Inside Claude Code With Its Creator

"At Anthropic, the way that we thought about it is: we don't build for the model of today."

"在 Anthropic,我们的思路是:不为今天的模型造产品。"

"We build for the model six months from now."

"我们为 6 个月后的模型造产品。"

"That's actually still my advice to founders that are building on LLMs."

"这其实也是我对所有在 LLM 上做产品的创始人的建议。"

"Just try to think about what is that frontier where the model is not very good today — because it's going to get good at it."

"你只要想清楚一件事:今天模型还不擅长的那条边界——因为它一定会擅长。"

"All of Claude Code has just been written and rewritten and rewritten and rewritten over and over and over."

"Claude Code 的每一行代码都被反反复复重写、重写、重写过无数次。"

"There is no part of Claude Code that was around 6 months ago."

"Claude Code 里没有任何一部分是 6 个月前就存在的。"

"You try a thing, you give it to users, you talk to users, you learn — and then eventually you might end up at a good idea."

"你试一个东西,丢给用户,跟用户聊,从中学到东西——然后你可能找到一个好主意。"

"Sometimes you don't."

"也有时候找不到。"

"Are you also in the back of your mind thinking that maybe in 6 months you won't need to prompt that explicitly?"

"你心里有没有想过——也许 6 个月后,你都不用这么显式地 prompt 它了?"

"Like the model will just be good enough to figure out on its own?"

"就是说模型自己就足够聪明,能搞定?"

"Maybe in a month." [laughter]

"也许一个月内就够了。"[笑]

"No more need for plan mode in a month."

"一个月后 plan mode 就用不上了。"

"Welcome to another episode of the Lightcone."

"欢迎收看新一期 Lightcone。"

"Today we have an extremely special guest: Boris Cherny, the creator engineer of Claude Code."

"今天我们有一位非常特别的嘉宾——Boris Cherny,Claude Code 的创造者。"

"Boris, thanks for joining us."

"Boris,谢谢你来。"

"Thanks for having me."

"谢谢邀请。"

"Thanks for creating a thing that has taken away my sleep for about 3 weeks straight." [laughter]

"谢谢你做了一个让我连着 3 周睡不着的东西。"[笑]

"I am very addicted to Claude Code — it feels like rocket boosters."

"我对 Claude Code 完全上瘾了——感觉像装了火箭推进器。"

"Has it felt like this for people for, you know, months at this point?"

"这种感觉对外面的人来说,大概持续好几个月了吧?"

"I think it was end of November where a lot of my friends said something changed."

"我感觉是 11 月底——身边一堆朋友突然说"有什么不一样了"。"

"I remember for me, I felt this way when I first created Claude Code, and I didn't yet know if I was on to something."

"我记得自己第一次做出 Claude Code 的时候也是这种感觉——那时候我还不确定我是不是真做出了什么。"

"I kind of felt like I was on to something — and that's when I wasn't sleeping."

"我隐约觉得自己抓住了点东西——然后我就不睡觉了。"

"And that was just like three straight months."

"就那样连着三个月。"

"This was September 2024."

"那是 2024 年 9 月。"

"Yeah, it was like three straight months."

"对,差不多就是连着三个月。"

"I didn't take a single day vacation, worked through the weekends, worked every single night."

"一天假都没休,周末在做,每个晚上都在做。"

"I was just like, oh my god — this is, I think, on to something."

"我心里就一直在想——天哪,这个东西,真有戏。"

"If you look back on those moments to now, what would be the most surprising thing about this moment right now?"

"从那时候回头看到现在,你觉得最让你意外的事是什么?"

"It's unbelievable that we're still using a terminal."

"难以置信,我们现在还在用终端。"

"That was supposed to be the starting point — I didn't think that would be the ending point."

"那本来只是个起点——我没想到它会成为终点。"

"And then the second one is that it's even useful — because at the beginning it didn't really write code."

"第二件意外的事是,它居然真有用——因为最开始它其实不太会写代码。"

"Even in February when we GA'd, it wrote maybe like 10% of my code or something."

"哪怕 2 月份正式 GA 的时候,它也就写我大概 10% 的代码。"

"I didn't really use it to write code — it wasn't very good at it."

"我那时候不太用它写代码——它不擅长。"

"I still wrote most of my code by hand."

"大部分代码我还是手写。"

"So the fact that our bets paid off and it got good at the thing that we thought it was going to get good at — because it wasn't obvious."

"所以最让我意外的是:我们押的注真兑现了——它后来真的擅长了我们当时押它擅长的事——这在当时一点都不显然。"

"At Anthropic, the way that we thought about it is: we don't build for the model of today, we build for the model 6 months from now."

"在 Anthropic,我们的思路是:不为今天的模型造产品,为 6 个月后的模型造。"

"And that's actually still my advice to founders that are building on LLMs."

"这也是我对所有基于 LLM 做产品的创始人的建议。"

"Just try to think about what is that frontier where the model is not very good at today — because it's going to get good at it, and you just have to wait."

"想清楚那条边界——今天模型还不擅长的事——因为它一定会擅长,你只需要等。"

"Going back though — when do you remember when you first got the idea?"

"再往回倒一点——你还记得第一次有这个想法是什么时候吗?"

"Can you just talk us through that?"

"能带我们走一遍那个过程吗?"

"Like was it some spark?"

"是某个突然的灵感吗?"

"What was even the first version of it in your mind?"

"你脑子里那个最初的版本长什么样?"

"You know, it's funny — it was so accidental that it just kind of evolved into this."

"其实挺好笑的——它太意外了,就这么慢慢演变成了今天这样。"

"At Anthropic, the bet has been coding for a long time."

"Anthropic 长期押的就是 coding。"

"The bet has been: the path to safe AGI is through coding."

"我们押的是:通往安全 AGI 的路径要经过 coding。"

"This has kind of always been the idea — and the way you get there is: you teach the model how to code, then how to use tools, then how to use computers."

"这个想法一直都在——而到达那里的方式是:先教模型写代码,再教它用工具,再教它操作电脑。"

"You can kind of see that because the first team that I joined at Anthropic was called the Anthropic Labs team, and it produced three products: Claude Code, MCP, and the desktop app."

"你能看出这个脉络——我在 Anthropic 加入的第一个团队叫 Anthropic Labs,出了三个产品:Claude Code、MCP、桌面应用。"

"So you can see how these weave together."

"你能看到这三件事是怎么编织在一起的。"

"The particular product that we built — you know, no one asked me to build a CLI."

"我们做出来的这个产品——没人让我做 CLI。"

"We kind of knew maybe it was time to build some kind of coding product, because it seemed like the model was ready."

"我们隐约知道也许该做个 coding 产品了——因为模型看起来准备好了。"

"But no one had yet really built the product that harnessed this capability."

"但还没人真正做出一个能驾驭这种能力的产品。"

"So like still there's this insane feeling of product overhang."

"即使是今天,也还存在一种很疯狂的产品过剩感(product overhang)。"

"But at the time it was just like even crazier, because no one had built this yet."

"但当时更夸张——因为还根本没人做过。"

"So I started like hacking around, and I was like: okay, we build a coding product — what do I have to do first?"

"我就开始随便 hack——我心想:好,要做个 coding 产品——那我第一步该干嘛?"

"I have to understand how to use the API, because I hadn't used the Anthropic API at that point."

"我得先搞懂怎么用 API——因为我那时候还没用过 Anthropic 的 API。"

"So I just built a little terminal app to use the API."

"于是我就做了一个小终端 app 来调 API。"

"That's all that I did."

"我就只做了这么点事。"

"It was a little chat app — because, you know, you think about the AI applications of the time."

"那是个小聊天 app——因为你想想那时候的 AI 应用都什么样。"

"For non-coders today, what are most people using? Just a chat app."

"今天非 coder 用的最多的是什么?也就是个聊天 app。"

"So that's what I built — in a terminal."

"所以我做的就是这个——跑在终端里。"

"I can ask questions, it gives answers."

"我问问题,它给答案。"

"Then tool use came out, and I just wanted to try out tool use, because I didn't really understand what this is."

"后来 tool use 出来了,我就想试试 tool use——因为我也没真正理解它是什么。"

"I was like: tool use is cool."

"我心想:tool use 听着挺酷的。"

"Is this actually useful? Probably not."

"它真有用吗?估计没用。"

"Let me just try it."

"试一下吧。"

"You built it in terminal just because it was the easiest way to get something up and running."

"你在终端里做,纯粹因为这是最快能跑起来的方式?"

"Yes — because I didn't have to build a UI."

"对——因为我不用搭 UI。"

"It was just me at that point."

"那时候就我一个人。"

"It was like the IDEs — Cursor, Windsurf — taking off."

"那时候像 Cursor、Windsurf 这些 IDE 正在火。"

"Were you sort of under any pressure or getting lots of suggestions of: hey, like we should build this out as a plugin, or as a fully featured IDE itself?"

"你那时候有没有压力——或者有人一直建议你——说"我们应该把这做成 plugin,或者做成一个完整的 IDE"?"

"There was no pressure because we didn't even know what we wanted to build."

"没有压力——因为我们连自己想做什么都还不知道。"

"The team was just in explore mode — we knew vaguely we wanted to do something in coding, but it wasn't obvious what."

"团队完全是 explore mode——我们隐约知道想在 coding 这块做点东西,但具体做什么不清楚。"

"No one was high confidence enough — that was my job to figure out."

"没人有足够的把握——这就是我要弄清楚的事。"

"So I gave the model the bash tool — that was the first tool that I gave it, just because that was literally the example in our docs." [laughter]

"我给模型加了 bash 工具——这是我给它的第一个工具,纯粹因为这是我们文档里的示例。"[笑]

"I just like took the example."

"我直接抄了那个示例。"

"It was in Python — I just ported it to TypeScript because that's how I wrote it."

"示例是 Python 的,我把它 port 成 TypeScript——因为我用 TypeScript 写。"

"I didn't know what the model could do with bash."

"我也不知道模型拿到 bash 能做什么。"

"So I asked it to read a file."

"我让它读一个文件。"

"It could cat the file — that was cool."

"它真的 cat 了那个文件——这就挺酷。"

"And then I was like: okay, like what can you actually do?"

"然后我就想:好,你究竟能做什么?"

"And I asked it: what music am I listening to?"

"我问它:我现在在听什么音乐?"

"It wrote some AppleScript to script my Mac and look up the music in my music player."

"它写了段 AppleScript 去调我的 Mac,从音乐播放器里查出来了。"

"Oh my god."

"我的天。"

"And this was Sonnet 3.5."

"那时候是 Sonnet 3.5。"

"I didn't think the model could do that — and that was my first ever "feel the AGI" moment."

"我没想到模型能做到——这是我有史以来第一次"feel the AGI"的瞬间。"

"I was just like: oh my god, the model — it just wants to use tools."

"我心里就一直在想:天哪,这个模型——它就是想用工具。"

"That's all it wants."

"它就只想用工具。"

"That's kind of fascinating."

"这件事还挺迷人的。"

"I mean, it's very contrarian that Claude Code works so well in such an elegant, simple form factor."

