師在實驗，token 的成本相對於他們的薪資或其他營運成本來說可能還是相對低的。 - So it it’s actually like not not a huge cost as the thing scales up.
所以當事情擴大規模時，它其實不是一個很大的成本。 - So like let’s say you know they build something awesome and then it takes a huge amount of tokens and then the cost becomes pretty big.
比如說他們建構了一個很棒的東西，然後它需要大量的 token，然後成本變得很大。 - That’s the point at which you want to optimize it.
那就是你想要最佳化它的時候。 - But don’t don’t do that too early.
但不要太早做那件事。 - » Have you seen companies where their uh token cost is higher than their salary?
» 你有看到 token 成本比薪資還高的公司嗎？
Token 成本與規模化
- Is that a trend you think we’re going to find and see?
你覺得這是我們會發現和看到的趨勢嗎？ - » You know, at Anthropic, we’re starting to see some engineers that are spending, you know, like hundreds of thousands a month in in tokens.
» 你知道，在 Anthropic，我們開始看到一些工程師每個月在 token 上花費數十萬美元。 - Um, so we’re starting to see this a little bit.
所以我們開始稍微看到這種情況了。 - Um, there’s some companies that are we’re starting to see similar things.
有一些公司我們也開始看到類似的情況。 - Yeah.
個人程式設計歷程
- » Going back to coding, do you miss writing code?
» 回到寫程式這件事，你會想念寫程式嗎？ - Is this something you’re kind of sad about that this is no longer a thing you will do as a software engineer?
作為一個軟體工程師，你會因為這不再是你會做的事情而感到難過嗎？ - It’s funny for me, you know, like when when I learned engineering, for me it was very practical.
對我來說很有趣，當我學工程的時候，對我來說是非常實用的。 - I learned engineering so I could build stuff and for me I was I was selftaught, you know, like I studied economics in school, but um I didn’t study CS, but I I taught myself engineering kind of early on.
我學工程是為了能建構東西，而且我是自學的，我在學校念的是經濟學，但我沒有學 CS，但我很早就自學了工程。 - I was programming in like middle school and from the very beginning it was very practical.
我在國中就開始寫程式了，從一開始就非常實用導向。 - So I actually like I learned to code so that I can cheat on a math test.
所以我其實學寫程式是為了在數學考試中作弊。 - That was like the first thing we had these like graphing calculators and you know I just programmed the answer into » TI83.
那是最初的事情，我們有那些繪圖計算機，我就把答案編程進去 » TI-83。 - » T83 plus.
- Yeah.
- Yeah.
- Exactly.
- » Plus.
- Yeah.
- So like I programmed the answers in and then the next like math test whatever like the next year that it was just like too hard.
所以我把答案編程進去，然後下一次數學考試，像是隔年那次就太難了。 - Like I couldn’t program all the answers in because I didn’t know what the questions were.
我沒辦法把所有答案都編程進去，因為我不知道題目是什麼。 - And so I had to write like a little solver so that it it was a program that would just like solve these like uh you know these al algebra questions or whatever.
所以我必須寫一個小的解題器，一個程式可以解這些代數題之類的。 - And then I figured out you can get a little cable, you can give the program to the rest of the class and then the whole class gets A’s.
然後我發現你可以拿一條小傳輸線，把程式傳給全班，然後全班都拿 A。 - But then we all got caught and the teacher told us to knock it off.
但後來我們都被抓到了，老師叫我們住手。 - But from the very beginning it’s it’s always just been very practical for me where programming is a way to build a thing.
但從一開始，對我來說寫程式一直都是非常實用的，寫程式是一種建構東西的方式。 - It’s not the end in itself.
它本身不是目的。 - At some point I personally fell into the rabbit hole of kind of like the the beauty of of programming.
在某個時候我個人掉進了程式設計之美的兔子洞。 - Um so like I I wrote a book about TypeScript.
比如我寫了一本關於 TypeScript 的書。 - Um, I started the actually at the time it was the world’s biggest uh, TypeScript meetup just because I fell in love with the language itself.
我創辦了當時世界上最大的 TypeScript meetup，純粹是因為我愛上了這個語言本身。 - Uh, and I kind of got in deep into like functional programming and and all this stuff.
然後我深入研究了函數式程式設計（functional programming）和所有這些東西。 - I think a lot of coders they get distracted by this.
我覺得很多程式設計師會被這些分心。 - For me, it was always sort of um they there is a beauty to programming and especially to functional programming.
對我來說，程式設計確實有它的美，特別是函數式程式設計。 - There’s a beauty to type systems.
型別系統有它的美。 - Um, there there’s a certain kind of like this like buzz that you get like when you solve like a really a really complicated uh math problem.
有一種特別的興奮感，就像你解開一個非常複雜的數學題時的那種感覺。 - It’s kind of similar when you kind of balance the types or you know the program is just like really beautiful but it’s really not the end of it.
當你把型別平衡好，或者程式真的很漂亮的時候，那種感覺很類似，但那真的不是終點。 - Um I think for me coding is very much a tool and it’s a way to do things.
對我來說寫程式非常像是一種工具，是一種做事的方式。 - Uh that said not everyone feels this way.
話雖如此，不是每個人都這麼想。 - So for example you know like there’s one engineer uh on the team Lena who you know was still writing C++ on the weekends by hand because you know for her she just really enjoys writing C++ by hand.
比如說團隊裡有一個工程師 Lena，她週末還是會手寫 C++，因為她就是很享受手寫 C++。 - And so everyone is different and I think even as this field changes, even as everything changes, there’s always space to do this, there’s always space to enjoy the art um and to and and to kind of do do things by hand uh if you want.
所以每個人都不一樣，我覺得即使這個領域改變了，即使一切都改變了，總是有空間去做這些事情，總是有空間去享受這門藝術，如果你想的話，總是可以手動做事。 - » Do you worry about your skills atrophing as an engineer?
» 你會擔心你作為工程師的技能退化嗎？
技能演進與工程的未來
- Is that something you worry about or is it just like, you know, this is just the way it’s going to go?
這是你會擔心的事情，還是就像「這就是事情會發展的方向」？ - » I think it’s just the way that that it happens.
» 我覺得這就是事情發生的方式。 - I I don’t worry about it too much personally.
我個人不太擔心這個。 - I think for me like programming is on is on a continuum and you know like way back in the day you know like software actually is like relatively new right like if you look at the way programs are written today like using software that’s running on a virtual machine or something this has been the way that we’ve been writing programs since probably the 1960s so you know it’s been you know like 60 years or something like that.
對我來說，程式設計是在一個連續光譜上的。回溯到過去，軟體其實是相對新的東西，如果你看今天程式的寫法，像是用在虛擬機器上執行的軟體，這是我們從大概 1960 年代以來一直在寫程式的方式，所以大概 60 年左右。 - Before that it was punch cards.
在那之前是打孔卡片。 - Before that it was switches.
在那之前是開關。 - Before that it was hardware.
在那之前是硬體。 - And before that it was just you know like literally pen and paper.
在那之前就是紙和筆。 - It was like a room a room full of people that were doing math on on paper.
就像一個房間裡滿是在紙上做數學運算的人。 - And so, you know, programming has always changed in this way.
所以程式設計一直都是這樣在改變的。 - In some ways, you still want to understand the layer under the layer because it helps you be a better engineer.
在某些方面，你仍然想要了解底層的底層，因為這幫助你成為更好的工程師。 - And I think this will be the case maybe for the next year or so.
我覺得大概在接下來一年左右還是這樣。 - Um, but I think pretty soon it just won’t really matter.
但我覺得很快它就不太重要了。 - It’s just going to be kind of like the the assembly code wring running under the programmer or something like this.
它就會變成像是在程式設計師底下執行的組合語言之類的東西。 - uh at an emotional level, you know, I I feel like I’ve always had to learn new things.
在情感層面上，我覺得我一直都必須學習新東西。 - And as a programmer, it’s actually not it doesn’t feel that new because there’s always new frameworks, there’s always new languages.
作為一個程式設計師，它其實感覺不是那麼新，因為總是有新的框架、新的語言。 - It’s just something that we’re quite comfortable with in the field.
這在這個領域裡是我們相當習慣的事情。 - But at the same time, I you know, this isn’t true for everyone.
但同時，這不是每個人都如此。 - And I think for some people, they’re going to feel a greater sense of, I don’t know, maybe like loss or nostalgia or atrophy or something like this.
我覺得對某些人來說，他們會感受到更大的，我不知道，也許是失落感或懷舊感或退化感之類的東西。 - I don’t know if you saw this, but Elon was saying that uh why isn’t the AI just writing binary straight to binary?
我不知道你有沒有看到，Elon 說為什麼 AI 不直接寫二進位到二進位？ - Uh because what’s the point of all this, you know, programming abstraction in the end?
因為這些程式設計抽象層到底有什麼意義？ - » Yeah, it’s a good question.
» 是的，這是一個好問題。 - I mean, it totally can do that if you wanted to.
我的意思是，如果你想的話它完全可以做到。
AI 轉型的歷史類比
- » Oh, man.
- So, what I’m hearing here is in terms there’s always this question, should I learn to code?
所以我在這裡聽到的是，一直都有這個問題：我應該學寫程式嗎？ - Should people in school learn to code?
在學校的人應該學寫程式嗎？ - Uh what I heard from you is your take is in like a year or two, you don’t really need to.
我從你那裡聽到的是，你的看法是大概一兩年後，你其實不太需要了。 - My take is I think for for people that are using um there that are using Claude Code that are using agents to code today you still have to understand the layer under but yeah in a year or two it’s not going to matter.
我的看法是，對於現在正在使用 Claude Code、使用 agent 來寫程式的人，你仍然需要了解底層，但是一兩年後就不重要了。 - I I was thinking about um what is the right like historical analog for this cuz like like somehow we have to situate this thing in history and and kind of figure out when have we gone through similar transitions.
我一直在想，什麼才是正確的歷史類比，因為我們必須把這件事放在歷史脈絡中，弄清楚我們什麼時候經歷過類似的轉變。
印刷術的類比
- What’s the right kind of mental model for this?
什麼才是正確的思維模型？ - I think the thing that’s come closest for me is the printing press.
我覺得最接近的是印刷術。 - And so you know if you look at Europe in uh you know like in the in the mid the mid400s literacy was actually very low.
如果你看歐洲在 1400 年代中期，識字率其實非常低。 - Uh there was sub 1% of the population it was scribes that uh you know they were the ones that did all the writing.
不到 1% 的人口，是抄寫員在做所有的書寫工作。 - They they were the ones that did all the reading.
他們也是做所有閱讀工作的人。 - They were employed by like lords and kings that often were not literate themselves.
他們受僱於領主和國王，而這些人自己往往不識字。 - And so you know it was their job of this very tiny percent of the population to do this.
所以這是人口中極小比例的人的工作。 - And at some point the you know Gutenberg and and the printing press came along and there was this crazy stat that in the 50 years after the printing press was uh built there was more printed material created than in the c in the in the thousand years before and so the the volume of printed material just went way up.
然後在某個時候，Gutenberg 和印刷術出現了，有一個瘋狂的統計數據：印刷術發明後的 50 年內產生的印刷品，比之前一千年還多，所以印刷品的數量急劇上升。 - Uh the cost went way down.
成本大幅下降。 - It went down something like 100x over the next 50 years.
在接下來的 50 年裡下降了大約 100 倍。 - And if you look at literacy, you know, it actually took a while because learning to read and write is, you know, it’s quite hard.
如果你看識字率，它其實花了一段時間，因為學習讀寫是很困難的。 - It takes an education system.
需要一個教育體系。 - It takes free time.
需要空閒時間。 - You it takes like not having to work on a farm all day so that you actually have time for education and things like this.
需要你不用整天在農場工作，這樣你才有時間接受教育之類的。 - But over the next 200 years, it went up to like 70% globally.
但在接下來的 200 年裡，全球識字率上升到了大約 70%。 - So I think this is the kind of thing that we might see is a similar kind of transition.
所以我覺得我們可能會看到類似的轉變。 - And there was uh there was actually this interesting um historical document where there was an interview with some like scribe in the 1400s about like how do you feel about the printing press?
其實有一份有趣的歷史文獻，裡面有一個 1400 年代某位抄寫員的訪談，問他對印刷術有什麼感覺。 - And they were actually very excited because they were like actually the thing that I don’t like doing is copying between books.
他們其實非常興奮，因為他們說：其實我不喜歡做的事情是在書本之間抄寫。 - The thing that I do like doing is drawing the art in books and then doing the book binding.
我喜歡做的是在書中畫插圖，然後做裝訂。 - And I’m really glad that now my time is freed up.
我很高興現在我的時間被釋放出來了。 - And it’s interesting like as an engineer I sort of felt like a peril with this.
有趣的是，作為一個工程師，我覺得這有種相似之處。 - It’s like this is sort of how I feel where I don’t have to do the tedious work anymore of coding because this has always been sort of the detail of it.
這就像是我的感受，我不用再做寫程式中那些繁瑣的工作了，因為這一直都是其中細節的部分。 - It’s always been the tedious part of it and kind of like messing with like git and kind of using all these different tools.
一直都是繁瑣的部分，像是搞 git 和使用所有這些不同的工具。 - That that was not the fun part.
那不是有趣的部分。 - The fun part is figuring out what to build and coming up with this.
有趣的部分是弄清楚要建造什麼，以及想出這些東西。 - It’s uh it’s talking to users.
是跟使用者交談。 - It’s thinking about these big systems.
是思考這些大型系統。 - It’s thinking about the future.
是思考未來。 - It’s collaborating with you know other people on the team.
是跟團隊中的其他人協作。 - And that’s what I get to do more of now.
而這些就是我現在能做更多的事情。 - And what’s amazing is that the tool you’re building allows anybody to do this.
令人驚奇的是，你正在打造的工具讓任何人都能做到這些。 - People that have no technical experience can do exactly what you’re describing.
沒有技術經驗的人也能做到你所描述的事情。
現代工程工作流程的轉變
- Like I’m I’ve been doing a bunch of random little projects and any it’s just like anytime you get stuck just like help me figure this out and you get on block.
像是我一直在做一堆隨機的小專案，任何時候你卡住了就說「幫我搞清楚這個」，然後你就解除阻塞了。 - Like I used to I was an engineer for early in my career for 10 years and I just remember spending so much time on like libraries and dependencies and things and just like oh my god what do I do and then looking on stack overflow and now it’s just like help me figure this out and here’s step by step one two three four okay we got this.