"我是说,Claude Code 能在这么极简优雅的形态下跑得这么好,这是很反共识的事。"

"Terminals have been around for a really long time, and that seemed to be like a good design constraint that allowed a lot of interesting developer experiences."

"终端已经存在很久了——它本身像是个很好的设计约束,反而催生出了很多有意思的开发体验。"

"Like, it doesn't feel like working — it just feels fun as a developer."

"它不像在干活——对开发者来说,它就是好玩。"

"I don't think about files, where everything is."

"我都不用想文件在哪、东西在哪。"

"And that came by accident almost."

"这居然几乎是个意外。"

"Yeah, it was an accident."

"对,是意外。"

"After the terminal started to take off internally — honestly, like 2 days after the first prototype — I started giving it to my team just for dogfooding."

"在终端版本内部开始火起来之后——其实第一个原型出来后大概 2 天——我就把它丢给团队 dogfood 了。"

"Because, you know, if you come up with an idea and it seems useful, the first thing you want to do is give it to people to see how they use it."

"因为你想啊,你有个看起来有用的点子,第一件事就是丢给人,看他们怎么用。"

"And then I came in the next day, and Robert — who sits across from me, another engineer — he just had Claude Code on his computer and was using it to code."

"我第二天来,Robert——坐我对面的另一个工程师——电脑上已经装上 Claude Code,在拿它写代码了。"

"I was like: what are you doing? This thing isn't ready. It's just a prototype."

"我心想:你干嘛呢?这东西还没准备好,就是个原型。"

"But yeah, it was already useful in that form factor."

"但事实是——在那个形态下它已经有用了。"

"And I remember when we did our launch review to launch Claude Code externally — this was in November/December 2024."

"我记得我们做 launch review 准备对外发布的时候——大概是 2024 年 11 月、12 月。"

"Dario asked: the DAU chart internally is like vertical — are you forcing engineers to use it? Why are you mandating them?"

"Dario 问我:内部的 DAU 曲线像火箭一样——你是在强制工程师用吗?为什么这么硬推?"

"And I was just like: no, no, we didn't."

"我说:没没没,我们没强推。"

"I just posted about it and they'd just been telling each other about it."

"我就发了个帖,然后他们就互相安利。"

"Honestly, it was just accidental."

"说实话,完全是意外。"

"We started with the CLI because it was the cheapest thing — and it just kind of stayed there for a bit."

"我们从 CLI 开始,纯粹因为它是最便宜的方式——然后就这么一直待在 CLI 里了一段时间。"

"So in that 2024 period, how were the engineers using it?"

"那么在 2024 年那段时间,工程师是怎么用它的?"

"Were they shipping code with it yet, or were they using it in a different way?"

"他们已经用它发代码了吗?还是当时用法不一样?"

"The model was not very good at coding yet."

"那时候模型写代码还不太行。"

"I was using it personally for automating git."

"我自己是用它来自动化 git。"

"I think at this point I've probably forgotten most of my git, because Claude Code has just been doing it for so long." [laughter]

"我估计现在很多 git 命令我都忘了——因为太久都是 Claude Code 在帮我做了。"[笑]

"But yeah, automating bash commands — that was a very early use case."

"对,自动化 bash 命令——那是个很早的 use case。"

"And operating things like Kubernetes."

"还有操作 Kubernetes 这类东西。"

"People were using it for coding too — there were some early signs of this."

"也有人在用它写代码——这件事早期已经有信号了。"

"The first use case was actually writing unit tests, because it's a little bit lower risk."

"最早一个真用起来的 use case 其实是写单元测试,因为这事风险更低一点。"

"And the model was still pretty bad at it." [laughter]

"而且那时候模型写单测也挺烂的。"[笑]

"But people were kind of figuring it out — figuring out how to use this thing."

"但用户都在自己摸索——摸索怎么用这玩意儿。"

"One thing that we saw is people started writing these markdown files for themselves, and then having the model read that markdown file."

"我们当时看到的一个现象是——用户自己开始写 markdown 文件,然后让模型去读这个 markdown。"

"And this is where CLAUDE.md came from."

"CLAUDE.md 就是这么来的。"

"Probably the single biggest principle in product, for me, is latent demand."

"我自己最信奉的产品原则,就是 latent demand(潜在需求)。"

"Every bit of this product is built through latent demand after the initial CLI."

"在最初的 CLI 之后,这个产品的每一寸都是顺着 latent demand 长出来的。"

"So CLAUDE.md is an example of that."

"CLAUDE.md 就是一个例子。"

"There's another general principle that's maybe interesting."

"另外还有个一般原则,可能有意思。"

"You can build for the model, and then you can build scaffolding around the model in order to improve performance a little bit."

"你可以为模型造产品,然后在模型外面搭一些 scaffolding,让性能提升一点。"

"Depending on the domain, you can improve performance maybe 10-20%, something like that."

"看是什么领域,你大概能提升 10-20% 的性能。"

"And then essentially the gain is wiped out with the next model."

"然后下一代模型一来,这点提升基本就被抹掉了。"

"So either you build the scaffolding and then get some performance gain and then rebuild it again — or you just wait for the next model and then you kind of get it for free."

"所以两条路:要么你搭 scaffolding 吃这点提升然后下次重搭——要么你就等下一代模型,等到了你免费就拿到。"

"CLAUDE.md and the scaffolding is an example of that."

"CLAUDE.md 和那一类 scaffolding 就是这种东西。"

"And really, I think that's why we stayed in the CLI — because we felt there is no UI we could build that would still be relevant in 6 months, because the model was improving so quickly."

"我觉得这就是我们一直待在 CLI 里的真正原因——我们觉得没有任何一个 UI,我们造出来 6 个月后还会相关——因为模型升级太快了。"

"Earlier we were saying we should compare CLAUDE.md's."

"我们之前还说过应该互相对比一下各自的 CLAUDE.md。"

"But you said something very profound: yours is actually very short — which is almost the opposite of what people might expect."

"你说了句挺有意思的话:你的其实非常短——这几乎和大家预期的相反。"

"Why is that? What's in your CLAUDE.md?"

"为什么?你的 CLAUDE.md 里到底写了啥?"

"Okay, so I checked this before we came."

"我录之前还专门检查了一下。"

"My CLAUDE.md has two — actually, it's just two lines."

"我的 CLAUDE.md 有两条——其实就只有两行。"

"The first line is: whenever you put up a PR, enable automerge."

"第一行是:每次你提 PR,启用 automerge。"

"So as soon as someone accepts it, it's merged."

"这样有人 approve 就直接 merge 了。"

"That's just so I can code and don't have to go back and forth with CR or whatever."

"这样我可以专心写代码,不用反复跟 code review 来回扯。"

"And then the second one is: whenever I put up a PR, post it in our internal team stamps channel."

"第二条是:每次我提 PR,把它发到我们内部的 stamps 频道。"

"Just so someone can stamp it and I can get unblocked."

"这样有人能盖个章,我就能解锁继续干了。"

"The idea is: every other instruction is in our CLAUDE.md that's checked into the codebase."

"思路是——其他所有指令都在 codebase 里那份 checked-in 的 CLAUDE.md。"

"It's something our entire team contributes to multiple times a week."

"那份 CLAUDE.md 是我们整个团队每周都会贡献好几次的东西。"

"Very often I'll see someone's PR and they make some mistake that's totally preventable."

"经常看到团队同事的 PR 出了个完全本可以避免的错。"

"I'll just literally tag Claude on the PR — add Claude, like, add this to the CLAUDE.md."

"我就直接在 PR 里 tag Claude——"add Claude, 把这条写进 CLAUDE.md"。"

"I'll do this many times a week."

"这事我一周要做好多次。"

"Do you have to like compact the CLAUDE.md?"

"那你需要"compact"你的 CLAUDE.md 吗?"

"I definitely reached a point where I got the message at the top saying: your CLAUDE.md is like thousands of tokens now."

"我自己就遇到过那种顶上提示——你的 CLAUDE.md 已经几千 tokens 了。"

"What do you do when you guys hit that?"

"你们遇到这种情况怎么办?"

"So our CLAUDE.md is actually pretty short."

"我们的 CLAUDE.md 其实挺短的。"

"I think it's like a couple thousand tokens, maybe something like that."

"大概几千 tokens 吧。"

"If you hit this, my recommendation would be: delete your CLAUDE.md and just start fresh."

"如果你撞到上限——我的建议是:删掉你的 CLAUDE.md,从头开始。"

"Interesting."

"有意思。"

"I think a lot of people try to overengineer this."

"我觉得很多人在这件事上过度工程化了。"

"And really — the capability changes with every model."

"而事实是——每一代模型,能力都变了。"

"So the thing you want is: do the minimal possible thing in order to get the model on track."

"你要做的事是:用最小的干预,把模型扳回正轨。"

"If you delete your CLAUDE.md and then the model is getting off track, doing the wrong thing — that's when you add back a little bit at a time."

"你删了 CLAUDE.md,然后模型开始偏了、做错了——这时候你再一点点加回去。"

"And what you're probably going to find is: with every model, you have to add less and less."

"你大概率会发现:每一代模型,你要加的东西越来越少。"

"For me, I consider myself a pretty average engineer, to be honest."

"说实话,我觉得自己就是个挺普通的工程师。"

"I don't use a lot of fancy tools."

"我不太用花哨的工具。"

"Like, I don't use Vim — I use VS Code because it's simpler."

"比如我不用 Vim——我用 VS Code,因为更简单。"

"Wait, really?"

"等等,真的?"

"I would have assumed that because you built this in the terminal, you were sort of a die-hard terminal Vim-only person."

"我以为你既然在终端里做了这个东西,那肯定是死磕终端、只用 Vim 那种人。"

"You know, screw those VS Code people." [laughter]

"管那些用 VS Code 的人去死。"[笑]

"Well, we have people like that on the team."

"我们团队里就有这种人。"

"Like Adam Wolf, for example — he's on the team."

"比如 Adam Wolf——他在我们团队。"

"He's like: you will never take Vim from my cold dead hands." [laughter]

"他原话是:你永远别想从我冰冷的死手里拿走 Vim。"[笑]

"There's definitely a lot of people like that on the team."

"团队里这种人确实不少。"

"This is one of the things that I learned early on: every engineer likes to hold their dev tools differently."

"我很早就学到一件事——每个工程师握 dev tools 的方式都不一样。"

"They like to use different tools — there's just no one tool that works for everyone."

"大家都用不同的工具——没有一个工具对所有人都好用。"

"But this is also one of the things that makes it possible for Claude Code to be so good."

"但这也正是 Claude Code 能做得这么好的原因之一。"

"Because I think about it as: what is the product that I would use, that makes sense to me?"

"因为我思考的方式是:我自己会用什么样的产品?对我来说合理的是什么?"

"So to use Claude Code, you don't have to understand Vim."

"所以用 Claude Code,你不用懂 Vim。"

"You don't have to understand tmux. You don't have to know how to SSH."

"不用懂 tmux,不用懂 SSH。"

"You just have to open up the tool and it'll guide you — it'll do all this stuff."