像是我以前，在我職涯早期做了 10 年工程師，我記得花了很多時間在函式庫和相依性之類的東西上，就像「天哪我該怎麼辦」然後去看 Stack Overflow，現在就是「幫我搞清楚這個」然後就是一步一步，一二三四，搞定了。 - » Yeah exactly exactly I was talking to an engineer earlier today they’re like they’re writing some service and go and you know it’s been like a month already and they they built up the service like it’s working quite well and then I was like okay so like how do you feel writing it?
» 對，完全正確。我今天稍早跟一個工程師聊天，他們在用 Go 寫一個服務，已經大概一個月了，他們把服務建起來了，運作得相當好，然後我就問「你寫起來感覺如何？」 - He was like, you know, like I I still don’t really know Go, but and I think we’re going to start to see more and more of this.
他說，你知道的，我其實還是不太懂 Go。但我覺得我們會開始看到越來越多這樣的情況。 - It’s like if you know that it works correctly and efficiently, then you you don’t actually have to know all the details.
就像是如果你知道它能正確且有效率地運作，那你其實不需要知道所有細節。 - Clearly, the life of a software engineer has changed dramatically.
很明顯，軟體工程師的生活已經發生了巨大的變化。 - It’s like a whole new job now as of the past year or two.
在過去一兩年裡，這就像是一份全新的工作。
AI 對鄰近角色的影響
- What do you think is the next role that will be most impacted by AI within either within tech like you know product managers, designers or even outside tech just like what do you think where do you think AI is going next?
你覺得下一個最受 AI 影響的角色會是什麼？在科技業內像是產品經理、設計師，或者甚至在科技業外，你覺得 AI 接下來會往哪裡走？ - » I think it’s going to be a lot of the roles that are adjacent to engineering.
» 我覺得會是很多與工程相鄰的角色。 - Um so yeah it could be like product managers, it could be design, could be data science.
所以可能是產品經理、可能是設計、可能是資料科學。 - It is going to expand to pretty much any kind of work that you can do on a computer because the model is just going to get better and better at this.
它會擴展到幾乎任何你能在電腦上做的工作，因為模型只會在這方面越來越好。 - Um, and you know, like this is the Cowork product is kind of the first way to get at this, but it’s just the first one.
這個 Cowork 產品算是第一個切入方式，但它只是第一個。 - And it’s the thing that I think brings AI to a agentic AI to people that haven’t really used it before, and people are starting just to to to get a sense of it for the first time.
它把自主式 AI 帶給了之前沒有真正用過的人，人們開始第一次感受到它。 - When I think back to engineering a year ago, no one really knew what an agent was.
當我回想一年前的工程界，沒有人真正知道 agent 是什麼。 - No one really used it.
沒有人真正在用它。 - But nowadays, it’s just the way that, you know, we do we do our work.
但現在，這就是我們工作的方式。 - And then when I look at non-technical work today um so you know like or maybe semi-technical like product work and you know like data science and things like this when you look at the kinds of AI that people are using it’s all it’s always these like conversational AI it’s like a chatbot or whatever but no one really has used an agent before and this word agent just gets thrown around all the time and it’s just like so misused it’s like lost all meaning but agent actually has like a very
然後當我看今天的非技術性工作，或者也許是半技術性的像產品工作和資料科學之類的，當你看人們在用的 AI 種類，全都是對話式 AI，就是聊天機器人之類的，但沒有人真正用過 agent，而這個詞一直被到處亂用，已經被濫用到失去所有意義了，但 agent 其實有一個非常 - specific technical meaning which is it’s a it’s a AI it’s a LM that’s able to use tools.
具體的技術含義，就是它是一個能夠使用工具的 AI，是一個能使用工具的語言模型。 - So it doesn’t just talk, it can actually act and it can interact with your system and you know this means like it can use your Google docs and it can it can send email.
所以它不只是說話，它實際上能行動，能跟你的系統互動，這意味著它可以使用你的 Google 文件、可以寄送電子郵件。 - It can run commands on your computer and do all this kind of stuff.
它可以在你的電腦上執行指令，做各種這類的事情。 - So I think like any kind of job where you do you use computer tools in this way.
所以我覺得任何你以這種方式使用電腦工具的工作。 - I think this is going to be next.
我覺得這些會是下一波。
社會與產業的迫切議題
- This is something we have to kind of figure out as a as a society.
這是我們作為一個社會必須去弄清楚的事情。 - This is something we have to figure out as an industry.
這是我們作為一個產業必須去弄清楚的事情。 - Um and I think for me also this is one of the reasons it it feels very important and urgent to do this work at Anthropic because I think we take this very very seriously.
我覺得對我來說，這也是為什麼在 Anthropic 做這件事感覺非常重要和緊迫的原因之一，因為我覺得我們非常非常認真地看待這件事。 - Um and so now you know we have economists we have uh policy folks we have social impact folks this is something we just want to talk about a lot so as society we can kind of figure out what to do because it shouldn’t be up to us.
所以現在我們有經濟學家、有政策人員、有社會影響力人員，這是我們想要多多討論的事情，這樣作為社會我們才能弄清楚該怎麼做，因為這不應該由我們來決定。
AI 與就業市場動態
- » So the big question which you’re kind of alluding to is jobs and job loss and things like that.
» 所以你暗示的那個大問題是工作和失業之類的事情。 - There’s this concept of Jevans paradox of just as we can do more we hire more and it’s not actually as scary as it looks.
有一個 Jevons 悖論的概念，就是當我們能做更多時，我們反而僱用更多人，它其實沒有看起來那麼可怕。 - What have you experienced so far I guess with AI becoming a big part of the engineering job?
到目前為止，隨著 AI 成為工程工作的重要部分，你的經驗是什麼？ - Just are you hiring more than if you didn’t have AI and just thoughts on jobs?
你們是否比沒有 AI 時僱用更多人？對就業有什麼想法？ - » Yeah, I mean for our team we’re we’re hiring.
» 是的，我們團隊正在招人。 - Um so Claudeco team is hiring.
Claude Code 團隊正在招人。 - Um if you’re interested just check out the jobs page on on Anthropic.
如果你有興趣，去看看 Anthropic 的職缺頁面。 - Personally, it’s, you know, all this stuff has just made me enjoy my work more.
就個人而言，所有這些東西只是讓我更享受我的工作。 - I have never enjoyed coding as much as I do today because I don’t have to deal with all the minutia.
我從來沒有像今天這樣享受寫程式，因為我不用再處理所有那些瑣碎的事情。 - So, for me personally, it’s been quite exciting.
所以對我個人來說，這非常令人興奮。 - This is something that we hear from a lot of customers where they love the tool, they love Claude Code because it just makes coding delightful again.
這是我們從很多客戶那裡聽到的，他們喜歡這個工具，他們喜歡 Claude Code，因為它讓寫程式再次變得愉快。 - Uh, and that’s just that’s just so fun for them.
這對他們來說就是很有趣。 - But it’s hard to know where this thing is going to go.
但很難知道這件事會走向何方。
AI 與創新的樂觀未來
- And I again I just like I have to reach for these historical analoges.
我又得去借用這些歷史類比了。 - Uh and I I think the printing press is just such a good one because what happened is this technology that was locked away to a small set of people like knowing how to read and write became accessible to everyone.
我覺得印刷術就是一個很好的例子，因為發生的事情是，這項原本被鎖在少數人手中的技術，像是知道如何讀和寫，變得每個人都能使用。 - It was just inherently democratizing.
它本質上就是民主化的。 - Everyone started to be able to do this.
每個人都開始能做到這件事。 - And if that wasn’t the case then something like the Renaissance just could never have happened because a lot of the Renaissance it was about like knowledge spreading.
如果不是這樣的話，像文藝復興這樣的事情就不可能發生，因為文藝復興的很大部分是關於知識的傳播。 - It was about like written records that people used to communicate.
是關於人們用來溝通的書面記錄。 - Um, you know, cuz there were no phones or anything like this.
因為那時候沒有電話或任何類似的東西。 - There was there was no internet at the time.
那時候沒有網路。 - So, it’s about like what does this enable next?
所以，重點是這接下來會催生什麼？ - And I think that’s the very optimistic version of it for me.
我覺得這對我來說是非常樂觀的版本。 - And that’s the part that I’m really excited about.
那也是我真正感到興奮的部分。 - It’s just unimaginable, you know, like we couldn’t be talking today if the printing press hadn’t been invented.
那是難以想像的，如果印刷術沒有被發明，我們今天就不可能在這裡對話。 - Like our microphones wouldn’t exist.
像我們的麥克風不會存在。 - None of the things around us would exist.
我們周圍的東西都不會存在。 - it just wouldn’t be possible to coordinate such a large group of people if that wasn’t the case.
如果不是這樣的話，就不可能協調這麼大一群人。 - And so I imagine a world, you know, a few years in the future where everyone is able to program.
所以我想像一個幾年後的世界，每個人都能寫程式。 - And what does that unlock?
那會解鎖什麼？ - Anyone can just build software anytime.
任何人隨時都能建造軟體。 - And I have no idea.
我完全不知道。 - It’s just the same way that, you know, in the 1400s, no one could have predicted this.
就像在 1400 年代，沒有人能預測到今天這樣。 - Um, I think it’s the same way.
我覺得道理是一樣的。 - But I do think in the meantime, it’s going to be very disruptive and it’s going to be painful for a lot of people.
但我確實覺得在此期間，它會非常具有破壞性，對很多人來說會很痛苦。 - Um, and again, as a society, this is a conversation that we have to have and this is a thing that we have to figure out together.
再次強調，作為一個社會，這是我們必須進行的對話，這是我們必須一起弄清楚的事情。
實驗 AI 工具
- » So, for folks hearing this that want to succeed and, you know, make it in this crazy turmoil we’re entering, any advice?
» 對於聽到這些、想要成功、想要在我們即將進入的瘋狂動盪中活下來的人，有什麼建議嗎？ - Is it, you know, play with AI tools, get really proficient at the latest stuff?
是玩 AI 工具、對最新的東西變得非常熟練嗎？ - Is there anything else that you recommend to help people uh stay ahead?
還有什麼你建議能幫助人們保持領先的嗎？ - Yeah, I think that’s pretty much it.
是的，我覺得差不多就是這樣。 - Uh, experiment with the tools, get to know them, don’t be scared of them.
去實驗這些工具，去認識它們，不要害怕它們。 - um just you know dive in try them be on the bleeding edge beyond the frontier.
就是投入進去，試試它們，站在最前沿。
擁抱通才與跨領域
- Maybe the second piece of advice is try to be a generalist more than you have in the past.
第二個建議可能是，試著比過去更成為一個通才。 - For example, in school a lot of people that study CS they learn to code and they don’t really learn much else.
例如在學校，很多讀資訊科學的人學會寫程式，但他們其實沒有學到太多其他的東西。 - Maybe they learn a little bit of systems architecture or something like this.
也許他們學了一點系統架構之類的。 - But some of the most effective engineers that I work with every day and some of the most effective, you know, like product managers and so on, they cross over disciplines.
但我每天合作的一些最有效率的工程師，以及一些最有效率的產品經理等等，他們會跨越不同領域。 - So on the Claude Code team, everyone codes.
在 Claude Code 團隊中，每個人都會寫程式。 - You know, our product manager codes, our engineering manager codes, our designer codes, our finance guy codes, our data scientist codes.
我們的產品經理會寫程式、我們的工程經理會寫程式、我們的設計師會寫程式、我們的財務人員會寫程式、我們的資料科學家會寫程式。 - Like everyone on the team codes.
團隊中的每個人都會寫程式。 - And and then if I look at particular engineers, people often cross different disciplines.
如果我看特定的工程師，人們經常跨越不同的領域。 - So some of the strongest engineers are hybrid product and infrastructure engineers or product engineers with really great design sense and they’re able to do design also or an engineer that has a really good sense of the business and can use that to figure out what to do next.
一些最強的工程師是混合型的產品和基礎設施工程師，或者是有很好設計感的產品工程師，他們也能做設計，或者是對商業有很好感覺的工程師，能用這個來判斷下一步該做什麼。 - or an engineer that also loves talking to users and can just really channel what what users want to figure out what’s next.
或者是一個也喜歡跟使用者交談的工程師，能夠真正傳達使用者想要什麼來判斷下一步。 - So I think a lot of the people that will be rewarded the most over the next few years, they won’t just be AI native and they don’t just know how to use these tools really well, but also they’re curious and they’re generalists and they cross over multiple disciplines and can think about the broader problem they’re solving rather than just the engineering part of it.
所以我覺得在接下來幾年會獲得最多回報的人，不只是 AI 原生的、不只是知道如何很好地使用這些工具，他們還充滿好奇心、是通才、跨越多個領域，能夠思考他們正在解決的更廣泛的問題，而不只是其中的工程部分。
角色定義的演變
- Do you find these three separate disciplines still useful as a way to think about the team?
你覺得這三個獨立的領域作為思考團隊的方式還有用嗎？ - They’re, you know, engineering, design, uh, product management.
就是工程、設計、產品管理。 - Do you find like those, even though they are now coding and contributing to thinking about what to build, do you feel like those are three roles that will persist long term, at least at this point?
你覺得即使他們現在都在寫程式、都在貢獻思考要建構什麼，你覺得這三個角色會長期持續存在嗎，至少在目前這個時間點？ - I think in the short term it’ll persist, but one thing that we’re starting to see is there’s maybe a 50% overlap in these roles where a lot of people are actually just doing the same thing and some people have specialties.
我覺得短期內會持續，但我們開始看到的一件事是這些角色之間可能有 50% 的重疊，很多人實際上在做同樣的事情，只是有些人有各自的專長。 - for example, I code a little bit more versus cat RPM does a little bit more, you know, coordination or planning or, you know, forecasting or things like this.
例如，我寫程式多一點，而 PM 做多一點的協調、規劃或預測之類的事情。 - » Stakeholder alignment.
» 利害關係人對齊。 - » Stakeholder alignment.
» 利害關係人對齊。 - Exactly.
沒錯。 - I I do think that there’s a future where I think by the end of the year what we’re going to start to see is these start to get even murkier murkier where I think in some places the title software engineer is going to start to go away and it’s just going to be replaced by builder or maybe it’s just everyone’s going to be a product manager and everyone codes or something like this.
我確實覺得有一個未來，我認為到今年年底我們會開始看到這些界線變得更加模糊，我覺得在某些地方「軟體工程師」這個頭銜會開始消失，取而代之的是「建造者」（builder），或者也許每個人都會是產品經理，而且每個人都會寫程式，類似這樣的概念。 - Who says hiring has to be fair?
誰說招募必須公平？
AI 招募優勢（MetaView）
- Every founder and hiring manager I’ve been speaking with these days is feeling the same pressure.
我最近跟每一位創辦人和招募主管交談時，他們都感受到同樣的壓力。 - Hire the best people as fast as possible.
盡可能快地僱用最好的人才。 - But recruiting is time consuming, alignment is hard, and competition for great talent keeps getting tighter.
但招募很耗時，對齊很困難，而優秀人才的競爭越來越激烈。 - That’s why teams like 11 Labs, Brex, Replet, Deal, and 5,000 other organizations use MetaView, the AI company giving high performance teams a real unfair advantage in hiring.
這就是為什麼像 11 Labs、Brex、Replet、Deal 和其他 5,000 個組織使用 MetaView，這家 AI 公司為高績效團隊在招募上提供了真正的不公平優勢。 - They give you a suite of AI agents that behave like recruiting Coworkers.
他們提供一套表現得像招募同事的 AI 代理。 - They find candidates for you based on your exact criteria, take interview notes automatically, gather insights across your hiring process, and help you identify the best candidates in your pipeline.
他們根據你的確切標準為你找到候選人，自動記錄面試筆記，在你的招募流程中收集洞察，並幫助你找出你的人才管線中最好的候選人。 - AI handles the recruiting toil and gives you a real source of truth.
AI 處理招募的繁瑣工作，給你一個真正的事實來源。 - That means hours saved per hire and a team focused on what matters most, winning the right candidates.
這意味著每次招募節省數小時，讓團隊專注於最重要的事情：贏得對的候選人。 - Don’t let your competitors outhire you.
不要讓你的競爭對手在招募上超越你。 - Metav customers close roles 30% faster.
MetaView 的客戶關閉職缺的速度快了 30%。 - Try Metaview today for free and get an extra month of sourcing at metaview.ai/lenny.
今天就免費試用 MetaView，在 metaview.ai/lenny 獲得額外一個月的人才搜尋。 - That’s me.
那就是我。 - Lenny.
- You talked about how you’re enjoying coding more.
你提到你越來越享受寫程式了。
調查結果：工程師、PM、設計師
- I actually did this little informal survey on Twitter.
我其實在 Twitter 上做了一個小型的非正式調查。 - I don’t know if you saw this where I just asked I did three different polls.
不知道你有沒有看到，我做了三個不同的投票。 - I asked engineers, are you enjoying your job more or less since adopting AI tools?
我問工程師，自從採用 AI 工具以來，你更享受還是更不享受你的工作？ - And then I did a separate one for PMs and one for designers.
然後我分別為 PM 和設計師做了一個。 - And both engineers and PMs, 70% of people said they are enjoying their job more and about 10% said they’re enjoying their job less.
工程師和 PM 兩者都有 70% 的人說他們更享受工作了，大約 10% 的人說他們更不享受工作了。 - Designers, interestingly, only 55% said they are enjoying their job more and 20% said they’re enjoying their job less.
有趣的是，設計師只有 55% 說他們更享受工作了，20% 說他們更不享受工作了。 - Thought that was really interesting.
我覺得這真的很有趣。
設計師角色與 AI 採用
- » That’s super interesting.
» 這太有趣了。 - I’ I’d love to talk to these people uh you know, both in the more bucket and the less bucket just to understand.
我很想跟這些人聊聊，不管是更享受的那群還是更不享受的那群，就是想了解一下。 - Do did you get to follow up with any of them?
你有跟他們中的任何人做後續追蹤嗎？ - They a few people replied and we’re actually doing a follow poll that we’ll link to in the show notes of going deeper into some of the stuff, but a lot of there’s like, you know, the factors that make it more fun and less fun.
有一些人回覆了，我們其實正在做一個後續投票，會在節目筆記中附上連結，深入探討其中一些內容，但有很多因素讓工作變得更有趣或更無趣。 - The designers, they didn’t share a lot actually of just like the people that are actually asked just like why are you enjoying your job less?
設計師們其實沒有分享太多，就是那些被問到為什麼更不享受工作的人。 - And I didn’t hear a lot.
我沒有聽到太多回覆。 - So, I’m curious what’s going on there.
所以我很好奇那邊發生了什麼事。 - » Yeah, I I’m seeing this a little bit with uh at Anthropic.
» 是的，我在 Anthropic 也有看到一點這樣的情況。 - I think everyone is fairly technical.
我覺得每個人都相當技術導向。 - This is something that we screen for, you know, when when people join.
這是我們在人們加入時會篩選的。 - We have there there’s a lot of technical interviews that people go go through even for non-technical functions.
即使是非技術職位，人們也要經歷很多技術面試。 - Uh and you know our designers largely code.
而且我們的設計師大多都會寫程式。 - So I think for them this is something that they have enjoyed from what I’ve seen because now instead of bugging engineers they can just like go in and code.
所以我覺得從我觀察到的來看，他們很享受這件事，因為現在他們不用去煩工程師，可以直接自己去寫程式。 - And even some designers that didn’t code before have just started to do it and for them it’s great cuz they can unblock themselves.
甚至一些之前不寫程式的設計師也開始這樣做了，對他們來說這很棒，因為他們可以自己解除阻塞。 - But I’d be really interested just to hear more people’s experiences cuz I I I bet it’s not uniform like that.
但我真的很想聽更多人的經驗，因為我打賭不會是那麼一致的。 - » Yeah.
- So maybe if you’re listening to this, leave a comment if you’re finding your jobs less fun and enjoying your job less cuz what you’re saying and what I’m hearing from most people, 70% of PMs and engineers are loving their job more.
所以如果你在聽這個節目，如果你覺得工作變得更無趣、更不享受的話，請留言，因為根據你說的和我從大多數人那裡聽到的，70% 的 PM 和工程師更喜歡他們的工作了。 - That’s like if you’re not in that bucket, you could something’s going on.
如果你不在那個群體裡，可能有些事情正在發生。 - » Yeah.
- Yeah.
- We do see that people use also different tools.
我們確實看到人們也使用不同的工具。
AI 工具與工作流程整合
- So for example, our designers, they use uh the claude desktop app a lot more to to do their coding.
例如，我們的設計師，他們更多地使用 Claude 桌面應用程式來寫程式。 - So you just download the desktop app.
你只要下載桌面應用程式。 - There’s a code tab.
有一個 code 分頁。 - Uh it’s right next to Cowork and it’s actually the same exact Claude Code.
它就在 Cowork 旁邊，而且實際上就是完全相同的 Claude Code。 - So it’s like the same agent and everything.
同樣的 agent，一切都一樣。 - We’ve had this for, you know, for many, many months.
我們已經有這個功能好幾個月了。 - Uh and so you can use this to code in a way that you don’t have to open a bunch of terminals, but you still get the power of Claude Code.
所以你可以用這個方式寫程式，不需要開一堆終端機，但仍然有 Claude Code 的能力。 - And the biggest thing is you can just run as many, you know, Claude sessions in parallel as you want.
最大的好處是你可以同時平行執行任意多個 Claude 工作階段。 - We, you know, we call this multi-Claudeing.
我們稱之為 multi-Claudeing。 - » So this is a it’s it’s a little more native, I think, for folks that are not engineers.
» 所以這對不是工程師的人來說，我覺得更加原生一些。 - And really, this is back to bringing the product to where the people are.
而這其實就是把產品帶到人們所在的地方。 - You don’t want to make people use a different workflow.
你不想讓人們使用不同的工作流程。 - You don’t want to make them go out of their way to learn a new thing.
你不想讓他們特地去學一個新東西。 - It’s whatever people are doing, if you can make that a little bit easier, then that’s just going to be a much better product that people enjoy more.
不管人們在做什麼，如果你能讓那件事稍微容易一點，那就會是一個人們更喜歡的好產品。 - And this is just this principle of latent demand, which I I think is just the the single most important principle in product.
而這就是潛在需求（latent demand）的原則，我覺得這是產品中最重要的一個原則。
產品開發中的潛在需求
- » Can you talk about that actually because I was going to go there.
» 你能談談這個嗎？因為我正要聊這個。 - Explain what this principle is and and and just what happens when you unlock this latent demand.
解釋一下這個原則是什麼，以及當你釋放這個潛在需求時會發生什麼。 - Latent demand is this idea that if you build a product in a way that can be hacked or can be kind of mi misused by people in a way it wasn’t really designed for to do kind of something that they want to do then this helps you as the product builder learn where to take the product next.
潛在需求（latent demand）的概念是，如果你建構一個產品的方式允許它被 hack 或被人們以非設計用途的方式「誤用」來做他們想做的事，那這就幫助你作為產品建造者了解產品下一步該往哪裡走。 - So an example of this is uh Facebook marketplace.
一個例子是 Facebook Marketplace。 - So the the manager for the team Fiona she she was actually the founding manager for uh the marketplace team and she talks about this a lot.
團隊的經理 Fiona，她實際上是 Marketplace 團隊的創始經理，她經常談論這個。 - Facebook Marketplace.
Facebook Marketplace。 - It started based on the observation back in uh this must have been like 20 2016 or or something like this that 40% of posts in Facebook groups are buying and selling stuff.
它的起源是基於大約 2016 年左右的觀察，Facebook 社團中 40% 的貼文是在買賣東西。 - So this is crazy.
這太瘋狂了。 - It’s like people are abusing the Facebook groups product to buy and sell.
就像是人們在「濫用」Facebook 社團產品來進行買賣。 - And it’s not it’s not abuse in kind of like a security sense.
這不是安全意義上的那種濫用。 - It’s abuse in that no one designed the product for this, but they’re kind of figuring it out because it’s just so useful for this.
而是沒有人為此目的設計這個產品，但人們自己想辦法這樣用，因為它對這件事太有用了。 - And so it was pretty obvious if you build a better product to let people buy and sell, they’re going to like it.
所以很明顯，如果你建構一個更好的產品讓人們買賣，他們會喜歡的。 - And it was just very obvious that marketplace would be a hit from this.
從這裡可以很明顯看出 Marketplace 會成功。 - And so the first thing was buy and sell groups.
所以第一個產品是買賣社團。 - So kind of special purpose groups to let people do that.
就是專門用途的社團讓人們做這件事。 - And the second product was marketplace.
第二個產品就是 Marketplace。 - Uh Facebook dating I think started in a pretty similar place.
Facebook Dating 我覺得也是從很類似的地方開始的。 - And I think that the observation was if you look at people looking if you look at uh profile views so people looking at each other’s profiles on Facebook 60% of profile views were people that are not friends with each other that are opposite gender.
我覺得那個觀察是，如果你看人們瀏覽彼此的個人資料，Facebook 上 60% 的個人資料瀏覽是來自彼此不是朋友且異性的人。 - And so this is this kind of like you know like traditional kind of dating setup but you know people are just like creeping on each other.
所以這就像是傳統的約會場景，但你知道，人們就是在偷偷瀏覽彼此。 - So maybe if you can build a product for this it’s you know it might work.
所以也許如果你能為此建構一個產品，它可能會成功。 - Um and so this idea of latent demand I think is just so powerful.
所以這個潛在需求的概念我覺得非常強大。 - And for example this is also where Cowork came from.
例如，Cowork 也是從這裡來的。 - We saw that for the last 6 months or so a lot of people using Claude Code were not using it to code.
我們看到在過去大約 6 個月裡，很多使用 Claude Code 的人並不是用它來寫程式。 - There was someone on Twitter that was using it to grow tomato plants.
Twitter 上有人用它來種番茄。 - There was someone else using it to analyze their genome.
有人用它來分析他們的基因組。 - Someone was using it to uh recover photos from a corrupted hard drive.
有人用它來從損壞的硬碟中復原照片。 - It was like uh wedding photos.
是婚禮照片。 - Uh there was someone that was using it for uh I think like uh they they were using it to analyze a MRI.
有人用它來分析 MRI。 - So there there’s just all these different use cases that are not technical at all.
所以有各種完全不是技術性的不同使用案例。 - And it was just really obvious like people are jumping through hoops to use a terminal to do this thing.
很明顯人們費盡周折地用終端機來做這些事。 - Maybe we should just build a product for them.
也許我們就應該為他們建構一個產品。 - And we saw this actually pretty early back in maybe May of last year.
我們其實很早就看到了這個現象，大概是去年五月。 - I remember walking into the office and our data scientist Brendan was had a Claude Code on his uh computer.
我記得走進辦公室，我們的資料科學家 Brendan 的電腦上開著 Claude Code。 - He just had a terminal up and I was like I was shocked.
他就開了一個終端機，我非常震驚。 - I was like Brendan what what are you doing?
我說 Brendan 你在做什麼？ - Like you you figured out how to open the terminal which is you know it’s a very engineering product.
你居然知道怎麼打開終端機，那可是一個非常工程向的產品。 - Even a lot of engineers don’t want to use a terminal.
甚至很多工程師都不想用終端機。 - It’s just like a it’s like just like the lowest level way to to do your work.
它就像是最底層的工作方式。 - Um just really really uh kind of in the weeds of the computer.
真的是深入電腦的最底層。 - And so he figured out how to use the terminal.
所以他自己搞懂了怎麼用終端機。 - He downloaded Node.js.
他下載了 Node.js。 - He downloaded Claude Code and he was doing SQL analysis in a terminal and it was crazy.
他下載了 Claude Code，然後在終端機裡做 SQL 分析，太瘋狂了。 - And then the next week all the data scientists were doing the same thing.
然後下一週所有的資料科學家都在做同樣的事。 - So when you see people abusing the product in this way, using it in a way that it wasn’t designed in order to do something that is useful for them, it’s just such a strong indicator that you should just build a product and and people are going to like that.
所以當你看到人們以這種方式「濫用」產品，用非設計用途的方式來做對他們有用的事情，這就是一個非常強烈的指標，你應該建構一個產品，人們會喜歡的。 - It’s something that’s special purpose for that.
就是為那個目的而專門設計的東西。 - I think now there there’s also this kind of interesting second dimension to latent demand.
我覺得現在潛在需求還有一個有趣的第二維度。
現代潛在需求：模型能力
- This is sort of the traditional framing is look at what people are doing, make that a little bit easier, empower them.
這是傳統的框架：看看人們在做什麼，讓那件事稍微容易一點，賦能他們。 - The modern framing that I’ve been seeing in the last 6 months is a little bit different and it’s look at what the model is trying to do and make that a little bit easier.
我在過去 6 個月看到的現代框架有點不同，是看看模型正在嘗試做什麼，然後讓那件事稍微容易一點。 - And so when we first started building Claude Code, I think a lot of the way that people approached designing things with LLMs is they kind of put the model in a box and they were like, here’s this application that I want to build.