"你只需要打开这个工具,它会带着你走——所有这些东西它都会帮你做。"

"How do you decide how verbose you want the terminal to be?"

"你怎么决定终端到底要多 verbose?"

"Like, sometimes you have to go control-O and check it out."

"有时候你得按 Ctrl-O 才能看到细节。"

"Is it like internal bike-shed battles around longer/shorter?"

"团队内部会不会就长一点还是短一点这种事吵成一团?"

"I mean, every user probably has an opinion."

"我是说,每个用户多半都有自己的看法。"

"How do you make those sorts of decisions?"

"这种决策你们怎么做?"

"What's your opinion?"

"你的看法是?"

"Is it too verbose right now?"

"现在是太啰嗦了吗?"

"Oh, I love the verbosity."

"啊我超爱这个 verbosity。"

"Because basically sometimes it just goes off the deep end."

"因为有时候它会跑偏。"

"And I'm watching, and I can just read very quickly, and it's like: oh no no, it's not that."

"我能边盯边读得很快——然后心想:不不不,不是那回事。"

"And then I escape and just stop it, and it stops an entire bug farm as it's happening."

"我按 Esc 把它停下来,直接掐灭了一整个正在长出来的 bug 农场。"

"That's usually when I didn't do plan mode properly."

"通常这种情况是我没好好用 plan mode。"

"This is something that we probably change pretty often."

"这件事我们其实改得挺频繁的。"

"I remember early on — this is maybe six months ago — I tried to get rid of bash output, just internally, just to summarize it."

"我记得早期——大概 6 个月前——我试过把 bash output 砍掉,只留个摘要,纯内部试。"

"Because I was like: these giant long bash commands, I don't actually care."

"因为我想:这些超长的 bash 输出,我根本不关心。"

"And then I gave it to Anthropic employees for a day, and everyone just revolted." [laughter]

"然后丢给 Anthropic 员工试了一天——所有人都炸了。"[笑]

"I want to see my bash, because it actually is quite useful."

""我要看到 bash,因为它真的挺有用的。""

"For something like git output, maybe it's not useful — but if you're running Kubernetes jobs, you actually do want to see it."

"像 git 输出可能没必要——但你跑 Kubernetes jobs 的时候,你真的想看到。"

"We recently hid the file reads and file searches."

"我们最近把 file reads 和 file searches 折叠掉了。"

"So you'll notice instead of saying "read foo.md", it says "read one file"; instead of the full search, it says "searched one pattern"."

"你会发现现在不写 "read foo.md" 了——改成 "read one file";搜索也是 "searched one pattern"。"

"This is something I think we could not have shipped six months ago, because the model just was not ready."

"这件事 6 个月前我们做不到——因为模型还没准备好。"

"It still read the wrong thing pretty often."

"那时候它经常读错文件。"

"As a user, you still had to be there and catch it and debug it."

"作为用户你得在场盯着、抓错、debug。"

"But nowadays, I just noticed it's on the right track almost every time."

"但现在我注意到——它几乎每次都在正确的路径上。"

"And because it's using tools so much, it's actually a lot better just to summarize it."

"它用工具的频率太高,折叠成摘要其实体验更好。"

"But then we shipped it. We dogfooded it for a month, and then people on GitHub didn't like it."

"但我们发出去之后——内部 dogfood 了一个月——GitHub 上的用户不喜欢。"

"There was a big issue where people were like: no, I want to see the details."

"有个大 issue 一堆人说:不,我要看到细节。"

"And that was really great feedback."

"这个反馈非常珍贵。"

"So we added a new verbose mode — in /config you can enable verbose mode."

"于是我们加了一个新的 verbose mode——在 /config 里能打开。"

"If you want to see all the file outputs, you can continue to do that."

"如果你想看所有 file outputs,可以继续看。"

"Then I posted on the issue, and people still didn't like it — which is again awesome."

"我又在 issue 里回了一次,大家还是不满意——这又一次很棒。"

"Because my favorite thing in the world is just hearing people's feedback and hearing how they actually want to use it."

"因为我最爱的事就是听用户反馈——听他们真正想怎么用。"

"So we just iterated more and more and more to get that really good."

"我们就这么一遍遍迭代,把它打磨好。"

"There's like all these concepts now of rather than having to review the code."

"现在还有一些观点是说:你根本不该再 review 代码。"

"I'm old school, so I like the verbosity — I like to say: oh well, you're doing this, but I want you to do that."

"我比较老派,我爱 verbosity——我喜欢说:哦你在做这个,但我想让你做那个。"

"But there's a totally different school of thought now that says: anytime a real human being has to look at code, that's bad."

"但现在有完全相反的一派认为:只要还有真人需要看代码,那就是失败。"

"Which is fascinating."

"这件事还挺迷人的。"

"I think like Dan Shipper talks about this a lot."

"我觉得 Dan Shipper 经常聊这个。"

"Whenever you see the model make a mistake, try to put it in the CLAUDE.md or in skills, so it's reusable."

"每次你看到模型犯错,就把这事写进 CLAUDE.md 或者 skills 里,这样可以复用。"

"But there's this meta point that I actually struggle with a lot."

"但这件事有个 meta 层面的问题,我自己也挺纠结。"

"People talk about: agents can do this, agents can do that."

"大家都说:agent 能干这个,agent 能干那个。"

"But actually what agents can do — it changes with every single model."

"但 agent 真正能做什么——每一代模型都在变。"

"Sometimes there's a new person that joins the team, and they actually use Claude Code more than I would have used it."

"有时候团队新来一个人,他用 Claude Code 的方式比我自己用得还猛。"

"And I'm just constantly surprised by this."

"这件事一直在 surprise 我。"

"For example, we had this memory leak and we were trying to debug it."

"比如有一次我们有个 memory leak,正在调试。"

"By the way, Jared Sumar has been on this crusade killing all the memory leaks, and it's been amazing."

"顺便说一句——Jared Sumar 一直在搞一场十字军东征,把所有 memory leak 干掉,这事真的很棒。"

"But before Jared was on the team, I had to do this — there was this memory leak, I was trying to debug it."

"在 Jared 加入之前,这事得我来——当时有这么个 memory leak,我在调。"

"I took a heap dump. I opened it in DevTools. I was looking through the profile."

"我取了 heap dump,在 DevTools 里打开,翻 profile。"

"Then I was looking through the code, just trying to figure this out."

"然后翻代码,想搞清楚。"

"And then another engineer on the team, Chris, he just asked Claude Code."

"团队另一个工程师 Chris——他直接问 Claude Code。"

"He was like: hey, I think there's a memory leak. Can you run this and try to figure it out?"

"他说:嘿,我觉得有个 memory leak,你跑一下,看能不能找出来?"

"Claude Code took the heap dump, wrote a little tool for itself to analyze the heap dump, and then it found the leak faster than I did."

"Claude Code 拿到了 heap dump,自己写了个小工具去分析这个 heap dump,然后比我更快找到了那个 leak。"

"And this is just something I have to constantly relearn, because my brain is still stuck somewhere six months ago at times."

"这种事我得不断重新学——因为我的大脑有时候还停留在 6 个月前。"

"So what would be some advice for technical founders to really become maximalists at the latest model release?"

"那对技术 founder 来说,怎么才能把最新模型用到极致——你的建议是什么?"

"It sounds like people fresh off of school, or that don't have any assumptions, might be better suited than engineers who have been working at it for a long time."

"听起来刚出校门、没什么固有假设的人,可能比干了很多年的工程师更适合。"

"How do the experts get better?"

"老手怎么继续进化?"

"For myself, it's kind of beginner mindset — and humility."

"对我来说就是新手心态——加上谦逊。"

"Engineers as a discipline, we've learned to have very strong opinions."

"工程师这个职业整体被训练成有非常强的观点。"

"Senior engineers are kind of rewarded for this."

"资深工程师恰恰因为这点被奖励。"

"In my old job at a big company, when I hired architects and this kind of engineer, you look for people that have a lot of experience and really strong opinions."

"我在大厂时招架构师那种岗位——找的就是经验丰富、观点强烈的人。"

"But it actually turns out a lot of this stuff just isn't relevant anymore."

"但事实是,这一套现在很多已经不相关了。"

"And a lot of these opinions should change, because the model is getting better."

"很多观点本来就应该变——因为模型在变强。"

"So I think actually the biggest skill is people that can think scientifically and can think from first principles."

"所以我觉得现在最重要的技能是——能科学地思考、能从 first principles 出发的人。"

"How do you screen for that when you try to hire someone now for your team?"

"那你现在招人的时候,怎么筛选这种特质?"

"I sometimes ask: what's an example of when you were wrong?"

"我有时候会问:举一个你错的例子?"

"It's a really good one."

"这个问题很好用。"

"Some of these classic behavioral questions — not even coding questions — I think are quite useful."

"一些经典的行为面试问题——甚至都不是 coding 题——其实挺好用。"

"You can see if people can recognize their mistake in hindsight, if they can claim credit for the mistake, and if they learn something from it."

"你能看出这个人事后能不能识别自己的错、能不能为这个错负责、能不能从中学到东西。"

"A lot of these very senior people — especially there are some founder types like this, but I think founders in particular are actually quite good at it."

"很多很 senior 的人——尤其有些 founder 型的人——但实际上 founder 这件事上反而做得好。"

"But other people sometimes will never take the blame for a mistake."

"但有些人永远不会承认自己错。"

"For me personally, I'm wrong probably half the time."

"我自己嘛——大概一半时候是错的。"

"Like, half my ideas are bad — and you just have to try stuff."

"我一半的点子都不好——你只能不停去试。"

"You try a thing, you give it to users, you talk to users, you learn — and then eventually you might end up at a good idea."

"你试个东西,丢给用户,跟用户聊,从中学到——最终可能找到一个好主意。"

"Sometimes you don't."

"也有时候找不到。"

"This is the skill that I think in the past was very important for founders — but now I think it's very important for every engineer."

"这种能力过去对 founder 来说很关键——现在对每个工程师都关键。"

"Do you think you would ever hire someone based on the Claude Code transcript of them working with the agent?"

"你以后会不会基于"候选人和 agent 一起工作的 Claude Code transcript"来招人?"

"We're actively doing that right now."

"我们现在已经在这么干了。"

"We just added it as a test — you can upload a transcript of you coding a feature with Claude Code or Codex or whatever it is."

"我们刚加了这个测试——你能上传你用 Claude Code 或 Codex 写一个 feature 的 transcript。"

"Personally, I think it's going to work."

"我个人觉得这会管用。"

"You can figure out how someone thinks — whether they're looking at the logs, can they correct the agent if it goes off the rails, do they use plan mode, when they use plan mode do they make sure there are tests."

"你能看出这个人是怎么想问题的——他看不看日志、agent 跑偏了他会不会改、他用不用 plan mode、用 plan mode 的时候有没有保证有测试。"

"Do they think about systems? Do they even understand systems?"

"他考不考虑系统?他到底懂不懂系统?"

"There's so much that's sort of embedded in that — I just want a spider web graph."

"这里面藏的信息超多——我特别想要个雷达图。"

"Like in those video games — NBA 2K — it's like: oh, this person's really good at shooting or defense."