所以當我們一開始建構 Claude Code 時，我覺得很多人設計 LLM 產品的方式是把模型放在一個框框裡，然後說，這是我要建構的應用程式。 - Here’s the thing that I wanted to do.
這是我要做的事情。 - model, you’re going to do this one component of it.
模型，你要負責其中這一個部分。 - Here’s the way that you’re going to interact with these tools and APIs and whatever.
這是你要跟這些工具和 API 互動的方式。 - And for Claude Code, we inverted that.
而對於 Claude Code，我們反轉了這個做法。 - We said the product is the model.
我們說產品就是模型。 - We want to expose it.
我們要把它暴露出來。 - We want to put the minimal scaffolding around it.
我們要在它周圍放最少的腳手架。 - Give it the minimal set of tools.
給它最少的工具集。 - So, it can do the things.
這樣它就能做事情。 - It can decide which tools to run.
它可以決定執行哪些工具。 - It can decide in what order to run them in and so on.
它可以決定以什麼順序執行它們，等等。 - And I I think a lot of this was just based on kind of latent demand of what the model wanted to do.
我覺得很多這些做法都是基於模型想要做什麼的潛在需求。 - And so, in research, we call this being on distribution.
在研究中，我們稱之為「在分佈上」（on distribution）。 - Uh you want to see like what the model is trying to do.
你要看模型正在嘗試做什麼。 - In product terms, latent demand is just the same exact concept but applied to a model.
用產品的術語來說，潛在需求就是同樣的概念，只是應用在模型上。 - » You talked about Cowork something that I saw you talk about when you launched that initially is you your team built that in 10 days.
» 你提到了 Cowork，我記得看到你在剛推出時說你的團隊在 10 天內建構了它。
Cowork 的開發與學習
- » That’s insane.
» 這太瘋狂了。 - Uh I it came out I think it was like you know used by millions of people pretty quickly something like that being built in 10 days.
它推出後我覺得很快就有數百萬人在使用，這樣的東西是在 10 天內建構出來的。 - Uh anything there any stories there other than just it was just you know we use Claude Code to build it and that’s it.
有什麼故事嗎？除了我們用 Claude Code 來建構它就這樣之外。 - » Yeah it’s funny.
» 是的，很有趣。 - Uh Claude Code like I said when we released it was not immediately a hit.
Claude Code 就像我說的，發布時並沒有立刻大受歡迎。 - it became a hit over time and there was a few inflection points.
它是隨著時間推移才成功的，而且有幾個轉折點。 - So one was you know like Opus 4 uh it just really really inflected and then in November it inflected and it just keeps inflecting.
其中一個是 Opus 4，它真的讓成長大幅轉折，然後在 11 月又轉折了，而且一直在持續加速。 - The growth just keeps getting steeper and steeper and steeper every day.
成長曲線每天都變得越來越陡峭。 - But you know for the first few months it wasn’t a hit.
但在頭幾個月它並不成功。 - Uh people used it but a lot of people couldn’t figure out how to use it.
人們用了它，但很多人不知道怎麼用。 - They didn’t know what it was for.
他們不知道它是用來做什麼的。 - The model still like wasn’t very good.
模型當時還不是很好。 - Cowork when we released it was just immediately a hit much more so than Claude Code was early on.
Cowork 發布時立刻就大受歡迎，比 Claude Code 早期成功得多。 - I think a lot of the credit honestly just goes to like Felix and and Sam and the and Jenny and the the team that built this.
我覺得很多功勞真的歸於 Felix、Sam、Jenny 和建構這個產品的團隊。 - It’s just an incredibly strong team.
這是一個非常強大的團隊。 - And again, the the place co came from is just this latent demand.
再次強調，Cowork 的起源就是這個潛在需求。 - Like we saw people using Claude Code for these non-technical things and we’re trying to figure out what do we do?
我們看到人們用 Claude Code 做這些非技術性的事情，然後我們試圖搞清楚我們該怎麼做。 - And so for a few months the team was exploring they were trying all sorts of different options and in the end someone was just like okay what what if we just take Claude Code and put it in the desktop app and that’s essentially the thing that worked.
所以有好幾個月團隊一直在探索，他們嘗試了各種不同的選項，最後有人就說好吧，如果我們把 Claude Code 放進桌面應用程式呢，而那基本上就是成功的做法。 - And so over 10 days they just completely use Claude Code to build it.
所以在 10 天內他們完全用 Claude Code 來建構它。 - Uh and you know Cowork is actually there’s this very sophisticated security system that’s that’s built in and essentially these guard rails to make sure that the model kind of does the right thing.
Cowork 實際上有一個非常精密的安全系統內建其中，本質上是這些護欄來確保模型做正確的事情。 - It doesn’t go off the rails.
不會偏離軌道。 - So for example we ship an entire virtual machine with it.
例如我們隨附了一整個虛擬機器。 - And Claude Code just wrote all of this code.
Claude Code 就寫了所有這些程式碼。 - So we just had to think about all right how do we make this a little bit safer a little more self-guided for uh people that are not engineers.
所以我們只需要思考好吧，我們要怎麼讓這個東西對不是工程師的人更安全一點、更能自我引導一點。 - It was fully implemented with Claude Code.
它完全是用 Claude Code 實作的。 - took about 10 days.
花了大約 10 天。 - We launched it early.
我們提早推出了它。 - You know, it was still pretty rough and it’s still pretty rough around the edges.
它還是相當粗糙的，邊邊角角還是相當粗糙。 - But this is kind of the way that we learn um both on the product side and on the safety side is we have to release things a little bit earlier than we think so that we can get the feedback so that we can talk to users.
但這就是我們學習的方式，在產品方面和安全方面都是，我們必須比我們覺得準備好的時間再早一點發布東西，這樣我們才能獲得回饋，才能跟使用者交談。 - We can understand what people want and and that will shape where the product goes in the future.
我們可以了解人們想要什麼，而那會塑造產品未來的走向。 - » Yeah, I think that point is so interesting and and it’s so unique.
» 是的，我覺得那個觀點非常有趣，而且非常獨特。
AI 安全與提早釋出策略
- There’s always been this idea release early, learn from users, get feedback, iterate.
一直以來都有這個概念：提早釋出，從使用者學習，獲得回饋，迭代。 - The fact that it’s hard to even know what the AI is capable of and how people will try to use it is like is a unique reason to start releasing things early that’ll help you as you exactly describe this idea of what is the latent demand in this thing that we didn’t really know.
事實上很難知道 AI 能做什麼以及人們會嘗試怎麼使用它，這是一個獨特的理由來提早釋出，這會幫助你，就像你描述的這個概念：這個東西裡面我們不知道的潛在需求是什麼。 - Let’s put it out there and see what people do with it.
讓我們把它放出去，看看人們會用它做什麼。 - » Yeah.
- And in Anthropic as a safety lab, the other dimension of that is safety because um you know like when you think about model safety, there’s a bunch of different ways to study it.
而在 Anthropic 作為一個安全實驗室，另一個維度是安全性，因為當你思考模型安全時，有很多不同的方式來研究它。 - Sort of the lowest level is alignment and mechanistic interpretability.
最底層是對齊（alignment）和機制可解釋性（mechanistic interpretability）。 - So this is when we train the model, we want to make sure that it’s safe.
當我們訓練模型時，我們要確保它是安全的。 - We at this point have like pretty sophisticated technology to understand what’s happening in the neurons to trace it.
我們現在已經有相當精密的技術來理解神經元中正在發生什麼，來追蹤它。 - And so for example like if there’s a neuron related to deception we can start we’re starting to get to the point where we can monitor it and understand that it’s activating.
例如，如果有一個與欺騙相關的神經元，我們開始能夠監控它並理解它正在被啟動。 - Um and so this is just this is alignment this is mechanistic interpretability.
所以這就是對齊，這就是機制可解釋性。 - It’s like the lowest layer.
這是最底層。 - The second layer is evolves and this is essentially a laboratory setting.
第二層是評估（evals），這本質上是一個實驗室環境。 - The model is in a petri dish and you study it and you put in a synthetic situation and just say okay like model what do you do and are you doing the right thing?
模型在一個培養皿中，你研究它，你把它放在一個合成的情境中，然後說好吧模型你會怎麼做，你做的是對的嗎？ - Is it aligned?
它是對齊的嗎？ - Is it safe?
它是安全的嗎？ - And then the third layer is seeing how the model behaves in the wild.
然後第三層是看模型在真實世界中的表現。 - And as the model gets more sophisticated, this this becomes so important because it might look very good on these first two layers but not great on the third one.
隨著模型變得更複雜，這變得非常重要，因為它可能在前兩層看起來很好，但在第三層表現不好。 - We released Claude Code really early because we wanted to study safety and we actually used it within Anthropic for I think four or 5 months or something before we released it because we weren’t really sure like this is the first agent that you know the first big agent that I think folks had released at that point.
我們很早就釋出了 Claude Code，因為我們想研究安全性，而且我們實際上在 Anthropic 內部使用了大約四到五個月才對外釋出，因為我們不太確定，這是第一個 agent，我認為是當時大家釋出的第一個大型 agent。 - um it was definitely the first uh you know coding agent that became broadly used and so we weren’t sure if it was safe and so we actually had to study it internally for a long time before we felt good about that and even since you know there’s a lot that we’ve learned about alignment there’s a lot that we’ve learned about safety that we’ve been able to put back into the model back into the product and for Cowork it’s pretty similar uh the model’s in this new setting it’s you know doing
它絕對是第一個被廣泛使用的程式設計 agent，所以我們不確定它是否安全，我們實際上必須在內部研究很長時間才覺得可以，而且即使到現在，我們學到了很多關於對齊和安全的知識，我們已經能夠把這些回饋到模型和產品中，而 Cowork 也很類似，模型在這個新的環境中，它在做 - these tasks that are not engineering tasks it’s an agent that’s acting on your behalf it looks good on alignment it looks good on evals we try to internally it looks good we it with a few customers, it looks good.
這些不是工程任務的事情，它是一個代表你行動的 agent，它在對齊上看起來不錯，在評估上看起來不錯，我們在內部嘗試看起來不錯，我們跟幾個客戶測試看起來不錯。 - Now, we have to make sure it’s safe in the real world.
現在，我們必須確保它在真實世界中是安全的。 - And so, that’s why we release a little early.
所以這就是為什麼我們提早釋出。 - That’s why we call it a research preview.
這就是為什麼我們稱它為研究預覽（research preview）。 - Um, but yeah, it’s just it’s constantly improving.
但是，它一直在持續改善。
AI 安全的三個層次
- Um, and this is really the only way to to make sure that over the long term the model is aligned and it’s doing the right things.
而這真的是確保模型長期對齊且做正確事情的唯一方式。 - It’s such a wild space that you work in where there’s this insane competition and pace.
你工作的領域是一個如此瘋狂的空間，有著瘋狂的競爭和節奏。 - At the same time, there’s this fear that if you get some if the the you know the god can escape and cause damage and just finding that balance must be so challenging.
同時，有一種恐懼是如果那個，你知道，那個神可能逃脫並造成傷害，找到那個平衡點一定非常具有挑戰性。 - What I’m hearing is there’s kind of these three layers and I know there’s like this could be a whole podcast conversation is how you all think about the safety piece but just what I’m hearing is there’s these three layers you work with.
我聽到的是有這三個層次，我知道關於你們如何思考安全這個部分可以是一整集 podcast 的對話，但我聽到的就是你們有這三個層次在運作。 - Uh there’s kind of like observing the model thinking and operating.
有一種是觀察模型的思考和運作。 - There’s tests eval that tell you this is doing bad things and then releasing it early.
有測試評估告訴你它正在做壞事，然後就是提早釋出。
機制可解釋性
- I haven’t actually heard a ton about that first piece.
關於第一個部分我其實沒有聽過太多。 - That is so cool.
那太酷了。 - So you guys can there’s an observability tool that can let you peek inside the model’s brain and see how it’s thinking and where it’s heading.
所以你們有一個可觀測性工具，可以讓你窺探模型的大腦內部，看它是怎麼思考的、它要往哪裡去。 - Yeah, you should uh you should at some point have Chris Ola on the podcast because uh he he’s just the industry expert on this.
是的，你應該找個時間請 Chris Olah 上 podcast，因為他就是這個領域的產業專家。 - He he invented this field of uh we call it mechanistic interpretability.
他發明了這個領域，我們稱之為機制可解釋性（mechanistic interpretability）。 - Uh and the the idea is uh you know like at its core like what is your brain?
核心的概念就是，你的大腦是什麼？ - Like what are what is it?
它到底是什麼？ - It’s like it’s a bunch of neurons that are connected.
它就是一堆相連的神經元。 - And so what you can do is like in a human brain or animal brain you can study it at this kind of mechanistic level to understand what the neurons are doing.
所以你能做的就像是在人腦或動物腦中，你可以在這種機制層面來研究它，理解神經元在做什麼。 - It turns out surprisingly a lot of this does translate to models also.
令人驚訝的是，很多這些研究確實也適用於模型。 - So model neurons are not the same as animal neurons but they behave similarly in a lot of ways.
模型的神經元跟動物的神經元不一樣，但在很多方面行為相似。 - And so we’ve been able to learn just a ton about the way these neurons work, about, you know, this layer or this neuron maps to this concept, how particular concepts are encoded, how the model does planning, how it how it thinks ahead, you know, like a long time ago, we weren’t sure if the model is just predicting the next token or is doing something a little bit deeper.
所以我們已經能夠學到大量關於這些神經元如何運作的知識，關於這一層或這個神經元對應到這個概念，特定概念是如何被編碼的，模型是如何做規劃的，它是如何提前思考的，就像很久以前我們不確定模型是只是在預測下一個 token，還是在做更深層的事情。 - Now, I think there’s actually quite strong evidence that it is doing something a little bit deeper.
現在，我覺得有相當強的證據顯示它確實在做更深層的事情。 - And then the structures that were to do this are pretty sophisticated now where as the models get bigger, it’s not just like a single neuron that corresponds to a concept.
而用來做這件事的結構現在相當精密，隨著模型變得更大，不再只是單一神經元對應一個概念。 - A single neuron might correspond to a dozen concepts.
一個神經元可能對應到十幾個概念。 - And if it’s activated together with other neurons, this is called superposition.
如果它與其他神經元一起被啟動，這叫做疊加（superposition）。 - And uh together it represents this more sophisticated concept.
它們一起代表了一個更複雜的概念。 - And it’s just something we’re learning about all the time, you know, and philAnthropic as as we think about the way this space evolves, doing this in a way that is safe and good for the world is just this is the reason that we exist and this is the reason that everyone is at Anthropic.
這是我們一直在學習的東西，而 Anthropic 在思考這個領域如何演進時，以安全且對世界有益的方式來做這件事，這就是我們存在的原因，這就是每個人在 Anthropic 的原因。 - Uh, everyone that is here, this is the reason why they’re here.
每一個在這裡的人，這就是他們在這裡的原因。 - So, a lot of this work we actually open source.
所以很多這些工作我們實際上是開源的。 - Uh, we publish it a lot.
我們大量發表這些研究。 - Um and you know we publish very freely to talk about this just so we can inspire other labs that are working on similar things to do it in a way that’s safe and this is something that we’ve been doing for Claude Code also we call this the race to the top uh internally and so for Claude Code for example we released an open source sandbox and this is a sandbox they can run the the agent in and it just makes sure that there’s certain boundaries and it can’t access like everything on your system.
我們非常自由地發表來談論這些，這樣我們可以啟發其他在做類似事情的實驗室以安全的方式來做，這也是我們為 Claude Code 做的事情，我們內部稱之為「向上競賽」（race to the top），例如對於 Claude Code 我們釋出了一個開源的沙盒（sandbox），這是一個可以在其中執行 agent 的沙盒，它確保有某些邊界，agent 不能存取你系統上的所有東西。 - Uh, and we made that open source and it actually works with any agent, not just Claude Code because we wanted to make it really easy for others to do the same thing.
我們把它開源了，而且它實際上可以跟任何 agent 一起運作，不只是 Claude Code，因為我們想讓其他人也能很容易地做同樣的事。 - Um, so this is just the same principle of race to the top.
所以這就是同樣的「向上競賽」原則。 - Um, we we want to make sure this thing goes well and this is just the this is the lever that we have.
我們想要確保這件事情順利進行，而這就是我們擁有的槓桿。 - » Incredible.
» 太了不起了。 - Okay, I definitely want to spend more time on that.
好的，我絕對想花更多時間在那上面。 - I I will follow up with this suggestion.
我會跟進這個建議的。
共同經歷與出身
- Something else that I’ve been noticing in the in the field across engineers, product managers, others that work with agents is there’s this kind of anxiety people feel when their agents aren’t working.
我注意到在這個領域中，工程師、產品經理和其他使用 agent 的人，當他們的 agent 沒有在運作時，會感到一種焦慮。 - There’s a sense that like, oh man, Nza has a question, I need to answer or it’s like blocked on something or it’s or I just like I I’m like there’s all this productivity I’m losing.
有一種感覺就是，天啊，它有個問題，我需要回答，或是它被什麼東西卡住了，或是我覺得我正在損失所有這些生產力。 - I can’t like I need to wake up and get it going again.
我必須醒來讓它重新開始運作。 - Is that something you feel?
這是你會感受到的嗎？ - Is that something your team feels?
這是你的團隊會感受到的嗎？ - Do you feel like this is a a problem we need to track and think about?
你覺得這是我們需要追蹤和思考的問題嗎？ - I always have a bunch of agents running.
我總是有一堆 agent 在執行。 - So like at the moment I have like five agents running and at any moment like you know like I I wake up and I I stored a bunch of agents.
像是此刻我大概有五個 agent 在執行，隨時都是這樣，我醒來時已經存了一堆 agent。 - Like the first thing I did when I woke up is like oh man I I want I really want to check this thing.
我醒來做的第一件事就是，天啊，我真的很想檢查這個東西。 - So like I opened up my phone Claude iOS app code tab uh you know like agent do do blah blah blah cuz I I wrote some code yesterday and I was like wait did did I do this right?
所以我打開手機上的 Claude iOS app 的 code tab，讓 agent 去做某件事，因為我昨天寫了一些程式，我在想，等等，我做對了嗎？ - I was like kind of double double guessing something and it and it was correct.
我有點在反覆懷疑某件事，結果它是正確的。 - But now it’s just like so easy to do this.
但現在做這件事就是這麼容易。 - So I don’t know, there is this little bit of anxiety.
所以我不知道，確實有一點點焦慮。 - Maybe I personally haven’t really felt it just cuz I have agents running all the time.
也許我個人沒有真的感受到，只是因為我一直都有 agent 在執行。 - Um, and I’m also just like not locked into a terminal anymore.
而且我也不再被鎖定在終端機裡了。 - Maybe a third of my code now is in the terminal, but also a third is uh using the desktop app and then a third is the iOS app, which is just so surprising cuz I did not think that this would be the way that I code uh in even in 2026.
現在大概三分之一的程式是在終端機裡寫的，另外三分之一是用桌面應用程式，然後三分之一是 iOS app，這真的很令人驚訝，因為我沒想到即使在 2026 年，這會是我寫程式的方式。 - I love that you describe it as coding still, which is just talking to the to Claude Code to code for you essentially.
我喜歡你仍然把它描述為寫程式，但本質上就是跟 Claude Code 對話，讓它替你寫程式。 - And it’s interesting that this is now like this is now coding.
有趣的是，這現在就是寫程式了。 - Coding now is describing what you want, not writing actual code.
寫程式現在是描述你想要什麼，而不是寫實際的程式碼。 - » I I I kind of wonder if uh the people that used to code using punch cards or whatever, if you show them software, what they would have said.
» 我有點好奇，那些以前用打孔卡寫程式的人，如果你給他們看軟體，他們會怎麼說。 - Isn’t that crazy?
這不是很瘋狂嗎？ - And I I remember reading something this was maybe like very early versions of like ACM uh like like magazine or something where people were saying no it’s not the same thing like this isn’t this isn’t really coding uh and you know like they called it programming I think coding is kind of a new word » but I kind of think about this like in the back in the you know my family is from the Soviet Union I you know I I was born in Ukraine um and my grandpa was actually one of the first programmers
我記得讀過一些東西，大概是很早期版本的 ACM 雜誌之類的，人們說不，這不一樣，這不是真正的程式設計，他們當時叫它 programming，我覺得 coding 是一個比較新的詞。但我會想到這個，我的家族來自蘇聯，我出生在烏克蘭，我爺爺其實是蘇聯最早的程式設計師之一。 - in the Soviet Union and he programmed using punch cards And uh you know like he he told my mom uh growing up told these stories of like or she she told these stories that when she was growing up he would bring these punch cards home and there was these like big stacks of punch cards and for her she would like draw all over them with crayons and that was like her childhood memory but for him that was like his experience of programming and he actually never saw the software transition but at
他用打孔卡寫程式。他跟我媽媽講過這些故事，或者說她講過這些故事，當她長大的時候，他會把打孔卡帶回家，有一大疊打孔卡，對她來說，她會用蠟筆在上面到處畫，那是她的童年記憶，但對他來說那是他的程式設計經驗。他其實從來沒有看到軟體的轉型，但在 - some point it did transition to software and I think there’s probably this older generation of programmers that just didn’t take software very seriously and they would have been like well you know it’s not really coding.
某個時候它確實轉型到軟體了，我覺得可能有一整代較老的程式設計師，他們就是不太認真看待軟體，他們會說，這不算真正的寫程式。 - But I I think this is a field that just has always been changing in this way.
但我覺得這個領域一直都是這樣在改變的。 - » Uh I don’t think you know this, but I was born in Ukraine also.
» 我不確定你知不知道，但我也是在烏克蘭出生的。 - » Oh, I don’t know.
» 哦，我不知道。 - » Yeah.
- Which time?
什麼時候的事？ - » I’m I’m from Odessa.
» 我來自敖德薩（Odessa）。 - » Oh, me too.
» 哦，我也是。 - » What?
- » Yeah, that’s crazy.
» 對，太瘋狂了。 - » Wow.
- Incredible.
太不可思議了。 - What a moment.
多麼巧的時刻。
個人旅程與反思
- Uh maybe related in some small way.
也許在某種小小的方面有關聯。 - » Uh what year did your home did you leave and your family leave?
» 你們家是哪一年離開的？ - » Uh we came in 95.
» 我們是 95 年來的。 - » Okay.
- We left in ’ 88.
我們是 88 年離開的。 - a little earlier.
早一點。 - » Oh, yeah.
» 哦，對。 - » What a different life that would have been to not to not leave, huh?
» 如果沒有離開，那會是多麼不同的人生，對吧？ - » Yeah.
- I just I feel I feel so lucky every day that uh get get to grow up here.
我只是每天都覺得自己好幸運，能夠在這裡長大。 - » Yeah.
- My family anytime there’s like a toaster or a meal, they’re just like to America.
我的家人每次只要有一台烤麵包機或一頓飯，他們就會舉杯說「敬美國」。 - It’s like, okay, enough about that.
就像是，好了，夠了。 - But you get it, you know, once you start really thinking about what life could have been.
但你懂的，一旦你開始認真想像人生可能會是什麼樣子。 - » Yeah.
- » Yeah.
- Exactly.
沒錯。 - » Yeah.
- We do we do the same toast, but it’s still vodka.
我們也舉同樣的杯，但還是喝伏特加。 - » It’s still vodka.
» 還是伏特加。 - Absolutely.
絕對是。
賦能模型，避免框限
- Oh, man.
天啊。 - Okay.
好的。 - Let me ask you a couple more things here.
讓我再問你幾件事。 - You shared some really cool tips for how to get the most out of AI, how to build on AI, how to build great products on AI.
你分享了一些很棒的技巧，關於如何從 AI 中獲得最大價值、如何在 AI 上建構、如何建構出色的 AI 產品。 - One tip you shared is give your team as many tokens as they want.
你分享的一個技巧是給你的團隊他們想要的 token 數量。 - Just like let them experiment.
就讓他們去實驗。 - You also shared just advice generally of just build towards the model where the model is going, not to where it is today.
你也分享了一般性的建議，就是朝著模型未來發展的方向去建構，而不是針對今天的模型。 - What other advice do you have for folks that are trying to build AI products?
對於那些試圖建構 AI 產品的人，你還有什麼其他建議？ - » I’d probably share a few more things.
» 我可能會再分享幾件事。 - So, one is don’t try to box the model in.
第一個是不要試圖把模型框住。 - Um I I think a lot of people’s instinct when they build on the model is they try to make it behave a very particular way.
我覺得很多人在模型上建構時的本能反應，是試圖讓它以非常特定的方式運作。 - They’re like this is a component of a bigger system.
他們把它當作一個更大系統的元件。 - I I think some examples of this are people layering like very strict workflows on the model for example you know to say like you must do step one then step two then step three and you have this like very fancy orchestrator doing this.
我覺得一些例子是人們在模型上疊加非常嚴格的工作流程，比如說你必須做第一步然後第二步然後第三步，然後你有一個非常花俏的編排器（orchestrator）在做這件事。 - But actually almost always you get better results if you just give the model tools you give it a goal and you let it figure it out.
但實際上幾乎總是，如果你只是給模型工具、給它一個目標、讓它自己搞定，你會得到更好的結果。 - I think a year ago you actually needed a lot of the scaffolding but nowadays you don’t really need it.
我覺得一年前你確實需要很多鷹架（scaffolding），但現在你真的不需要了。 - So, you know, I I don’t know what to call this principle, but it’s like, you know, like ask not what the model can do for you.
所以，我不知道該怎麼稱呼這個原則，但它有點像是，不要問模型能為你做什麼。 - Maybe maybe it’s something like this.
也許是類似這樣的東西。 - Just think about how do you give the model the tools to do things.
只要想想你怎麼給模型工具去做事情。 - Don’t try to overcurate it.
不要試圖過度策展它。 - Don’t try to put it into a box.
不要試圖把它放進盒子裡。 - Don’t try to give it a bunch of context up front.
不要試圖事先給它一堆上下文。 - Give it a tool so that it can get the context it needs.
給它一個工具，讓它可以取得它需要的上下文。 - You’re just going to get better results.
你就是會得到更好的結果。
擁抱苦澀的教訓
- I think a second one is um maybe actually like a a more even more general version of this principle is just the bitter lesson.
我覺得第二個，也許其實是這個原則的一個更一般性的版本，就是苦澀的教訓（the bitter lesson）。 - Uh and actually for the Claude Code team we have a you know hopefully hopefully um listeners have have read this but Rich Sutton had this blog post maybe 10 years ago called the bitter lesson.
對於 Claude Code 團隊，我們有，希望聽眾們讀過這個，Rich Sutton 大約 10 年前寫了一篇部落格文章叫做「苦澀的教訓」（The Bitter Lesson）。 - Uh and it’s actually a really simple idea.
這其實是一個非常簡單的想法。 - His idea was that the more general model will always outperform the more specific model and I think for him he was talking about like self-driving cars and other domains like this but actually there’s just so many corlaries to the bitter lesson.
他的想法是更通用的模型永遠會優於更特定的模型，我覺得他當時談的是自動駕駛汽車和其他類似的領域，但實際上苦澀的教訓有非常多的推論。 - And for me, the biggest one is just always bet on the more general model and you know over the long term like don’t don’t try to use tiny models for stuff.
對我來說，最大的一個就是永遠押注在更通用的模型上，從長遠來看，不要試圖用小模型來做事情。 - Don’t try to like fine-tune.
不要試圖去微調（fine-tune）。 - Don’t try to do any of this stuff.
不要試圖做任何這些事情。 - There’s like some applications you know there’s some reasons to do this but almost always try to bet on the more general model if you can if you have that flexibility.
有一些應用場景，有一些理由這樣做，但幾乎總是應該盡可能押注在更通用的模型上，如果你有那個彈性的話。 - Um and so these workflows are essentially a way that uh you know it’s it’s not it’s not a general model.
所以這些工作流程本質上是一種方式，它不是一個通用模型。 - It’s putting the scaffolding around it.
它是在模型周圍搭鷹架。 - And in general what we see is maybe scaffolding can improve performance maybe 10 20% something like this but often these gains just get wiped out with the next model.
一般來說我們看到的是，也許鷹架可以提升效能大概 10%、20% 之類的，但通常這些增益會被下一個模型完全抵消掉。 - Uh so it’s almost better to just wait for the next one.
所以幾乎不如直接等下一個模型。
為未來模型而建
- And I think maybe this is a final principle and something that Claude Code I think got right in hindsight.
我覺得也許這是最後一個原則，也是 Claude Code 回頭看來做對的事情。 - From the very beginning, we bet on building for the model six months from now, not for the model of today.