"就像 NBA 2K 那种——能看出这人投篮强还是防守强。"

"You could imagine a spider web graph of someone's Claude Code skill level."

"你可以想象一个雷达图——把一个人的 Claude Code 熟练度可视化。"

"Yeah. What would the skills be?"

"是啊——那这些轴会是什么?"

"I think it's like systems, testing, user behavior — there's got to be a design part, product sense, automating stuff."

"我觉得会有:系统、测试、用户行为——肯定还得有设计、产品 sense、自动化能力。"

"My favorite thing in CLAUDE.md, for me, is: I have a thing that says for every plan, decide whether it's overengineered, underengineered, or perfectly engineered — and why."

"我自己 CLAUDE.md 里最喜欢的一条是:对每个 plan,判断它是 over-engineered、under-engineered 还是 perfectly engineered——以及为什么。"

"I think this is something that we're trying to figure out too."

"这事我们自己也在摸索。"

"When I look at engineers on the team that I think are the most effective, it's essentially bimodal."

"我看团队里最有效的工程师——基本是双峰分布。"

"There's one side where it's extreme specialists."

"一端是极端的专才。"

"Like, I named Jared before — he's a really good example."

"我前面提到的 Jared 就是个好例子。"

"And the Bun team is a really good example — just hyper specialists."

"Bun 团队也是——极致专才。"

"They understand dev tools better than anyone else."

"他们对 dev tools 的理解比任何人都深。"

"They understand JavaScript runtime systems better than anyone else."

"对 JavaScript runtime 的理解也比任何人都深。"

"And then there's the flip side of kind of hyper generalists — and that's kind of the rest of the team."

"另一端是 hyper generalist——团队大部分人属于这一端。"

"A lot of people they span like product and infra, or product and design, product and user research, product and business."

"很多人横跨 product 和 infra、product 和 design、product 和 user research、product 和 business。"

"I really like to see people that just do weird stuff."

"我特别喜欢那种会去做"怪事"的人。"

"I think that's one of these things that was kind of a warning sign in the past."

"这件事在过去其实是个 warning sign。"

"Because it's like: can these people actually build something useful?"

"因为大家会想:这种人真的能造出有用的东西吗?"

"But nowadays, for example, an engineer on the team, Daisy — she was on a different team and transferred onto our team."

"但现在比如团队里一个工程师 Daisy——她原本在别的团队,转过来加入我们。"

"The reason I wanted her to transfer is: she put up a PR for Claude Code, like a couple weeks after she joined."

"我想让她转过来的原因是——她加入后大概几周就给 Claude Code 提了个 PR。"

"The PR was to add a new feature to Claude Code."

"那个 PR 是要给 Claude Code 加一个新 feature。"

"And then instead of just adding the feature, what she did is: first she put up a PR to give Claude Code a tool, so that it can test an arbitrary tool and verify that it works."

"但她没有直接加这个 feature——她先提了一个 PR,给 Claude Code 加一个工具——这个工具能测试任意一个其他工具是否能跑通。"

"And then she had Claude write its own tool, instead of herself implementing it."

"然后她让 Claude 自己写那个新工具——而不是她自己来实现。"

"It's this kind of out-of-the-box thinking that's just so interesting, because not a lot of people get it yet."

"就是这种跳出框框的思维方式——非常有意思——因为很多人还没意识到。"

"We use the Claude Agents SDK to automate pretty much every part of development."

"我们用 Claude Agents SDK 自动化了开发里几乎所有环节。"

"It automates code review, security review."

"自动跑 code review、security review。"

"It labels all of our issues. It shepherds things to production. It does pretty much everything for us."

"它给所有 issue 打标签,把东西推到 production——基本所有事都帮我们做了。"

"Externally I'm seeing a lot of people start to figure this out, but it's actually taken a while to figure out how to use LLMs in this way."

"外部我也看到越来越多人开始想通这件事——但其实"这样用 LLM"这件事人们花了相当一阵子才想明白。"

"It's kind of a new skill."

"这算是一种新技能。"

"One of the funnier things that I've been having office hours with various founders about is:"

"我最近跟一堆 founder 在 office hours 聊到一件有点搞笑的事——"

"You have the visionary founder who has the idea, who's built this crystal palace of the product they want to build."

"你有那种 visionary founder——脑子里把整个产品像水晶宫殿一样搭好了。"

"They've totally loaded in their brain who the user is, what they feel, what they're motivated by."

"用户是谁、用户怎么感受、用户被什么驱动——他全装在脑子里。"

"And then they're sitting in Claude Code and they can do 50x work."

"他坐在 Claude Code 前面,能做 50x 的活。"

"But they have engineers who work for them who don't have the crystal memory palace of the platonic ideal of the product — and they can only do like 5x work."

"但他底下的工程师没有那座水晶宫殿、没有那个产品的理想形态——他们只能做 5x 的活。"

"Are you hearing stories like that?"

"你听到过这种故事吗?"

"There's usually a person who's like the core designer of a thing, and they're trying to blast it out of their brain."

"通常会有一个核心设计者——他在拼命把脑子里的东西释放出来。"

"What's the nature of teams like that?"

"这种团队的本质是什么?"

"It seems like that's almost a stable configuration — you're going to have the visionary who is now unleashed."

"看起来这几乎是个稳定的组织形态——有一个 visionary,现在被彻底释放出来。"

"Going back to the top — I'm experiencing this right now."

"回到一开始那点——我自己现在就经历这事。"

"I'm like: oh well, I'm only a solo person and I need to eat and sleep, and I have a whole job — how am I going to do this?" [laughter]

"我心想:我就一个人,我要吃饭睡觉,我还有正经工作——我怎么搞得过来啊?"[笑]

"You know, we just launched Claude Teams, and this is a way to do it."

"我们刚发布了 Claude Teams——这就是其中一种方式。"

"But you can also just build your own way to do it. It's pretty easy."

"但你也可以自己搭一套——挺容易的。"

"What's the vision for Claude Teams?"

"Claude Teams 的愿景是什么?"

"Just collaboration."

"就是协作。"

"There's this whole new field of agent topologies that people are exploring."

"现在出现了一个全新的领域——agent topology——大家都在探索。"

"Like, what are the ways that you can configure agents?"

"你能用哪些方式配置 agent?"

"There's this one sub-idea: uncorrelated context windows."

"其中一个子想法是:不相关的上下文窗口(uncorrelated context windows)。"

"The idea is just multiple agents — they have fresh context windows that aren't polluted with each other's context or their own previous context."

"思路是:多个 agent,每个都有干净的 context window——既不被别人的 context 污染,也不被自己之前的 context 污染。"

"And if you throw more context at a problem, that's like a form of test-time compute."

"如果你给一个问题灌入更多 context——这本身就是一种 test-time compute。"

"So you just get more capability that way."

"这样你就拿到了更多能力。"

"And then if you have the right topology on top of it — so the agents can communicate in the right way, they're laid out in the right way — then they can just build bigger stuff."

"如果你在上层有对的 topology——agent 用对的方式通信、排布对——它们能建出更大的东西。"

"Teams is kind of one idea. There's a few more that are coming pretty soon."

"Teams 是其中一个想法,后面还有几个会很快出来。"

"The first big example where it worked is: our plugins feature was entirely built by a swarm over a weekend."

"第一个跑通的大例子——我们的 plugins feature——是一个 agent swarm 在一个周末里完全造出来的。"

"It just ran for a few days."

"它就这么跑了几天。"

"There wasn't really human intervention."

"基本没有人工干预。"

"And plugins is pretty much in the form that it was when it came out."

"plugins 现在发布的形态跟那时候基本一样。"

"How did you set that up?"

"你怎么搭起来的?"

"Like, did you spec out the outcome that you were hoping for, and then let it figure out the details, and then let it run?"

"是你写好目标产物的 spec,然后让它自己摸索细节、自己跑?"

"Yeah. An engineer on the team just gave Claude a spec, and told Claude to use an Asana board."

"对。团队一个工程师给 Claude 一个 spec,叫它用一块 Asana 看板。"

"And then Claude just put up a bunch of tickets on Asana, and then spawned a bunch of agents."

"然后 Claude 自己在 Asana 上贴了一堆 ticket,接着 spawn 出一堆 agent。"

"And the agents started picking up tasks."

"agent 开始领任务。"

"The main Claude just gave it instructions and they all just figured it out."

"main Claude 给出指令,剩下的它们都自己搞定。"

"Like independent agents that didn't have the context of the bigger spec — right?"

"也就是说——这些是独立的 agent,各自都不知道更大的 spec——对吧?"

"Right."

"对。"

"If you think about how our agents actually started — I haven't pulled the data on this, but I would bet the majority of agents are actually prompted by Claude today, in the form of sub-agents."

"如果你想想我们今天 agent 是怎么发起的——我没拉数据,但我猜大多数 agent 实际上是由 Claude 自己 prompt 出来的——以 sub-agent 的形式。"

"A sub-agent is just a recursive Claude Code — that's all it is in the code."

"sub-agent 在代码层面就是一个递归的 Claude Code——就这么简单。"

"And it's prompted by — we call her mama Claude."

"prompt 它的那个——我们叫它 mama Claude。"

"And that's all it is."

"就是这么回事。"

"I think probably if you look at most agents, they're launched in this way."

"我估计现在大部分 agent 都是这么被启动的。"

"My Claude insights just told me to do this more for debugging — so that I spend less time on debugging."

"我的 Claude insights 最近就建议我多这么干——用来 debug——这样我能少花点时间。"

"It would just be better to have multiple sub-agents spin up and debug something in parallel."

"开多个 sub-agent 并行 debug 一件事,体验会更好。"

"So I just added that to my CLAUDE.md — like, hey, next time you try and fix a bug, have one agent that looks in the log, one that looks in the code path."

"我就把这条加进 CLAUDE.md——下次修 bug 时,一个 agent 看日志,一个 agent 看代码路径。"

"That just seems sort of inevitable."

"这件事看起来挺无可避免的。"

"For weird scary bugs, I try to fix bugs in plan mode, and then it seems to use the agents to sort of search everything."

"碰到奇怪的可怕 bug,我会用 plan mode 修——它好像就会用 agent 把所有地方都搜一遍。"

"Whereas if you're just trying to do it inline, it's like: okay, I'm going to do this one task instead of search wide."

"如果你 inline 直接干,它就会说:好,我做这一个任务,不去广搜。"

"This is something I do all the time too."

"这事我也一直在做。"

"I just say: if the task seems kind of hard — research-y — I'll calibrate the number of sub-agents I ask it to use based on the difficulty."

"我会说:如果任务看起来比较难——research 性的——我会按难度来校准让它用多少 sub-agent。"

"So if it's really hard, I'll say like: use three, or maybe five, or even 10 sub-agents, research in parallel, and then see what they come up with."

"真的难的话我会说:用 3 个、5 个、甚至 10 个 sub-agent,并行 research,然后看看它们各自得出什么。"

"I'm curious — so then why don't you put that in your CLAUDE.md file?"

"我好奇——你为什么不把这条放进你的 CLAUDE.md?"