從一開始，我們就押注在為六個月後的模型而建構，而不是為今天的模型。 - And for the very early versions of the product, it just wrote so little of my code cuz I I didn’t trust it cuz, you know, it was like sonnet 3.5, then it was like 3.6 or forget 3 3.5 new, whatever whatever whatever name we gave it.
在產品的最早期版本中，它幾乎沒有幫我寫多少程式，因為我不信任它，當時是 Sonnet 3.5，然後是 3.6 還是 3.5 new，不管我們給它取了什麼名字。 - Um, these models just weren’t very good at coding yet.
這些模型在寫程式方面還不是很好。 - Um, they were they were getting there, but it was still pretty early.
它們正在進步，但還是很早期。 - So back then the model did uh you used git for me it automated some things but it it really wasn’t doing a huge amount of my coding and so the bet with Claude Code was at some point the model gets good enough that it can just write a lot of the code and this is a thing that we first started seeing with opus 4 and sonnet 4 and opus 4 was our first kind of ASL3 class model uh that we released back in May and we just saw this inflection because everyone started to use Claude Code for the first time
所以那時候模型幫我用 git，它自動化了一些事情，但它真的沒有幫我寫大量的程式。所以 Claude Code 的賭注是，在某個時候模型會變得夠好，可以直接寫大量的程式碼。這是我們最先在 Opus 4 和 Sonnet 4 上看到的，Opus 4 是我們第一個 ASL3 等級的模型，我們在五月發布的，我們就看到了這個轉折點，因為每個人第一次開始使用 Claude Code。 - and that was kind of when our growth really went exponential and like I said it’s kind of it stayed there.
那就是我們的成長真正變成指數級的時候，而且如我所說，它一直維持在那裡。 - So I think this is some this is advice that I actually give to to a lot of folks especially people building startups.
所以我覺得這是我實際上會給很多人的建議，特別是建構新創公司的人。 - It’s going to be uncomfortable cuz your product market fit won’t be very good for the first 6 months but if you build for the model 6 months out when that model comes out you’re just going to hit the ground running and the product is going to click and and start to work.
這會很不舒服，因為你的產品市場適配度（product market fit）在頭六個月不會很好，但如果你為六個月後的模型而建構，當那個模型出來時，你就能馬上開始奔跑，產品會到位並開始運作。 - And when you say build for the model 6 months out what is what is it that you think people can assume will happen?
當你說為六個月後的模型而建構，你覺得人們可以假設什麼事情會發生？ - Is it just generally it will get better at things?
只是一般性地它會在各方面變得更好嗎？ - Is it just like okay, it’s like almost good enough and that’s a sign that it’ll probably get better at that thing.
還是說，它現在幾乎夠好了，那就是一個信號表示它可能會在那件事上變得更好。 - Is there any advice there?
有什麼建議嗎？ - » I think that’s a good way to do it.
» 我覺得那是一個好的做法。 - Like, you know, obviously within an AI lab, we get to see the specific ways that it gets better.
顯然在一個 AI 實驗室裡，我們可以看到它具體在哪些方面變得更好。 - » So, it’s a it’s a little unfair, but we we also we try to talk about this.
» 所以這有一點不公平，但我們也試著談論這些。 - So, you know, like one of the ways that it’s going to get better is it’s going to get better and better at using tools and using computers.
它會變得更好的方式之一是，它在使用工具和使用電腦方面會越來越好。 - This is a bet that I would make.
這是我會下的一個賭注。 - Uh, another one is it’s going to get better and better for long, for running for long periods of time.
另一個是它在長時間執行方面會越來越好。 - And this is a place, you know, like there’s all sorts of studies about this, but if you just trace the trajectory or, you know, maybe even like for my own experience when I used Sonnet 3.5 back, you know, a year ago, it could run for maybe 15 or 30 seconds before, before it started going off the rails and you just really had to hold its hand through any kind of complicated task.
這方面有各種研究，但如果你追蹤這個軌跡，或者甚至從我自己的經驗來看，大約一年前我使用 Sonnet 3.5 時，它大概只能執行 15 到 30 秒就開始偏離軌道，你真的必須在任何複雜的任務中牽著它的手。 - But nowadays with Opus 4.6 fix, you know, on average, it’ll run maybe 10, 30, 20, 30 minutes unattended and I’ll just like start another Claude and have it do something else.
但現在用 Opus 4.6，平均來說它可以無人看管地執行 10 分鐘、20 分鐘、30 分鐘，我就再開另一個 Claude 讓它做別的事情。 - And you know, like I said, I always have a bunch of Claudes running.
如我所說，我總是有一堆 Claude 在執行。 - Uh, and they can also run for hours or even days at a time.
而且它們也可以一次執行好幾個小時甚至好幾天。
未來模型能力
- I think there are some examples where they ran for many weeks.
我覺得有一些例子是它們執行了好幾個星期。 - And so I think over time this is going to become more and more normal where the models are running for a very, very long period of time and you, you don’t have to sit there and babysit them anymore.
所以我覺得隨著時間推移，模型長時間執行會變得越來越正常，你不用再坐在那裡看著它們了。 - » So, we just talked about tips for building AI products.
» 所以我們剛才談了建構 AI 產品的技巧。 - Any tips for someone just using Claude Code, say for the first time, or just someone already using Claude Code that wants to get better?
對於剛開始使用 Claude Code 的人，或是已經在使用但想要用得更好的人，有什麼技巧嗎？ - What are like a couple pro tips that you could share?
你能分享幾個專業技巧嗎？
善用最強大的模型
- » I will give a caveat, which is there’s no one right way to use Claude Code.
» 我要先說一個但書，就是使用 Claude Code 沒有唯一正確的方式。 - So I, I can share some tips, but honestly, this is a dev tool.
所以我可以分享一些技巧，但說實話，這是一個開發工具。 - Developers are all different.
開發者都是不同的。 - Developers have different preferences.
開發者有不同的偏好。 - They have different environments.
他們有不同的環境。 - So there’s just so many ways to use these tools.
所以有非常多種使用這些工具的方式。 - There’s no one right way.
沒有唯一正確的方式。 - Um, you, you sort of have to find your own path.
你必須找到自己的路。 - Luckily, you can ask Claude Code.
幸運的是，你可以問 Claude Code。 - Uh, it’s able to make recommendations.
它能夠給出建議。 - It can edit your settings.
它可以編輯你的設定。 - It kind of knows about itself.
它對自己有一定的了解。 - So, it can help, it can help with that.
所以它可以幫忙。 - A few tips that generally I find pretty useful.
幾個我一般覺得蠻有用的技巧。 - So, number one is just use the most capable model.
第一個就是用最強大的模型。 - Um, currently that’s Opus 4.6.
目前那是 Opus 4.6。 - I have maximum effort enabled always.
我一直開著最大努力模式（maximum effort）。 - The thing that happens is sometimes people try to use a less expensive model like Sonnet or something like this.
有時候人們會試著用比較便宜的模型，像是 Sonnet 之類的。 - But because it’s less intelligent, it actually takes more tokens in the end to do the same task.
但因為它沒那麼聰明，最後實際上要用更多 token 來完成同樣的任務。 - Um, and so it’s actually not obvious that it’s cheaper if you use a less expensive model.
所以用比較便宜的模型其實不見得更便宜。 - Often it’s actually cheaper and less token intensive if you use the most capable model because it can just do the same thing much faster with less correction, less, uh, less handholding and so on.
通常用最強大的模型反而更便宜、消耗更少 token，因為它可以更快地完成同樣的事情，需要更少的修正、更少的手把手指導等等。 - So that’s the first tip is just use the best model.
所以第一個技巧就是用最好的模型。
有效運用 Plan Mode
- The second one is use plan mode.
第二個是使用 plan mode。 - I start almost all of my tasks in plan mode, maybe like 80%.
我幾乎所有的任務都從 plan mode 開始，大概 80%。 - And plan mode is actually really simple.
Plan mode 其實非常簡單。 - All it is is we inject one sentence into the model’s prompt to say, please don’t write any code yet.
它就是我們在模型的 prompt 中注入一句話，說請先不要寫任何程式碼。 - That’s it.
就這樣。（Ernest: 只有我想罵 __ 嗎！） - Like there’s, there’s actually like nothing fancy going on.
其實沒有什麼花俏的東西。 - It’s just the simplest thing.
就是最簡單的事情。 - » Um, and so for people that are in the terminal, it’s just shift tab twice and that gets you into plan mode.
» 所以對於在終端機裡的人，只要按兩次 shift tab 就可以進入 plan mode。 - Uh, for people in the desktop app, there’s a little button.
對於在桌面應用程式裡的人，有一個小按鈕。 - On web, there’s a little button.
在網頁上也有一個小按鈕。 - It’s coming pretty soon to mobile also.
行動版也很快會有。 - Uh, and we just launched it for the SWAC integration too.
我們也剛為 Slack 整合推出了它。 - Uh, so plan mode is the second one.
所以 plan mode 是第二個。 - And uh, essentially the model would just go back and forth with you.
基本上模型會跟你來回討論。 - Once the plan looks good, then you let the model execute.
一旦計劃看起來不錯，你就讓模型去執行。 - I auto accept edits after that because if the plan looks good, it’s just going to oneshot it.
之後我就自動接受編輯，因為如果計劃看起來不錯，它就會一次搞定。
探索不同介面
- It’ll get it right the first time almost every time with Opus 4.6.
用 Opus 4.6 的話，它幾乎每次都能第一次就做對。 - And then maybe the third tip is just play around with different interfaces.
然後也許第三個技巧就是玩玩不同的介面。 - I think a lot of people when they think about Claude Code, they think about a terminal.
我覺得很多人想到 Claude Code 的時候，他們會想到終端機。 - Um, and you know, of course, we support every terminal.
當然，我們支援每一種終端機。 - We support like Mac, Windows, you know, like whatever terminal you might use, it works perfectly.
我們支援 Mac、Windows，不管你用什麼終端機，都能完美運作。 - But we actually support a lot of other form factors too, like, you know, we have like iOS and Android apps.
但我們其實也支援很多其他形式，像是我們有 iOS 和 Android 的 app。 - We have a desktop app.
我們有桌面應用程式。 - There’s, uh, you know, the Slack integration.
有 Slack 整合。 - There’s all sorts of things that we support.
我們支援各種各樣的東西。 - So I would just like play around with these.
所以我會建議就去玩玩這些。 - And again it’s like every engineer is different.
再說一次，每個工程師都不一樣。 - Everyone that’s building is different.
每個在建構東西的人都不一樣。 - Just find the thing that feels right to you and and use that.
找到對你來說感覺對的東西，然後用它。 - You don’t have to use a terminal.
你不一定要用終端機。 - It’s the same Claude agent running everywhere.
到處執行的都是同一個 Claude agent。
競爭態勢與使用者聚焦
- » Amazing.
» 太棒了。 - Okay.
好的。 - Just a couple more questions to round things out.
再問幾個問題來收尾。 - What’s your take on Codeex?
你怎麼看 Codex？ - How do you feel about that product?
你對那個產品有什麼感覺？ - How do you feel about where they’re going?
你覺得他們的方向怎麼樣？ - Just kind of competing in this very competitive space uh in coding agents.
就是在程式設計 agent 這個競爭非常激烈的領域裡競爭。 - Yeah, I actually haven’t really used it, but uh I I think I did use it maybe when it came out.
其實我沒怎麼用過，但我想我可能在它剛出來的時候用過。 - It looked a lot like Claude Code to me, so that was kind of flattering.
它看起來很像 Claude Code，所以這蠻讓人受寵若驚的。 - It’s I think it’s actually good, you know, to have more competition cuz people should get to choose and hopefully it forces all of us to like do a even better job.
我覺得有更多競爭其實是好事，因為人們應該有選擇，而且希望這能迫使我們所有人做得更好。 - Honestly, for our team though, we’re just focused on solving the problems that users have.
但老實說，對我們團隊來說，我們就是專注在解決使用者的問題。 - Um so for us, you know, we don’t spend a lot of time looking at competing products.
所以我們不會花很多時間去看競爭產品。 - We don’t really try the other products.
我們不太會去試其他產品。 - I you know you kind of you want to be aware of them.
你會想要知道它們的存在。 - You want to know they exist but for me I just I love talking to users.
你會想知道它們存在，但對我來說，我就是喜歡跟使用者聊天。 - I love making the product better.
我喜歡把產品做得更好。 - Um I I love just acting on on feedback.
我喜歡根據回饋來行動。 - So it’s really just about building a building a good product.
所以真的就是專注在建構一個好產品。
AGI 之後的計畫與做味噌
- » Maybe a last question.
» 也許最後一個問題。 - So I talked to Ben man co-founder of Anthropic.
我跟 Anthropic 的共同創辦人 Ben Mann 聊過。 - What what to talk to you about.
聊要跟你談什麼。 - He had a bunch of suggestions which I’ve integrated throughout our chat.
他有一堆建議，我已經整合到我們的對話中了。 - One question he had for you is what’s your plan post AGI?
他有一個問題要問你，就是你在 AGI 之後有什麼計劃？ - What do you think you’re going to be doing?
你覺得你會做什麼？ - What’s your life like once we hit AGI?
一旦我們達到 AGI，你的生活會是什麼樣子？ - whatever that means.
不管那代表什麼。 - » So before I joined Anthropic, um I was actually living in rural Japan and it was like a totally different lifestyle.
» 在我加入 Anthropic 之前，我其實住在日本鄉下，那是一種完全不同的生活方式。 - Um I was like the only engineer in the town.
我是鎮上唯一的工程師。 - I was the only English speaker in the town.
我是鎮上唯一說英語的人。 - It was just like a totally different vibe.
那就是一種完全不同的氛圍。 - Like a couple times a week I would like bike to the farmers market.
一個星期有幾次我會騎腳踏車去農夫市集。 - Uh and you know you like bike by like rice patties and stuff.
你會騎過稻田之類的地方。 - It was just like a totally different speed than just complete opposite of San Francisco.
那是一種完全不同的節奏，跟舊金山完全相反。 - One of the things that I really liked is a way that we got to know our neighbors and we kind of built friendships is by trading like pickles.
我很喜歡的一件事是，我們認識鄰居和建立友誼的方式是交換醃漬物。（Ernest: 可惡想要！） - So in that in the town where we lived, it was actually like everyone made like miso.
在我們住的那個小鎮，每個人都會做味噌。 - Everyone made pickles.
每個人都會做醃漬物。 - Uh and so I actually got like decently good at making miso.
所以我做味噌其實做得還不錯。 - Um and you know I made a bunch of batches and um this is something that I still make.
我做了好幾批，而且這是我至今仍在做的事。 - Uh miso is this interesting thing where it teaches you to think on these longtime skills.