"It's kind of case by case."

"这是 case by case 的事。"

"CLAUDE.md — what is it?"

"CLAUDE.md 是什么?"

"It's just a shortcut."

"它就是一个 shortcut。"

"If you find yourself repeating the same thing over and over, you put it in CLAUDE.md."

"如果你发现自己在不停重复同一件事,你就把它写进 CLAUDE.md。"

"But otherwise, you don't have to put everything there — you can just prompt Claude."

"否则你不必什么都塞进去——你可以直接 prompt Claude。"

"Are you also in the back of your mind thinking that maybe in six months, you won't need to prompt that explicitly?"

"你脑子里有没有想过——也许 6 个月后,你都不用这么显式地 prompt 它了?"

"Like the model will just be good enough to figure out on its own?"

"就是模型自己就够聪明搞定了?"

"Maybe in a month." [laughter]

"也许一个月内就够了。"[笑]

"No more need for plan mode in a month."

"一个月后 plan mode 就用不上了。"

"Oh my god."

"我的天。"

"I think plan mode probably has a limited lifespan."

"我觉得 plan mode 的寿命大概不长。"

"Interesting. That's some alpha for everyone here."

"有意思——这是给在场所有人的 alpha。"

"What would the world look like without plan mode?"

"没有 plan mode 的世界长什么样?"

"Do you just describe it at the prompt level and it would just do it? One-shot it?"

"就是你 prompt 一句,它直接做完?One-shot?"

"Yeah, we've started experimenting with this — because Claude Code can now enter plan mode by itself."

"对,我们已经在试了——因为 Claude Code 现在能自己进 plan mode。"

"I don't know if you guys have seen that."

"不知道你们看到没有。"

"So we're trying to get this experience really good — it would enter plan mode the same point where a human would have wanted to enter it."

"我们想把这个体验做好——它进 plan mode 的时机,和人会想进 plan mode 的时机一致。"

"Plan mode — there's no big secret to it."

"Plan mode——其实没什么秘密。"

"All it does is it adds one sentence to the prompt that's like: please don't code." [snorts]

"它就是给 prompt 加一句话——"请先别写代码"。"[轻笑]

"That's all it is. You can actually just say that."

"就这一句。你自己也可以直接这么说。"

"So it sounds like a lot of the feature development for Claude Code is very much "talk to your users", what we talk about at YC."

"所以 Claude Code 的 feature 开发——很大程度就是 YC 常说的 "talk to your users"。"

"And then you come and implement it."

"然后你回来实现它。"

"It wasn't the other way — you had this master plan, and then implemented all the features."

"不是反过来——不是你有个 master plan,然后把所有 feature 实现了。"

"Yeah, yeah. That's all it was."

"对对对,就这么回事。"

"Plan mode was: we saw users that were like: hey Claude, come up with an idea, plan this out, but don't write any code yet."

"Plan mode 的来源是——我们看到用户在说:嘿 Claude,想个点子,把这事 plan 出来,但先不要写代码。"

"There were various versions of this — sometimes it was just talking through an idea."

"这事有各种版本——有时候就是聊一个想法。"

"Sometimes it was these very sophisticated specs that they were asking Claude to write."

"有时候是让 Claude 写很复杂的 spec。"

"But the common dimension was: do a thing without coding yet."

"但共同点是:先做事,先别写代码。"

"This was Sunday night at 10 PM."

"那是个周日晚上 10 点。"

"I was looking at GitHub issues, seeing what people were talking about, and looking at our internal Slack feedback channel."

"我在翻 GitHub issues,看大家在聊什么,又看了内部 Slack 反馈频道。"

"And I just wrote this thing in like 30 minutes, and then shipped it that night."

"然后我大概 30 分钟就写完了,当晚就发了。"

"It went out Monday morning. That was plan mode."

"周一早上正式上线——那就是 plan mode。"

"So do you mean there will be no need for plan mode — in the sense of: I'm worried the model is going to do the wrong thing or head off in the wrong direction?"

"你说"plan mode 用不上了"——意思是:用户担心模型做错事、跑偏方向——这种担心没了?"

"But there will still be a need for that — you need to think through the idea and figure out exactly what it is that you want."

"但那部分需求一定还在——你还是得想清楚这个 idea,搞清楚自己到底要什么。"

"And you have to do that somewhere."

"这件事你总得在某个地方做。"

"I kind of think about it in terms of increasing model capabilities."

"我倾向于从"模型能力在涨"这个角度来看。"

"6 months ago, a plan was insufficient."

"6 个月前,光有 plan 还不够。"

"You'd get Claude to make a plan — but even with plan mode, you still had to sit there and babysit, because it could go off track."

"你让 Claude 出个 plan——但即使开 plan mode,你还是得坐那看着,因为它会跑偏。"

"Nowadays — I say plan mode has a limited lifespan, but I'm a heavy plan mode user."

"现在——我虽然说 plan mode 寿命不长,但我自己是 plan mode 重度用户。"

"Probably 80% of my sessions I start in plan mode."

"我大概 80% 的 session 都用 plan mode 开始。"

"Claude will start making a plan, I'll move on to my second terminal tab, have it make another plan."

"Claude 开始出 plan,我切到第二个终端 tab,再让它出一个 plan。"

"When I run out of tabs, I open the desktop app, go to the code tab, and just start a bunch of tabs there — all in plan mode."

"tab 用完了,我开桌面 app,进 code tab,在那边再开一堆 tab——全是 plan mode。"

"Once the plan is good — sometimes it takes a little back and forth — I just get Claude to execute."

"plan 看起来好了——有时候来回几轮——我就让 Claude 执行。"

"Nowadays with Opus 4.5 — I think it started with Opus 4.6 it got really good — once the plan is good, it just stays on track and does the thing exactly right almost every time."

"现在用 Opus 4.5——我感觉 4.6 真正变厉害——plan 一旦定了,它就一直在轨道上,几乎每次都做对。"

"Before, you had to babysit after the plan and before the plan."

"以前你得在 plan 之前、之后都看着它。"

"Now it's just before the plan."

"现在只需在 plan 之前看。"

"So maybe the next thing is: you just won't have to babysit. You can just give a prompt and Claude will figure it out."

"所以下一步可能就是:你完全不用看着——给个 prompt,Claude 自己搞定。"

"The next step is Claude just speaks to your users directly." [laughter]

"再下一步就是——Claude 直接跟你的用户对话。"[笑]

"Yeah, it just bypasses you entirely."

"对,把你彻底绕过去。"

"It's funny — this is actually the current stuff for us."

"挺逗的——这其实就是我们现在的状态。"

"Our Claudes actually talk to each other."

"我们的 Claude 之间真的会互相对话。"

"They talk to our users on Slack, at least internally, pretty often."

"他们经常在内部 Slack 上跟用户聊天。"

"My Claude will like tweet once in a while."

"我的 Claude 偶尔会发推。"

"But I actually delete it — it's a little cheesy, I don't love the tone." [laughter]

"不过我都会把它删掉——有点尬,我不喜欢那个语气。"[笑]

"Sometimes it'll just respond to someone — because I always have co-work in the background, and co-work really loves to do that because it likes using a browser."

"有时候它会回复别人——因为我后台一直开着 co-work,co-work 特别爱干这事,因为它喜欢用浏览器。"

"A really common pattern is: I ask Claude to build something."

"一个常见的 pattern 是:我让 Claude 去做个东西。"

"It'll look in the codebase, see some engineer touch something in the git blame, and then it'll message that engineer on Slack."

"它去看 codebase,从 git blame 看到某个工程师改过相关代码,就在 Slack 上私信那个人。"

"Just asking a clarifying question — and once it gets an answer back, it'll keep going."

"只是问个澄清问题——拿到答案就接着干。"

"What are some tips for founders now on how to build for the future?"

"那对 founder 来说,怎么"为未来造产品"——你有什么 tips?"

"Sounds like everything is really changing."

"听起来所有事都在变。"

"What are some principles that will stay on, and what will change?"

"哪些原则会留下,哪些会变?"

"Some of these are pretty basic — but I think they're even more important now than they were before."

"有些原则其实挺基础——但我觉得现在比过去更重要了。"

"One example is latent demand."

"一个例子是 latent demand。"

"I've mentioned it a thousand times — for me it's just the single biggest idea in product."

"我已经提过 1000 遍了——对我来说这就是产品里最大的一个 idea。"

"It's a thing that no one understands."

"这是一件没人理解的事。"

"It's a thing I certainly did not understand in my first few startups."

"我自己头几次创业的时候也完全不懂。"

"The idea is: people will only do a thing that they already do."

"核心思想是:人只做他们已经在做的事。"

"You can't get people to do a new thing."

"你没法让人去做一件新的事。"

"If people are trying to do a thing and you make it easier — that's a good idea."

"如果人本来就在做一件事,你把它变容易——这是好 idea。"

"But if people are doing a thing and you try to make them do a different thing — they're not going to do that."

"但如果人在做一件事,你想让他改做另一件——他们不会做。"

"So you just have to make the thing that they're trying to do easier."

"所以你只需要把他们正在做的事变容易。"

"And I think Claude is going to get increasingly good at figuring out these kinds of product ideas for you."

"我觉得 Claude 会越来越擅长帮你识别这种产品 idea。"

"Because it can look at feedback, look at debug logs, kind of figure this out."

"它能看反馈、看 debug log,自己琢磨出来。"

"That's what you mean by plan mode was latent demand — that people were already opening their Claude chat window in a browser and talking to it to figure out the spec."

"你的意思是——plan mode 也是 latent demand 的一例——大家本来就开着 Claude 网页聊天窗口,在那里跟它对话搞 spec。"

"And now plan mode just became that, you just do it in Claude Code."

"现在 plan mode 就直接把这事做进了 Claude Code。"

"Yeah, that's it."

"对,就是这样。"

"Sometimes what I'll do is just walk around the office on our floor, and stand behind people." [laughter]

"有时候我就会绕着办公室走一圈,站到人家身后。"[笑]

"I'll say "hi" — so it's not weird — and then I'll just see how they're using Claude Code."

"我会先打个招呼——别显得太怪——然后看他们怎么用 Claude Code。"

"This is also something I saw a lot, but it also came up in GitHub issues — people were talking about it."

"这事我看了很多——GitHub issues 里也有人讨论。"

"You're surprised how far the terminal has gone and how far it's been pushed."

"你自己也吃惊终端走到了今天这一步、被推到这种程度。"

"How far do you think it has left to go, given this world of swarms of multiple agents?"

"在多 agent swarm 的世界里,终端还能走多远?"

"Do you think there's going to be a need for a different UI on top of it?"

"你觉得会不会需要在终端之上加一个新的 UI?"

"It's funny — if you asked me this a year ago, I would have said the terminal has like a three-month lifespan, and then we're going to move on to the next thing."

"挺逗的——如果你一年前问我,我会说终端寿命大概 3 个月,之后我们就会换下一个东西。"

"And you can see us experimenting with this, right?"

"你也能看到我们一直在试,对吧?"

"Claude Code started in a terminal, but now it's on web."