味噌是一個很有趣的東西，它教你用長時間尺度去思考。 - That’s just very different than engineering cuz like uh you know like a batch of white miso takes like at least three months to make and a red miso is like you know 2 3 4 years.
這跟工程非常不同，因為一批白味噌至少要三個月才能做好，紅味噌則要兩三四年。 - You just have to be very patient.
你就是得非常有耐心。 - You kind of mix it up and then you just like wet it sit.
你把它混合好，然後就讓它靜置。 - You have to be very very patient.
你必須非常非常有耐心。 - » So I the thing that I love about it is just thinking in these longtime skills.
» 我喜歡它的地方就是用這種長時間尺度去思考。 - Uh, and yeah, I think postGI or if I wasn’t at Anthropic, I’d probably be making miso.
我覺得 AGI 之後，或是如果我不在 Anthropic 的話，我大概會在做味噌。 - » I love this answer.
» 我喜歡這個回答。 - Uh, Ben asked me to ask you about what’s the deal with you and miso and so I love that you answered it.
Ben 叫我問你，你跟味噌是怎麼回事，所以我很高興你回答了。 - Okay, so the future the future might be just going deep into miso, getting really good at get making miso.
好的，所以未來可能就是深入味噌的世界，把味噌做得非常好。 - Uh, amazing.
太棒了。
Anthropic 的開發哲學
- Uh, Boris, this was incredible.
Boris，這真的太精彩了。 - I feel like we’re we’re brothers now from Ukraine.
我覺得我們現在是來自烏克蘭的兄弟了。 - Uh before we get to a very exciting lightning round, is there anything else that you wanted to share?
在我們進入令人興奮的快問快答之前，還有什麼你想分享的嗎？ - Is there anything you want to leave listeners with?
有什麼你想留給聽眾的嗎？ - Anything you want uh you want to double down on?
有什麼你想要加倍強調的嗎？ - » Yeah, I I think I would just like underscore, you know, like for for Anthropic since the beginning, this idea of like starting at coding, then getting to tool use, then getting to computer use has just been the way that we think about things.
» 我想強調的是，對 Anthropic 來說，從一開始，從程式設計開始、然後到工具使用、再到電腦使用，這就是我們思考事情的方式。 - And we this is the way that we know the models are going to develop or, you know, the way that we want to build our models.
這是我們知道模型會發展的方式，也是我們想要建構模型的方式。 - And it’s also the way that we get to learn about safety, study it, and improve it the most.
這也是我們學習安全性、研究它、並最大程度改進它的方式。
AI 採用的早期階段
- So, you know, everything that’s happening right now around, you know, just like Claude Code becoming this huge, you know, multi-billion dollar business and, you know, like now all of my friends use Claude Code and they just text me about it all the time.
現在正在發生的一切，Claude Code 變成這個巨大的、數十億美元的業務，現在我所有的朋友都在用 Claude Code，他們一直傳訊息跟我說。 - Uh, so just like, you know, this thing getting kind of big and in some ways it’s a total surprise because this isn’t kind of the we didn’t know that it would be this product.
這個東西變得這麼大，在某些方面完全是個驚喜，因為我們不知道會是這個產品。 - We didn’t know that it would start in a terminal or anything like this.
我們不知道它會從終端機開始，或任何像這樣的事。 - But in some ways, it’s just totally unsurprising because this has been our belief as a company for for a long time.
但在某些方面，這完全不意外，因為這一直是我們公司長久以來的信念。 - At the same time, it just feels still very early, you know, like most of the world still does not use Claude Code.
同時，感覺還是非常早期，世界上大多數人仍然沒有在用 Claude Code。 - Most of the world still does not use AI.
世界上大多數人仍然沒有在用 AI。 - So, it just feels like this is 1% done and there’s so much more to go.
所以感覺這才完成了 1%，還有很多路要走。
使用者回饋驅動改進
- » Yeah.
- Man, that’s insane to think seeing the numbers that are coming out.
看到這些數字，想想就覺得瘋狂。 - You guys just raised a bazillion dollars.
你們剛募了一大筆錢。 - Uh I think Claude Code alone is making$2 billion dollars in revenue.
我覺得光是 Claude Code 就在產生 20 億美元的營收。 - you think Anthropic, I think the number you guys put out, you’re making 15 billion in revenue.
你想想 Anthropic，我覺得你們公布的數字是 150 億美元的營收。 - It’s uh insane to just think this is how early it still is and just the numbers we’re seeing.
想到這還是多麼早期，再看看這些數字，真的太瘋狂了。 - » Yeah.
- Yeah.
- Yeah.
- It’s crazy.
真的很瘋狂。 - And and I mean like the the way that Claude Code has kept growing is honestly just the users.
Claude Code 持續成長的原因，老實說就是使用者。 - Like we so many people use it.
有太多人在用它了。 - They’re so passionate about it.
他們對它充滿熱情。 - They fall in love with the product and then they tell us about stuff that doesn’t work, stuff that they want.
他們愛上這個產品，然後告訴我們什麼不能用、他們想要什麼。 - And so like the only reason that it keeps improving is because everyone is using it.
所以它持續進步的唯一原因就是因為每個人都在用它。 - Everyone is talking about it.
每個人都在談論它。 - Everyone keeps giving feedback and this is just the single most important thing and you know for me this is the way that I love to spend my day is just talking to users and making it better for them » and making me so » and making me so well the you know the miso is like not super involved it just you just got to wait you just got to wait » well Boris with that we’ve reached our very exciting lightning round I’ve got five questions for you are you ready » let’s do it first question what
每個人都持續給回饋，這是最重要的一件事。對我來說，這是我喜歡度過一天的方式，就是跟使用者聊天，為他們把產品做得更好。» 還有做味噌。» 還有做味噌。味噌其實不需要太多時間投入，你就是得等，你就是得等。» Boris，說到這裡，我們到了令人興奮的快問快答環節。我有五個問題要問你，你準備好了嗎？» 來吧。第一個問題，
推薦書籍
- are two or three books that you find yourself recommending most to other people » I I’m a greeter.
你最常推薦給別人的兩三本書是什麼？» 我是一個大量閱讀的人。 - Uh I would start with the technical book one is it it is functional programming in Scola.
我先從技術書開始，是 Functional Programming in Scala。 - This is the single best technical book I’ve ever read.
這是我讀過最好的一本技術書。 - It’s very weird because you’re probably not going to use Scola and I don’t know how much this matters in the future now but there’s this just elegance to functional programming and thinking in types and this is just the way that I code and the way that I can’t stop thinking about coding.
這很奇怪，因為你可能不會用 Scala，而且我不知道這在未來還有多重要，但函數式程式設計和型別思考就是有一種優雅，這就是我寫程式的方式，也是我無法停止思考程式設計的方式。 - So you know you could think of it as a historical artifact.
你可以把它當成一個歷史文物。 - You could think of it as something that will level you up.
也可以把它當成能讓你升級的東西。 - » I love this neverbeforementioned book.
» 我喜歡這本從來沒有被提到過的書。 - My favorite.
我的最愛。 - » Oh, amazing.
» 太棒了。 - Amazing.
太棒了。 - Uh, okay.
好的。 - Second one is uh Accelerondo by Straws.
第二本是 Charles Stross 的 Accelerando。 - This is probably, you know, like my my big genre is uh is sci-fi.
我最喜歡的類型大概就是科幻小說。 - Uh like probably sci-fi and fiction.
科幻和小說。 - Accelerondo is just this incredible book and it it it’s just so fast-paced.
Accelerando 是一本令人難以置信的書，它的節奏非常快。 - The pace gets faster and faster and faster.
節奏越來越快、越來越快。 - And I just feel like it captures the essence of this moment that we’re in more than any other book that I’ve read.
我覺得它比我讀過的任何其他書都更能捕捉我們所處的這個時刻的本質。 - Just the speed of it.
就是那種速度感。 - And it starts as a liftoff is starting to happen and you know starting to approach the singularity and it ends with like this like collective lobster consciousness orbiting Jupiter.
它從起飛開始發生、開始接近奇點開始，最後以一個繞著木星運行的集體龍蝦意識結束。 - Um and you know this happens over like the span of a few decades or something.
這一切發生在幾十年的時間跨度裡。 - So the the pace is just incredible.
所以那個節奏真的很驚人。 - I I really love it.
我真的很喜歡。 - Maybe I’ll I’ll do one more book.
也許我再說一本書。 - Uh the wandering earth uh wandering earth by Cixin Liu .
劉慈欣的流浪地球。 - So he’s the guy that did uh three body problem.
他就是寫三體的那個人。 - I think a lot of people know him for that.
我想很多人因為那個認識他。 - I actually I think your body problem was awesome, but I actually liked his short stories even more.
我覺得三體很棒，但我其實更喜歡他的短篇小說。 - So, Wandering Earth is one of the short story collections and it just has some really really amazing stories and it it’s also just quite interesting to see uh Chinese sci-fi because it has a very different perspective than Western sci-fi and kind of the way that um at least he as a writer thinks about it.
流浪地球是其中一本短篇小說集，裡面有一些非常精彩的故事，而且看中國科幻小說也很有趣，因為它跟西方科幻有非常不同的視角，至少他作為一個作家思考方式是不同的。 - It’s so interesting how sci-fi has prepared us to think about where things are going.
科幻小說幫助我們思考事情的走向，這真的很有趣。 - Just like it creates these mounts to models of like okay I see I’ve read about this sort of world.
它建立了這些心智模型，讓你覺得「好的，我讀過這種世界」。 - Yeah.
- I think I think for me this is like the reason that I joined Anthropic actually cuz uh you know like like I said I was living in this rural place.
我覺得對我來說，這其實就是我加入 Anthropic 的原因，因為如我所說，我住在那個鄉下的地方。 - I was thinking these longtime skills because everything is just so slow out there at least compared to SF.
我在用長時間尺度思考，因為那裡的一切都好慢，至少跟舊金山比起來。 - Um and just like all the things that you do are based around the seasons and it’s based around this food that takes many many months.
你做的所有事情都圍繞著季節，圍繞著需要好幾個月才能完成的食物。 - That’s the way that kind of like social events are organized.
社交活動就是這樣組織的。 - That’s the way you kind of organize your time.
你就是這樣安排你的時間。 - You like you go to the farmers market and it’s like it’s pimmen season and you know that because there’s like 20 pimmen vendors and then the next week the season is done and it’s like grape season and you kind of see this.
你去農夫市集，現在是青椒的季節，你知道因為有 20 個賣青椒的攤販，然後下一週季節就結束了，變成葡萄的季節，你就這樣看著。 - So it’s like these kind of longtime skills and I was also reading a bunch of sci-fi at the time and just like being in this moment I was like you know just thinking about these long time scales.
所以就是這種長時間尺度，我那時也在讀一堆科幻小說，身處那個時刻，我在思考這些長時間尺度。 - I know how this thing can go and I just I felt like I had to contribute to it going a little bit better and that’s actually why I ended up at Ant and Ben man was also a big part of that too.
我知道這件事可能會怎麼發展，我覺得我必須為它能走得好一點做出貢獻，這其實就是我最後來到 Anthropic 的原因，Ben Mann 也是很重要的一部分。 - » I feel like I want to do a whole podcast just talking about your time in Japan and the journey of Boris through Japan to Anthropic but we’ll keep it we’ll keep it short.
» 我覺得我想做一整集 podcast 只談你在日本的時光，以及 Boris 從日本到 Anthropic 的旅程，但我們就簡短帶過。 - Uh I’ll quickly recommend a sci-fi book to you if you haven’t read it.
我快速推薦一本科幻小說給你，如果你還沒讀過的話。 - Have you read Fire Upon the Deep?
你讀過 A Fire Upon the Deep 嗎？ - » Uh this is Ving, right?
» 這是 Vernor Vinge 的，對吧？ - Yeah.
對。 - It’s great.
很棒。 - » Yes.
» 對。
對科幻小說與 AI 的個人反思
- Okay.
好的。 - That one’s like it’s like so interesting from a AI AGI perspective.
從 AI AGI 的角度來看，那本書真的很有趣。 - Uh so few people have read that so um I myself » Yeah.
很少人讀過那本書，所以我自己…… » 對。 - It’s like a lot.
那本書很長。 - » Yeah.
- Yeah.
- Yeah.
- I like Deepness in the Sky also.
我也喜歡 A Deepness in the Sky。 - I think those sequels, right?
那些是續集，對吧？ - Or » Yeah.
或者…… » 對。 - » Yeah.
- Yeah.
- Yeah.
- I think so.
我想是的。 - » Yeah.
- It’s very long and like complex to get into but so good.
它很長，很難進入狀態，但真的很棒。 - Okay.
好的。
快問快答開始
- We’ll keep going through a lightning round.
我們繼續快問快答。 - Uh do you have a favorite recent movie or TV show you really enjoyed?
你有最近喜歡的電影或影集嗎？
有限的影視觀看
- » So, I actually don’t really watch TV or movies.
» 其實我不太看電視或電影。
喜愛三體影集
- I just don’t really have time these days.
我最近真的沒什麼時間。 - Um, I did watch I I I’m going to bring up another Cixin Liu, but the three body problem series on Netflix I I really loved.
我又要提到劉慈欣了，但 Netflix 上的三體影集我真的很喜歡。 - Um, I thought that was like a great rendition of the book series.
我覺得那是對小說系列的一個很棒的改編。 - » So, the common pattern across uh AI leaders is no time to watch TV or movies, which I completely understand.
» 所以 AI 領導者的共同模式就是沒時間看電視或電影，我完全能理解。
最愛產品：Cowork
- Uh, is there a favorite product you’ve recently discovered that you really love?
有沒有你最近發現的最愛產品？ - » I’m going to like chill a little bit and just say Cowork cuz this is legitimately the the one product that’s been pretty life-changing for me.
» 我要稍微自賣自誇一下，就說 Cowork，因為這真的是對我來說相當改變生活的一個產品。 - uh just because I I have it running all the time and uh the the Chrome integration in particular is just really excellent.
因為我一直讓它在執行，而且 Chrome 整合特別出色。 - Uh so it’s been like it paid a traffic fine for me.
它幫我繳了一張交通罰單。 - It like canceled a couple subscriptions for me.
幫我取消了幾個訂閱。 - Uh just like the amount of like tedious work it gets out of the way is awesome.
它幫你處理掉的那些繁瑣工作量真的太棒了。 - I I also don’t know if it’s a product, but maybe I’ll I’ll uh also another podcast that I really love obviously besides uh besides Venny is » obviously » Yeah, it’s uh it’s the acquired uh podcast by Ben Ben and David.