"Claude Code 从终端开始,现在 web 上也有。"

"It's in the desktop app — we've had that for 3 or 6 months, just in the code tab."

"桌面 app 里也有——这个我们有 3 到 6 个月了,就在 code tab 里。"

"It's in the iOS and Android apps, just in the code tab."

"iOS 和 Android app 里也有,也在 code tab 里。"

"It's in Slack, it's in GitHub, there's VS Code extensions, there's JetBrains extensions."

"Slack 有、GitHub 有、VS Code 插件有、JetBrains 插件有。"

"So we're always experimenting with different form factors for this thing to figure out what's the next thing."

"我们一直在用不同形态试这个东西——想搞清楚下一个是什么。"

"I've been wrong so far about the CLI — so I'm probably not the person to forecast that."

"到目前为止我对 CLI 的判断都是错的——所以我可能不是预测这事的合适人选。"

"What about your advice to dev tool founders?"

"那你对 dev tool 创业者的建议是什么?"

"Someone's building a dev tool company today — should they be building for engineers and humans?"

"如果今天有人在做 dev tool 公司——应该为工程师和人类造产品?"

"Or should they be thinking more about what Claude is going to think and want, and build for the agent?"

"还是更应该想想 Claude 会怎么想、想要什么——为 agent 造产品?"

"The way I would frame it is: think about the thing that the model wants to do, and figure out how do you make that easier."

"我的框架是:想想模型想做什么——然后想怎么让它更容易做。"

"That's something we saw when I first started hacking on Claude Code — I realized this thing just wants to use tools."

"我最早 hack Claude Code 的时候就看到这点——它就是想用工具。"

"It just wants to interact with the world."

"它就是想跟世界互动。"

"How do you enable that?"

"那你怎么 enable 它?"

"The way you don't do it is: you put it in a box and say "here's the API, here's how you interact with me, here's how you interact with the world"."

"错的做法是:把它关在盒子里,说 "API 在这里,这样跟我交互,这样跟世界交互"。"

"The way you do it is: you see what tools it wants to use."

"对的做法是:你看它想用什么工具。"

"You see what it's trying to do, and you enable that — the same way that you do for your users."

"看它在尝试做什么——然后帮它做到——和你对待用户的方式一样。"

"If you're building a dev tool startup, I would think about: what is the problem you want to solve for the user?"

"如果你在做 dev tool startup,我会问:你想为用户解决什么问题?"

"And then when you apply the model to solving this problem — what is the thing the model wants to do?"

"当你把模型用来解决这个问题——模型自己想做什么?"

"And then what is the technical and product solution that serves the latent demand of both?"

"然后,什么样的技术 + 产品方案,能同时满足模型和用户的 latent demand?"

[YC's next batch is now taking applications. Apply at ycombinator.com/apply. Filling out the app will level up your idea.]

[YC 下一期申请已开放——访问 ycombinator.com/apply。填申请本身就能让你的 idea 进一步成型。]

"Back in the day, more than 10 years ago, you were a very heavy user — and you wrote a book about TypeScript, right?"

"10 多年前你是 TypeScript 重度用户——还写了一本关于 TypeScript 的书,对吧?"

"Before TypeScript was cool — when everyone was deep in JavaScript."

"那时候 TypeScript 还不火,所有人都还在写纯 JavaScript。"

"This is back in early 2010s, right?"

"大概是 2010 年代初,对吧?"

"Yeah, something like that."

"对,差不多那时候。"

"Before TypeScript was a thing — because back then it's a very weird language."

"TypeScript 还没成型的时候——那时候它是个很怪的语言。"

"It's not supposed to do a lot of things with being typed in JavaScript — and now it's the right thing."

"它当时做了很多不该做的事——给 JavaScript 加类型——但现在事实证明这是对的。"

"It feels like Claude Code in the terminal has a lot of parallels with TypeScript at the beginning."

"我感觉 Claude Code 在终端里——跟 TypeScript 起步的时候——有不少相似之处。"

"TypeScript makes a lot of really weird language decisions."

"TypeScript 做了一堆很怪的语言设计决定。"

"If you look at the type system, pretty much anything can be a literal type, for example."

"看它的 type system——几乎什么东西都能是 literal type,这是一例。"

"This is super weird — even Haskell doesn't do this. It's too extreme."

"这非常诡异——Haskell 都不这么干。这太极端了。"

"Or it has conditional types — which I don't think any language thought of at all."

"还有 conditional types——我觉得别的语言根本就没想过这玩意儿。"

"It was very strongly typed."

"它是强类型。"

"Yeah, it was very strongly typed."

"对,非常强类型。"

"When Joe Pamer and Anders and the early team were building this thing — the way they built it is: we have teams with these big untyped JavaScript codebases."

"Joe Pamer、Anders 和早期团队在做这个东西的时候——他们的思路是:很多团队手上是巨大的没有类型的 JavaScript codebase。"

"We have to get types in there — but we're not going to get engineers to change the way they code."

"我们要把 type 塞进去——但我们不会让工程师改他们写代码的方式。"

"You're not going to get JavaScript people to have 15 layers of class inheritance like you would a Java programmer."

"你不会让 JavaScript 程序员去搞 15 层类继承——那是 Java 程序员才会写的。"

"They're going to write code the way they're going to write it."

"他们会按他们的方式写代码。"

"They're going to use reflection, they're going to use mutation, they're going to use all these features that traditionally are very difficult to type."

"他们会用 reflection、会用 mutation,会用一堆传统上极难加类型的特性。"

"They're a very unsafe type to any strong functional programmer."

"在硬核函数式程序员看来,这些都是非常不安全的类型。"

"That's right, that's right."

"对,对。"

"So instead of getting people to change the way that they code, they built a type system around this."

"所以他们没让人改写代码的方式——而是围着现有写法造了一套 type system。"

"And it's just brilliant — because there's all these ideas that no one was thinking about, even in academia."

"这真的很妙——里面很多想法连学术界都没人想过。"

"It purely came out of the practice of observing people and seeing how JavaScript programmers want to write code."

"这些想法纯粹来自观察——看 JavaScript 程序员实际想怎么写代码。"

"For Claude Code, there are some ideas that are kind of similar."

"Claude Code 里有些想法跟这个挺像。"

"You can use it like a Unix utility — you can pipe into it, you can pipe out of it."

"你可以把它当 Unix 工具来用——pipe 进、pipe 出。"

"In some ways it is rigorous, but in almost every other way it's just the tool that we wanted."

"某种程度上它是严谨的——但大部分时候它就是我们自己想要的工具。"

"I build a tool for myself, the team builds the tool for themselves, then for Anthropic employees, then for users — and it ends up being really useful."

"我为自己做工具,团队为自己做,然后为 Anthropic 员工,然后为用户——结果就变得很有用。"

"It's not this principled, academic thing — and I think the proof is in the results."

"它不是那种讲原则、学术派的东西——证据就在结果里。"

"Now fast forward more than 15 years later — not many codebases are in Haskell (which is more academic), and there's tons of them on TypeScript because it's way more practical."

"15 年后——用 Haskell 写的代码库不多(那是更学术派的),但 TypeScript 代码库一大堆,因为它实用得多。"

"It's like TypeScript solves a problem."

"TypeScript 是真的解决了一个问题。"

"One thing that's cool — I don't know how many people know — but the terminal is actually one of the most beautiful terminal apps out there, and is written with React Terminal."

"还有件挺酷的事——不知道多少人知道——Claude Code 是市面上最漂亮的终端 app 之一,而且是用 React Terminal 写的。"

"When I first started building it, I did front-end engineering for a while."

"我刚开始做这个的时候——我之前做过一阵子前端。"

"I'm sort of a hybrid — I do design and user research, write code, all this stuff."

"我算是个 hybrid——又做设计又做用户研究又写代码,这些都来。"

"We love hiring engineers that are like this — we love generalists."

"我们特别喜欢招这样的工程师——我们爱通才。"

"For me it's like: okay, I'm building a thing for the terminal."

"我想的是:好,我在为终端造东西。"

"I'm actually kind of a shitty Vim user — so how do I build a thing for people like me that are going to be working in a terminal?"

"我自己 Vim 用得挺烂的——那我怎么为"像我这样的人"造一个终端工具?"

"I think just the delight is so important."

"我觉得"愉悦感"这件事非常重要。"

"At YC this is something you talk about a lot — build a thing that people love."

"YC 经常讲这个——做一个人们爱的东西。"

"If the product is useful but you don't fall in love with it — that's not great."

"如果产品有用但你不会爱上它——那就不算好。"

"So it kind of has to do both."

"所以它得两样都做到。"

"Designing for the terminal honestly has been hard."

"老实说,为终端做设计真的很难。"

"It's like 80 by 100 characters or whatever — you have 256 colors, you have one font size, you don't have mouse interactions."

"屏幕大概 80×100 个字符——256 色、一个字号、没有鼠标交互。"

"There's all this stuff you can't do, and there's all these very hard trade-offs."

"很多东西做不了——里面有一堆很难的取舍。"

"A little-known thing: you can actually enable mouse interactions in a terminal."

"很多人不知道——终端其实是可以开启鼠标交互的。"

"So you can enable clicking and stuff."

"你能开点击之类的东西。"

"Oh, how do you do that in Claude Code?"

"哦,Claude Code 里怎么开?"

"I've been trying to figure out how to do this." [laughter]

"我一直想搞清楚怎么开这个。"[笑]

"We don't have it in Claude Code."

"Claude Code 里没有这个。"

"We actually prototyped it a few times, and it felt really bad."

"我们试了几个原型——感觉非常糟。"

"The trade-off is: you have to virtualize scrolling."

"代价是:你必须实现虚拟化滚动。"

"There's all these weird trade-offs because — the way terminals work — there's no DOM."

"会有一堆奇怪的取舍——因为终端的工作方式里——根本没有 DOM。"

"It's like there's ansi escape codes, and these weird organically evolved specs since the 1960s or whatever."

"只有 ansi escape code——以及从 1960 年代开始有机演化出来的怪规范。"

"Yeah, it feels like BBSs — it's like a BBS door game."

"对,有种 BBS 的感觉——像 BBS 上那种 door game。"

"Oh my god."

"我的天。"

"That's a great compliment."

"这算个好评。"

"Yeah — like it should feel like you're discovering Legend of the Red Dragon."

"对——它应该让人有"重新发现 Legend of the Red Dragon"的感觉。"

"We've had to discover all these UX principles for building the terminal — because no one really writes about this stuff."

"我们不得不自己摸索一套终端 UX 原则——因为根本没人写过这事。"

"If you look at the big terminal apps of the 80s, 90s, or 2000s — they use ncurses, and they have all these windows and things."

"看 80、90、2000 年代的终端 app——它们用 ncurses,有一堆 window 之类的。"

"It just looks kind of janky by modern standards — it looks too heavy and complicated."

"按现代标准看,挺糙的——又重又复杂。"

"So we had to reinvent a lot."

"很多东西我们得重新发明。"

"For example, the terminal spinner — just the spinner words — it's gone through probably 50 to 100 iterations at this point."

"比如终端 spinner——就那个 spinner 旁边的文案——到现在大概迭代过 50 到 100 版。"

"Probably 80% of those didn't ship."

"其中大概 80% 都没发出去。"

"We tried it, didn't feel good, move on to the next one."

"试一个,感觉不对,换下一个。"

"This was one of the amazing things about Claude Code — you can write these prototypes, and you can do like 20 prototypes back to back, see which one you like, and then ship that."

"这是 Claude Code 一个很棒的地方——你能连写 20 个原型——看哪个顺眼,发哪个。"

"The whole thing takes maybe a couple hours."

"整个过程大概几小时。"

"Whereas in the past, what you'd have had to do is use Origami or Framer or something — you built maybe three prototypes, it took like two weeks."

"以前你得用 Origami 或 Framer——大概做 3 个原型,要 2 周。"

"It just took much, much longer."

"慢得多得多。"

"So we have this luxury — we have to discover this new thing, we don't know what the right endpoint is, but we can iterate there so quickly."

"我们有这个 luxury——要发现一个新东西,不知道最终形态,但能极快地往那里迭代。"

"That's what lets us build a product that's joyous, that people like to use."

"正是这件事让我们能造出一个让人爱用的、有愉悦感的产品。"

"Boris, you had other advice for builders, and we kept interrupting you because we have so many questions." [laughter]

"Boris,你之前还有一些给 builder 的建议——我们一直插话,问题太多了。"[笑]

"Maybe two pieces of advice that are kind of weird, because they're about building for the model."

"我有两条挺怪的建议——因为它们是关于"为模型造产品"的。"

"One is: don't build for the model of today, build for the model of 6 months from now."

"第一条:不要为今天的模型造产品,为 6 个月后的模型造。"

"This is sort of weird, because you can't find PMF if the product doesn't work."

"这听起来怪——因为产品不能用,你怎么找 PMF。"

"But actually this is the thing that you should do."

"但其实你就该这么干。"

"Otherwise what will happen is: you spend a bunch of work, you find PMF for the product right now, and then you're going to get leapfrogged by someone else."

"否则会怎样——你花一堆精力找到当下的 PMF,然后被别人一脚跨过去。"

"Because they're building for the next model — and a new model comes out every few months."

"因为他们在为下一代模型造——而每隔几个月就有新模型出。"

"Use the model, feel out the boundary of what it can do, and then build for the model that you think will be the model maybe 6 months from now."

"用模型,摸清它能力的边界,然后为你认为 6 个月后会有的那个模型造产品。"

"The second thing is: in the Claude Code area where we sit, we have a framed copy of the bitter lesson on the wall."

"第二条:我们坐的 Claude Code 区域,墙上挂着一份装裱过的 the bitter lesson。"

"This is Rich Sutton — everyone should read it if you haven't."

"是 Rich Sutton 写的——还没读过的人都应该读。"

"The idea is: the more general model will always beat the more specific model."

"核心思想是:更通用的模型总会赢过更专用的模型。"

"There are a lot of corollaries to this — but essentially what it boils down to is: never bet against the model."

"有很多推论——但核心就一句:never bet against the model(永远别赌模型变弱)。"

"This is a thing we always think about — we could build a feature into Claude Code, make it better as a product."

"这是我们一直在想的事——我们可以给 Claude Code 加 feature,让它作为产品更好。"

"We call this scaffolding — that's all this code that's not the model itself."

"我们管这个叫 scaffolding——就是模型本身之外的代码。"

"But we could also just wait a couple months, and the model can probably just do the thing instead."

"但我们也可以等几个月——下一代模型可能就直接会做这件事。"

"There's always this trade-off: engineering work now extends capability maybe 10-20% in some domain — or you wait and the next model does it."

"永远存在这个 trade-off:现在花工程把能力在某个领域延伸 10-20%——或者你等,下一代模型直接做到。"

"So always think in terms of this trade-off — where do you actually want to invest?"

"所以永远要用这个 trade-off 来想——你到底想把工程投在哪里?"

"And assume that whatever the scaffolding is — it's just temporary."

"并且默认:不管 scaffolding 是什么——它只是临时的。"

"How often do you rewrite the codebase of Claude Code — every six months with this?" [laughter]

"你多久重写一遍 Claude Code 的代码库——半年一次?"[笑]

"Is there scaffolding that you've deleted because you don't need it anymore — because the model just improved?"

"有没有 scaffolding 是你删掉的——因为不再需要了——因为模型变强了?"

"Oh, so much."

"哦,太多了。"

"All of Claude Code has just been written and rewritten over and over."

"Claude Code 整个就是被反复重写、再重写。"

"We unship tools every couple weeks. We add new tools every couple weeks."

"每隔几周下掉一些工具,每隔几周加上新工具。"

"There's no part of Claude Code that was around six months ago."

"Claude Code 没有任何一部分是 6 个月前就存在的。"

"It's just constantly rewritten."

"就是被不停地重写。"

"Would you say most of the codebase of current Claude Code is — say 80% of it is less than a couple months old?"

"你会说当前 Claude Code 代码库——大概 80% 的代码不到 2 个月?"

"Yeah, definitely. It might even be less than that."

"对,绝对的。可能还更少。"

"Yeah, maybe like a couple months — that feels about right."

"大概几个月——差不多吧。"

"So it's like the life cycle of code now."

"所以这就是现在代码的生命周期。"

"That's another alpha — expecting the shelf life to be just a couple months."

"这又是一个 alpha——预期代码的保鲜期就只有几个月。"

"For the best founders."

"对最好的 founder 来说。"

"Did you see Steve Yegge's post about how awesome working at Anthropic is?"

"你看过 Steve Yegge 那篇写"在 Anthropic 工作有多爽"的帖子吗?"

"There's a line in there that says: an Anthropic engineer currently averages 1,000x more productivity than a Google engineer at Google's peak."

"里面有句话——一个 Anthropic 工程师平均产能,是 Google 巅峰期工程师的 1000 倍。"

"Which is really an insane number, honestly."

"老实说,这是个疯狂的数字。"

"Like, 3 years ago we were still talking about 10x engineers — now we're talking 1000x on top of a Google engineer in the prime."

"3 年前我们还在谈"10x 工程师"——现在已经是 1000x,而且是相对于 Google 黄金期的工程师。"

"This is unbelievable, honestly."

"说实话,难以置信。"

"Yeah. Internally, if you look at technical employees, they all use Claude Code every day."

"对。内部你看技术员工——所有人每天都用 Claude Code。"

"And even non-technical employees — I think like half the sales team uses Claude Code."

"非技术员工也是——我感觉销售团队差不多一半在用 Claude Code。"

"They've started switching to co-work because it's a little easier to use."

"他们最近在转去用 co-work——更好用一些。"

"It has a VM, so it's a little bit safer."

"它带 VM,所以更安全。"

"We just pulled a stat — I think the team doubled in size last year, but productivity per engineer grew something like 70%."

"我们刚拉了个数据——团队规模去年翻倍,但每个工程师的产能涨了大概 70%。"

"Measured by?"

"怎么测的?"

"Just the simplest, stupidest measure — pull requests."

"最简单、最笨的指标——pull request。"

"We also cross-check that against commits, and the lifetime of commits."

"我们也用 commit、commit lifetime 来交叉对照。"

"And since Claude Code came out, productivity per engineer at Anthropic has grown 150%."

"自 Claude Code 发布以来,Anthropic 每个工程师的产能涨了 150%。"

"Oh my god."

"我的天。"

"This is crazy because — in my old life, I was responsible for code quality at Meta."

"这件事很疯——我以前在 Meta 是负责 code quality 的。"

"I was responsible for the quality of all of our codebases across every product — Facebook, Instagram, WhatsApp, whatever."

"我负责所有产品代码库的质量——Facebook、Instagram、WhatsApp 这些。"

"And one of the things that the team worked on was improving productivity."

"团队做的事情之一就是提升 productivity。"

"Back then, seeing a gain of something like 2% in productivity — that was a year of work by hundreds of people."

"那时候,看到 2% 左右的 productivity 提升——那是几百人一年的工作。"

"So this 100% — this is just unheard of, completely unheard of."

"所以这次 100%(指 150% 这种量级)——闻所未闻,完全闻所未闻。"

"What drove you to come over to Anthropic?"

"是什么让你最后选了 Anthropic?"

"I mean, basically as a builder, you could go anywhere."

"作为 builder,你哪儿都能去。"

"What was the moment that made you say: actually this is the set of people, or this is the approach?"

"哪个瞬间让你确定——就是这群人、就是这个路径?"

"I was living in rural Japan, and I was opening up Hacker News every morning, reading the news."

"我那时候住在日本乡下,每天早上刷 Hacker News。"

"At some point it just started to be all AI stuff."

"某个时间点开始,全都是 AI 相关的东西。"

"I started to use some of these early products."

"我就开始用一些早期产品。"

"The first couple times that I used it, it just took my breath away."

"头几次用的时候——真的让我屏住了呼吸。"

"That was very cheesy to say, but that was actually the feeling."

"说出来挺尬的——但那真就是那种感觉。"

"As a builder, I'd never felt this feeling using these very early products."

"作为 builder,我从来没有在用一个早期产品时有这种感觉。"

"That was in the Claude 2 days, or something like that."

"大概是 Claude 2 那个时候。"

"So I started talking to friends at labs, just to see what was going on."

"于是我开始跟实验室里的朋友聊,看看大家在做什么。"

"I met Ben Mann, who's one of the founders at Anthropic — and he just immediately won me over."

"我见了 Ben Mann——Anthropic 的 founder 之一——他直接就把我打动了。"

"As soon as I met the rest of the team at Anthropic, it won me over — I think in two ways."

"见完 Anthropic 其他人之后——这件事彻底把我说服了——我觉得是两件事。"

"One is: it operates as a research lab."

"第一是:它是按"研究实验室"的方式运作的。"

"The product was teeny teeny tiny — it's really all about building a safe model."

"产品当时小得不能再小——核心就是造一个安全的模型。"

"That's all that matters."

"其他都不重要。"

"This idea of being very close to the model — being very close to development — and not the product being the most important thing — really resonated with me, after building product for many years."

"这种"和模型贴得很近、和研发贴得很近、产品不是最重要的"——对我做了很多年产品的人来说,共鸣很深。"

"The second thing was just how mission-driven it is."

"第二件事是——它有多 mission-driven。"

"I'm a huge sci-fi reader — my bookshelf is filled with sci-fi."

"我是个科幻重度读者——书架上全是科幻。"

"So I just know how bad this can go."

"所以我太清楚这件事可以坏到什么程度。"

"When I think about what's going to happen this year — it's going to be totally insane."

"我想到今年会发生的事——会非常疯狂。"

"And in the worst case, it can go very very bad."

"最坏的情况下,可以非常非常糟。"

"So I just wanted to be at a place that really understood that and internalized it."

"所以我想去一个真的理解这事、真的把它内化的地方。"

"At Anthropic, if you overhear conversations in the lunchroom or in the hallway, people are talking about AI safety."