我也不知道這算不算產品，但我還有另一個很愛的 podcast，除了 Lenny 的之外，當然。» 當然。» 對，是 Ben 和 David 的 Acquired podcast。
最愛 Podcast：Acquired
- » Uh it’s it’s just like super it’s super awesome.
» 它真的超棒的。 - Um, I feel like the way that they get into like business history and bring it alive is is really really good.
我覺得他們深入商業歷史並讓它活起來的方式真的非常好。 - And I would start with a Nintendo episode if uh if you haven’t listened to it.
如果你還沒聽過的話，我建議從任天堂那集開始。
有效使用 Cowork
- » Great tip uh with Cowork just so people understand if they haven’t tried this like basically you type something you want to get done and it can launch Chrome and just do things for you.
» 關於 Cowork 的好建議，讓還沒試過的人了解一下，基本上你打字輸入你想完成的事，它就能啟動 Chrome 幫你做事。 - I saw one of the, someone went on pat leave from Anthropic and he had it fill out these, like, medical forms for him.
我看到有人從 Anthropic 請了育嬰假，然後他讓它幫他填那些醫療表格。 - These, like, really annoying PDFs where it just, like, loads up the browser, logs in, fills them out, and submits them.
那些很煩人的 PDF，它就是開啟瀏覽器、登入、填寫、然後提交。 - » Yeah, exactly.
» 對，沒錯。 - Exactly.
完全正確。 - And and it actually just kind of works.
而且它真的就是能用。 - Like, we tried this experiment like a year ago and it didn’t really work ‘cause the model wasn’t ready, but now, now it actually just works.
我們大約一年前試過這個實驗，那時候不太行因為模型還沒準備好，但現在，現在它真的就能用了。 - And it’s amazing.
真的很驚人。 - I think a lot of people just don’t really understand what this is because they haven’t used agents before.
我覺得很多人不太理解這是什麼，因為他們之前沒用過 agent。 - And it, it just feels very, very similar to me to Claude.ai a year ago.
對我來說，它感覺非常像一年前的 Claude.ai。 - Um, but like I said, it’s just growing much faster than Claude.ai did in the early days.
但如我所說，它的成長速度比 Claude.ai 早期快得多。 - So, I think it’s starting to, it’s starting to break through a bit.
所以我覺得它正在開始突破。 - » And there’s also this Chrome extension that you mentioned that you could just use standalone that sits in Chrome and you could just talk to Claude, uh, looking at your screen, at your browser and have it do stuff, have it tell you about what you’re looking at, summarize what you’re looking at, things like that.
» 還有你提到的那個 Chrome 擴充套件，可以獨立使用，它就在 Chrome 裡面，你可以跟 Claude 對話，它看著你的螢幕、你的瀏覽器，讓它做事、告訴你正在看什麼、摘要你正在看的東西之類的。 - » Exactly.
» 沒錯。 - Exactly.
完全正確。 - For, for people that are, like, just starting to use Cowork, the thing I recommend is, so you download the Claude desktop app, you go to the Cowork tab.
對於剛開始用 Cowork 的人，我建議的是，你下載 Claude 桌面應用程式，去 Cowork 的分頁。 - It’s right next to the code tab.
就在 code 分頁旁邊。 - Um, the thing that I recommend doing is, like, start by having it use a tool.
我建議做的是，先讓它使用一個工具。 - So, like, clean up your desktop or, like, summarize your email or something like this, or, you know, like, respond to the top three emails, like, it actually just responds to emails for me now, too.
比如清理你的桌面，或摘要你的電子郵件之類的，或回覆前三封郵件，它現在也真的會幫我回覆郵件。 - The second thing is connect tools.
第二件事是連接工具。 - So, like, if you connect, like, if you say, look at my top emails and then send Slack messages or, you know, like, put them in a spreadsheet or something.
比如你說，看看我最重要的郵件然後發 Slack 訊息，或把它們放到試算表之類的。 - Or, for example, like, I use it for all my project management.
或者，比如我用它來做所有的專案管理。 - So we have a single spreadsheet for the whole team.
我們整個團隊有一個試算表。 - There’s like a row per engineer.
每個工程師一行。 - Every week, everyone fills out a status and every Monday, Cowork just goes through and it messages every engineer on Slack that hasn’t filled out their status and so I don’t have to do this anymore.
每週每個人填寫狀態，每週一 Cowork 就會自動檢查，然後在 Slack 上通知每個還沒填寫狀態的工程師，所以我不用再做這件事了。 - And this is just one prompt.
這只是一個 prompt。 - It’ll do everything.
它會做好所有事。 - And then the third thing is just run a bunch of Claude.ais in parallel.
然後第三件事就是同時執行一堆 Claude.ai。 - So, we can Cowork, you can have as many tasks running as you want.
在 Cowork 裡，你可以同時執行任意多個任務。 - So it’s like, start one task, you know, I have this project management thing running, then I’ll have it do something else, then something else and I’ll kick these off and then I just go get a coffee while it runs.
就像是啟動一個任務，我讓專案管理的那個跑著，然後再讓它做另一件事，再做另一件事，我把這些都啟動，然後就去倒杯咖啡等它跑。 - There’s a post I’ll link to that shares a bunch of ways people use, uh, what was previously Claude.ai and now just you can do through Cowork because a lot of this is just like, “Oh, wow, I hadn’t thought I could use it for that.”
有一篇文章我會附連結，分享了很多人使用以前 Claude.ai 的方式，現在都可以透過 Cowork 來做，因為很多時候就是「哇，我沒想到可以這樣用」。
產品評估與洞察
- And once you see, like, these examples, I think are what people need to hear of just like, “Oh, wow, I didn’t know I could do that.” » So.
一旦你看到這些例子，我覺得這就是人們需要聽到的，就是「哇，我不知道還可以這樣用」。» 沒錯。 - » Yeah, I think a lot of this was also, » Some of this was also inspired by you.
» 對，我覺得很多這些也是…… » 有一些也是受到你的啟發。 - Any » You, you had this post about, uh, it was like 50 non-technical use cases for Claude.ai or something like this.
» 你之前發過一篇文章，大概是 50 個 Claude.ai 的非技術用途之類的。 - » So we actually, one of our PMs used that as a way to evaluate Cowork before we released it.
» 我們其實有一個 PM 用那個來評估 Cowork，在我們發布之前。 - Um, and I think at the point where we hit where Cowork was able to do like 48 out of the 50, they were like, “Okay, it’s pretty good.” » Wow.
我覺得當 Cowork 能做到 50 個裡面的 48 個的時候，他們就說「好的，這蠻不錯的了」。» 哇。 - I did not know that.
我不知道這件事。 - That is awesome.
太棒了。 - Uh, it’s I’ve become an eval.
我變成了一個評測基準。 - » Yeah.
» 對。 - How does that feel?
感覺如何？ - » Amazing.
» 太棒了。 - I feel like I’m valuable to the future of AI.
我覺得我對 AI 的未來很有價值。 - » This is like reverse breaking through.
» 這就像是反向突破。 - » Wow, that is so cool.
» 哇，太酷了。 - Wow.
- Okay.
好的。 - I wonder what those last two are.
我很好奇最後那兩個是什麼。
最後問題的引言
- Anyway, okay, two more questions.
好的，再問兩個問題。
人生座右銘：運用常識
- Um, do you have a favorite life motto that you often come back to in work or in life?
你有沒有一句在工作或生活中經常回想的座右銘？ - » Use common sense.
» 運用常識。 - I think a lot of the failures that I see in, especially in a work environment, is people just failing to use common sense.
我覺得我看到的很多失敗，特別是在工作環境中，就是人們沒有運用常識。 - Like they follow a process without thinking about it.
他們照著流程走，卻沒有去思考。 - Um, they just do a thing without thinking about it or they’re working on a product that’s like not a good product or not a good idea and they’re just following the momentum and not thinking about it.
他們就是做一件事卻不去想，或者他們在做一個不好的產品或不好的想法，只是跟著慣性走而不去思考。 - I think the best results that I see are people thinking from first principles and just developing their own common sense.
我覺得我看到最好的結果，是人們從第一原理思考，發展自己的常識。 - Like if something smells weird, then you know it’s probably not a good idea.
如果某件事聞起來怪怪的，那它大概就不是個好主意。 - So I think I think just this this is the single advice that I give, you know, to Coworkers more more than anything too.
所以我覺得這是我給同事最多的建議。 - And » I feel like that alone could be its own podcast conversation.
» 我覺得光是這個就可以是一整集 podcast 的對話。 - What is common sense?
什麼是常識？ - How do you build?
你怎麼建立它？ - But we’ll keep this short.
但我們就簡短帶過。
轉向 Twitter/X
- Uh final question.
最後一個問題。
Twitter/X 的使用經驗與互動
- Uh so you’ve been got more active on Twitterx.
你在 Twitter/X 上變得更活躍了。 - I’m curious just uh why and just what’s your experience been with with Twitter, the world of Twitter?
我很好奇為什麼，還有你在 Twitter 這個世界的經驗是什麼？ - Uh because you get a lot of engagement on on Twitterx.
因為你在 Twitter/X 上得到很多互動。 - » So for a long time I used Threads exclusively because I actually helped build threads a little bit back in the day.
» 很長一段時間我只用 Threads，因為我以前其實有參與 Threads 的建構。 - Um and I also just like the design.
而且我也就是喜歡那個設計。 - It’s like a very clean product.
它是一個非常乾淨的產品。 - I I just really like that.
我就是很喜歡那個。 - » I started using Threads cuz actually I was bored.
» 我開始用 Threads 是因為我無聊。 - Um so in in December I was in Europe.
十二月的時候我在歐洲。 - » You started using Twitter, you mean?
» 你是說開始用 Twitter？ - » Oh yeah.
» 噢對。 - Yeah.
- Yeah.
- I started I started using uh Twitter because I was bored.
我開始用 Twitter 是因為我無聊。 - So my my wife and I were uh we were traveling around in in Europe for December.
我和太太十二月的時候在歐洲到處旅行。 - We’re just kind of nomading around.
我們就是到處遊牧。 - We went to like Copenhagen, went to like a few different countries.
我們去了哥本哈根，去了幾個不同的國家。 - Um and for me it was just like a coding vacation.
對我來說那就像是一個寫程式的假期。 - So every day I was coding and that’s like my favorite kind of vacation just to just like code all day.
所以每天我都在寫程式，那是我最喜歡的假期方式，就是整天寫程式。 - It’s the best.
最棒了。 - And at some point I just kind of got bored and like I ran out of ideas for you know like a few hours.
然後在某個時間點我有點無聊，有幾個小時想不到要做什麼。 - I was like okay what do I want to do next?
我想說好吧，接下來要做什麼？ - And so I opened Twitter.
所以我打開了 Twitter。 - I saw some people like tweeting about Claude Code and then I just started responding and then I was like okay maybe actually I think I should do is just like look for people look for bugs that people have maybe people have like bugs or kind of feedback they have and so kind of introduce myself ask for if people had a bunch of bugs and feedback and I think they were kind of surprised by like the pace at which we’re able to address feedback nowadays.
我看到一些人在推文談 Claude Code，然後我就開始回覆，然後我想也許我應該做的就是去找人們有的 bug，也許人們有 bug 或回饋，所以我就自我介紹，問人們有沒有 bug 和回饋，我覺得他們對我們現在處理回饋的速度感到蠻驚訝的。 - Um, for me it’s just like so normal like if someone has a bug like I can probably fix it within a few minutes because I just sort of Claude and as long as the description is good it’ll just go and do it and then I’ll I’ll go do something else and answer the next thing.
對我來說這很正常，如果有人有 bug，我大概幾分鐘內就能修好，因為我就用 Claude，只要描述夠好它就會去做，然後我就去做其他事情、回答下一個問題。 - But I think for a lot of people was pretty surprising.
但我覺得對很多人來說這蠻驚訝的。 - So that was really cool and yeah the experience on Twitter has been pretty great.
所以那真的很酷，在 Twitter 上的經驗蠻棒的。 - It’s it’s been awesome just engaging with people and seeing what people want uh hearing hearing about bugs, hearing about features.
跟人們互動、看到人們想要什麼、聽到 bug、聽到功能需求，這真的很棒。 - I saw complaints to Nikita Beer the other day on Twitter of just you they’re like posting many threads and it was breaking and just like oh man what’s going on here.
我前幾天在 Twitter 上看到有人跟 Nikita Beer 抱怨，說他們發很多串文然後就壞掉了，就覺得天啊怎麼回事。 - » Yeah.
- Yeah.
- Yeah.
- There there was a bug.
那裡有一個 bug。 - I hope it’s fixed now.
我希望現在已經修好了。 - Amazing.
太棒了。
結語
- Oh man, Boris, I could chat with you for hours.
天啊 Boris，我可以跟你聊好幾個小時。
如何找到與互動
- Uh I’ll let you go.
我就讓你走了。 - Thank you so much for doing this.
非常感謝你來做這個。 - Uh you’re wonderful.
你太棒了。 - Um where can folks find you online?
大家可以在網路上哪裡找到你？ - How can listeners be useful to you?
聽眾可以怎麼幫助你？ - » Yeah, find me on threads or on Twitter.
» 在 Threads 或 Twitter 上找我。 - That’s the that’s the easiest place.
那是最容易的地方。 - And please just tag me on stuff.
請標記我。 - Um, send bugs, send feature requests, what’s missing, what can we do to make the products better?
送 bug、送功能需求，什麼缺少了，我們可以怎麼讓產品更好？ - What do you like?
你喜歡什麼？ - What do you want?
你想要什麼？ - Um, I I love love hearing » Amazing.
我非常喜歡聽到…… » 太棒了。 - Boris, thank you so much for being here.
Boris，非常感謝你來到這裡。 - » Cool.
» 酷。 - Thanks, Funny.
謝謝，Lenny。
Podcast 結尾與訂閱
- » Bye, everyone.
» 大家再見。 - » Thank you so much for listening.
» 非常感謝你的收聽。 - If you found this valuable, you can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app.
如果你覺得這有價值，你可以在 Apple Podcasts、Spotify 或你喜歡的 podcast app 上訂閱這個節目。 - Also, please consider giving us a rating or leaving a review as that really helps other listeners find the podcast.
也請考慮給我們評分或留下評論，這真的有助於其他聽眾找到這個 podcast。 - You can find all past episodes or learn more about the show at lennispodcast.com.
你可以在 lennispodcast.com 找到所有過去的集數或了解更多關於這個節目的資訊。 - See you in the next episode.
下一集見。