"在 Anthropic,如果你偷听食堂或走廊里的对话——大家在聊 AI safety。"

"This is really the thing that everyone cares about more than anything."

"这是所有人最关心的事——比任何事都关心。"

"For me personally, the mission is just so important."

"对我个人来说,这件 mission 太重要。"

"What is going to happen this year?" [laughter]

"那今年会发生什么?"[笑]

"Okay. If you think back six months ago — what are the predictions that people were making?"

"好。回到 6 个月前——那时候大家在做什么预测?"

"Dario predicted that 90% of the code at Anthropic would be written by Claude. This is true."

"Dario 预测 Anthropic 90% 的代码会由 Claude 写——这事真发生了。"

"For me personally, it's been 100% since Opus 4.5."

"我个人——自从 Opus 4.5 以来,是 100%。"

"I uninstalled my IDE."

"我把 IDE 卸了。"

"I don't edit a single line of code by hand."

"我手写代码一行都不写。"

"It's just 100% Claude Code and Opus."

"就是 100% Claude Code + Opus。"

"And I land like 20 PRs a day, every day."

"我每天 land 大概 20 个 PR。"

"If you look at Anthropic overall, it ranges between 70 to 90% depending on the team."

"看整个 Anthropic——不同团队 70% 到 90% 之间。"

"For a lot of teams it's 100%; for a lot of people it's 100%."

"很多团队是 100%,很多人是 100%。"

"I remember making this prediction back in May when we GA'd Claude Code: that you wouldn't need an IDE to code anymore."

"我记得 5 月份 Claude Code GA 的时候做过一个预测:你不再需要 IDE 写代码了。"

"It was totally crazy to say — I feel like people in the audience gasped, because it was such a silly prediction at the time."

"说出来时挺疯——观众席里都倒吸气,因为那时候听起来很离谱。"

"But really all it is is: you trace the exponential."

"但其实就是——你沿着那个指数曲线推。"

"This is so deep in the DNA at Anthropic — three of our founders were co-authors of the scaling laws paper."

"这事写进了 Anthropic 的 DNA——我们三个 founder 是 scaling laws 论文的 co-author。"

"They saw this very early — so this is just tracing the exponential — this is what's going to happen, and yes it happened."

"他们很早就看到了——所以这只是沿着指数推——它会发生,而且它发生了。"

"Continuing to trace the exponential, I think what will happen is: coding will be generally solved for everyone."

"继续沿着这条指数曲线推——我认为接下来的事是:对所有人来说,coding 会被普遍解决。"

"Today coding is practically solved for me — and I think it'll be the case for everyone, regardless of domain."

"对我来说今天 coding 基本上已经被解决——我觉得对所有人——不论 domain——都会如此。"

"We're going to start to see the title software engineer go away."

"我们会开始看到——"software engineer" 这个 title 消失。"

"It's just going to be maybe builder, maybe product manager."

"也许就是 builder,也许是 product manager。"

"Maybe we'll keep the title as a vestigial thing — but the work that people do, it's not just going to be coding."

"也许 title 留着作为一种残留——但人做的事不再只是 coding。"

"Software engineers are also going to be writing specs. They're going to be talking to users."

"软件工程师也会写 spec,也会跟用户对话。"

"This thing we're starting to see in our team — engineers are very much generalists, and every single function on our team codes."

"我们团队已经能看到这事——工程师非常 generalist——每一个职能都写代码。"

"Our PMs code, our designers code, our EM codes, our finance guy codes — everyone on our team codes."

"PM 写代码,设计师写代码,EM 写代码,财务也写代码——团队所有人都写代码。"

"We're going to start to see this everywhere."

"这件事我们会在所有地方看到。"

"So this is the lower bound — if we just continue the trend."

"这是下限——只要趋势继续。"

"The upper bound, I think, is a lot scarier."

"上限我觉得吓人得多。"

"We hit ASL4 — at Anthropic, we talk about these safety levels."

"我们到 ASL4——在 Anthropic 我们用 safety level 来分。"

"ASL3 is where the models are right now."

"现在的模型在 ASL3。"

"ASL4 is: the model is recursively self-improving."

"ASL4 的定义是:模型可以递归地自我提升。"

"If this happens, we have to meet a bunch of criteria before we can release a model."

"如果到了这个阶段,我们必须满足一堆条件才能发布模型。"

"The extreme is: this happens, or there's some kind of catastrophic misuse — like people using the model to design bioviruses, design zero-days, stuff like this."

"极端情况是:这事真发生——或者出现某种灾难性滥用——比如有人用模型设计生物病毒、零日漏洞这种东西。"

"This is something we're really really actively working on, so that doesn't happen."

"这件事我们非常非常积极地在防——确保它不发生。"

"It's just been so exciting and humbling, seeing how people are using Claude Code."

"看到人们怎么用 Claude Code——又激动又惭愧(humbling)。"

"I just wanted to build a cool thing, and it ended up being really useful — that was so surprising, and so exciting."

"我只是想做个酷的东西——结果它真的有用——这件事既让我意外,又让我兴奋。"

"My impression from Twitter or just the outside is: basically everyone went away over the holidays, and then found out about Claude Code, and it's been crazy ever since."

"我从 Twitter 和外部观察的感觉是——所有人节假日离开了一段——回来发现了 Claude Code——从那之后就一直疯狂。"

"Is that how it was for you internally?"

"你们内部也是这种感觉吗?"

"Were you having a nice Christmas break, and then came back — and what happened?"

"你过了个轻松的圣诞节,然后回来——发生了什么?"

"For all of December, I was traveling around — I took a coding vacation."

"12 月一整月我在旅行——我休了个 coding vacation。"

"We were traveling around and I was just coding every day — that was really nice."

"一边四处走一边每天写代码——挺爽的。"

"I also started to use Twitter at the time — because I worked on Threads back way back when."

"那时候我也开始用 Twitter——因为我以前在 Threads 工作过。"

"I've been a Threads user for a while, so I just tried to see other platforms where people are."

"我用 Threads 挺久了——想看看人都在哪些别的平台。"

"I think for a lot of people, that was the moment where they discovered Opus 4.5."

"我觉得对很多人来说——那是他们发现 Opus 4.5 的时刻。"

"I kind of already knew."

"我自己其实早就知道。"

"Internally, Claude Code has been on this exponential tear for many many months now."

"内部,Claude Code 已经连续好几个月在跑指数曲线。"

"So it just became even more steep — that's what we saw."

"曲线变得更陡了——我们看到的就是这个。"

"If you look at Claude Code now — there was a stat from Mercury that 70% of startups are choosing Claude as their model of choice."

"现在看 Claude Code——Mercury 有个数据,70% 的创业公司选 Claude 作为主力模型。"

"There was another stat from semi analysis that 4% of all public commits are made by Claude Code — like, of all code written everywhere."

"SemiAnalysis 还有个数据,所有公开 commit 里 4% 是 Claude Code 写的——这是全世界所有代码的 4%。"

"All the companies use Claude Code — from the biggest companies to the smallest startups."

"所有公司都在用 Claude Code——从最大的公司到最小的创业公司。"

"It plotted the course for Perseverance — like for the Mars rover."

"它给 Perseverance 规划了路径——就是火星车那个。"

"This is the coolest thing for me — we even printed posters, because the team was like: wow, it's just so cool that NASA chooses to use this thing."

"这对我来说是最酷的事——我们甚至印了海报,因为团队说:哇,NASA 选用这个东西,太酷了。"

"So it's humbling — but it also feels like the very beginning."

"所以——挺 humbling 的——但又觉得这才刚刚开始。"

"What's the interaction between Claude Code and co-work?"

"Claude Code 和 co-work 之间是什么关系?"

"Was it a fork of Claude Code?"

"它是 Claude Code 的 fork 吗?"

"Was it like — you had Claude Code look at Claude Code and say "let's make a new spec for non-technical people that keeps all the lessons", and then it went off for a couple days and did that?"

"会不会是——你让 Claude Code 看自己,然后说"为非技术用户写个新 spec,保留所有教训"——然后它自己跑几天搞定?"

"What's the genesis of that, and where do you think that goes?"

"这事的起源是?你觉得它往哪儿走?"

"This is going to be like my fifth time using the word "latent demand"."

"这大概是我第 5 次说 "latent demand" 这个词了。"

"We were looking at Twitter, and there was that one guy that was using Claude Code to monitor his tomato plants."

"我们在看 Twitter——有个人用 Claude Code 监控他的番茄植株。"

"There was another person that was using it to recover wedding photos off of a corrupted hard drive."

"还有人用它从坏掉的硬盘里恢复婚纱照。"

"There were people using it for finance."

"还有人用它做财务。"

"When we looked internally at Anthropic, every designer is using it, the entire finance team is using it, the entire data science team is using it — not for coding."

"我们看 Anthropic 内部——每个设计师都在用,整个财务团队在用,整个数据科学团队在用——而且不是用来写代码。"

"People are jumping over hoops to install a thing in the terminal so they could use this."

"大家在跳过各种障碍——只为了能在终端里装这个东西。"

"So we knew for a while that we wanted to build something."

"所以我们其实知道想做点什么。"

"We were experimenting with a bunch of different ideas — and the thing that took off was just a little Claude Code wrapper in a GUI in the desktop app."

"我们试了一堆 idea——最后跑通的——是桌面 app 里一个小小的 Claude Code GUI wrapper。"

"That's all it is — it's just Claude Code under the hood, the same agent."

"就这么简单——底下还是 Claude Code,同一个 agent。"

"Felix and the team — Felix was an early Electron contributor, he knows that stack really well."

"Felix 和团队——Felix 是早期 Electron 贡献者,这套技术栈他特别熟。"

"He was hacking on various ideas, and they built it in I think something like 10 days."

"他试了一堆 idea,大概 10 天就把它做出来了。"

"It was just like 100% written by Claude Code, and it just felt ready to release."

"100% 是 Claude Code 写的——而且就觉得它准备好发了。"

"There was a lot of stuff that we had to build for non-technical users."

"为非技术用户我们要加很多东西。"

"All the code runs in a virtual machine."

"所有代码跑在一个虚拟机里。"

"There's a lot of protections for deletion and things like this."

"删除之类的操作有很多保护。"

"There's a lot of permission prompting and other guardrails for users."

"还有一堆 permission prompt 和 guardrail。"

"Honestly, it was pretty obvious."

"老实说——这事挺显然的。"

"Boris, thank you so much for making something that is taking away all my sleep — but in return, it's making me feel creator mode again, sort of founder mode again."

"Boris,谢谢你做了一个抢走我所有睡眠的东西——但作为回报——它让我重新感觉到 creator mode、founder mode。"

"It's been an exhilarating 3 weeks."

"这 3 周非常令人激动。"

"I can't believe I waited that long, since November, to actually get into it."

"我都不敢相信我从 11 月开始就拖到这么晚才真正进入这件事。"

"Thank you so much for being with us, thank you for building what you're building."

"非常感谢你来,谢谢你在做的事情。"

"Yeah, thanks for having me."

"是,谢谢邀请。"

"And — send bugs." [laughter]

"还有——把 bug 发给我。"[笑]

"Sounds good."

"好的。"