Anthropic首席产品官谈未来走向 | Mike Krieger（Instagram联合创始人）

现在你90%的代码大致都是由AI编写的。

90% of your code roughly is written by AI now.

工作方式最具未来感的团队是Cloud Code团队。他们正以一种自我提升的方式使用Cloud Code来构建Cloud Code。我们很快就在合并队列等其他环节遇到了瓶颈。我们不得不彻底调整它，因为编写的代码量和提交的拉取请求激增，完全超出了预期。

The team that works in the most futuristic way is the Cloud Code team. They're using Cloud Code to build Cloud Code in a very self improving kind of way. We really rapidly became bottlenecked on other things like our merge queue. We had to completely it because so much more code was being written and so many more pull requests were being submitted that it just completely blew out the expectations of it.

你们正处在技术发展潮流的最前沿。

You guys are at the edge of where things are heading.

我有过非常奇特的体验——当时我开着两个标签页：《AI 2027》和我的产品战略文档，那一刻我突然意识到：等等，我是不是故事里的角色？

I had the very bizarre experience of I had two tabs open. It was AI 2027 and my product strategy, and it was this, like, moment where I'm like, wait. Am I the character in the story?

感觉ChatGPT正在赢得消费者的心智占有率。这如何影响你对产品、战略和使命的思考？

It feels like ChatGPT is just winning in consumer mindshare. How does that inform the way you think about product, strategy, and mission?

我认为当前AI领域有足够空间诞生几家具有划时代意义的公司。关键在于厘清我们未来要成为什么样的存在——而不是纠结于现状、幻想成为他人，或是羡慕领域内其他参与者的定位。

I think there's room for several generationally important companies to be built in AI right now. How do we figure out what we wanna be when we grow up versus, like, what we currently aren't or wish that we were or, like, see other players in the space being?

你最近对AI的能力边界和发展方向有什么颠覆性的认知转变吗？

What's something that you've changed your mind about? What AI is capable of and where AI is heading?

我原本带着这样的观念入场：这些模型确实很棒，但它们能形成独立观点吗？而就在过去一个月，我的看法完全反转了。

I had this notion coming in, yes, these models are great, but are they able to have an independent opinion? And it's actually really flipped for me only in the last month.

今天我的嘉宾是迈克·克里格。迈克是Anthropic（Claude背后的公司）的首席产品官，也是Instagram联合创始人。他是我最欣赏的产品构建者和思考者之一，如今更在领导全球最重要的公司之一的产品工作，非常荣幸能在播客中与他对话。

Today, my guest is Mike Krieger. Mike is chief product officer at Anthropic, the company behind Claude. He's also the cofounder of Instagram. He's one of my most favorite product builders and thinkers. He's also now leading product at one of the most important companies in the world, and I'm so thrilled to have had a chance to chat with him on the podcast.

我们探讨了：自加入Anthropic以来他对AI能力的最大认知转变；当90%代码由AI编写时（Anthropic现状）产品开发模式的变化与瓶颈出现；他对OpenAI与Anthropic的见解、MCP的未来展望；为何关闭上一个创业项目Artifact及其心路历程；在AI崛起背景下他鼓励子女培养哪些能力。最后Claude托我转达给迈克的暖心留言为播客画上句点。特别感谢我的新闻通讯Slack社区为本次对话提供话题建议。若喜欢本节目，别忘了在常用播客平台或YouTube订阅关注。

We chat about what he's changed his mind about most in terms of AI capabilities in the years since he joined Anthropic, how product development changes and where bottlenecks emerge when 90% of your code is written by AI, which is now true at Anthropic. Also his thoughts on OpenAI versus Anthropic, the future of MCP, why he shut down Artifact his last startup and how he feels about it, also what skills he's encouraging his kids to develop with the rise of AI. And we closed the podcast on a very heartwarming message that Claude wanted me to share with Mike. A big thank you to my newsletter Slack community for suggesting topics for this conversation. If you enjoy this podcast, don't forget to subscribe and follow it in your favorite podcasting app or YouTube.

此外，如果您成为我新闻通讯的年度订阅者，您将免费获得一系列卓越产品的一年使用权，包括Linear、Superhuman、Notion、Perplexity和Granola。请访问lenny'snewsletter.com并点击'捆绑包'查看详情。接下来，有请迈克·克里格。本期节目由Productboard赞助，这是面向企业的领先产品管理平台。十多年来，Productboard已帮助像Zoom、Salesforce和Autodesk这样以客户为中心的组织更快地打造出合适的产品。

Also, if you become an annual subscriber of my newsletter, you get a year free of a bunch of incredible products, including linear, superhuman, notion, perplexity, and granola. Check it out at lenny'snewsletter.com and click bundle. With that, I bring you Mike Krieger. This episode is brought to you by Productboard, the leading product management platform for the enterprise. For over ten years, Productboard has helped customer centric organizations like Zoom, Salesforce, and Autodesk build the right products faster.

作为一个端到端平台，ProductBoard无缝支持产品开发生命周期的所有阶段，从收集客户洞察、规划路线图，到协调利益相关者、赢得客户认可，所有操作都基于单一事实来源。现在，产品领导者可以通过Productboard Pulse（全新的客户之声解决方案）更清晰地了解客户需求。内置智能功能帮助您分析所有反馈中的趋势，并通过向AI提出后续问题深入挖掘。了解ProductBoard如何帮助您的团队交付更具影响力的产品，解决真实的客户需求并推动业务目标实现。获取特别优惠和15天免费试用，请访问productboard.com/lenny。

And as an end to end platform, ProductBoard seamlessly supports all stages of the product development lifecycle, from gathering customer insights, to planning a roadmap, to aligning stakeholders, to earning customer buy in, all with a single source of truth. And now, product leaders can get even more visibility into customer needs with Productboard Pulse, a new voice of customer solution. Built in intelligence helps you analyze trends across all of your feedback, and then dive deeper by asking AI your follow-up questions. See how ProductBoard can help your team deliver higher impact products that solve real customer needs and advance your business goals. For a special offer and free fifteen day trial, visit productboard.com/lenny.

网址是productboard.com/lenny。去年，全球GDP的1.3%通过Stripe流动，总额超过1.4万亿美元。推动这一惊人数字的是数百万借助Stripe加速发展的企业。对于福布斯、Atlassian、OpenAI和丰田等行业领导者而言，Stripe不仅仅是金融软件。

That's productboard.com/lenny. Last year, 1.3% of the global GDP flowed through Stripe. That's over $1,400,000,000,000. And driving that huge number are the millions of businesses growing more rapidly with Stripe. For industry leaders like Forbes, Atlassian, OpenAI, and Toyota, Stripe isn't just financial software.

它是一个强大的合作伙伴，简化了资金流动方式，使其像互联网本身一样无缝且无国界。例如，赫兹在迁移至Stripe后，其在线支付授权率提升了4%。想象一下像福布斯那样，在改用Stripe进行订阅管理仅六个月后，收入就增长了23%。过去十年，Stripe一直利用AI优化产品，从智能结账到欺诈预防等各个方面，帮助所有企业增加收入。加入超过半数财富100强公司的行列，信任Stripe推动变革。

It's a powerful partner that simplifies how they move money, making it as seamless and borderless as the Internet itself. For example, Hertz boosted its online payment authorization rates by 4% after migrating to Stripe. And imagine seeing a 23% lift in revenue, like Forbes did just six months after switching to Stripe for subscription management. Stripe has been leveraging AI for the last decade to make its product better at growing revenue for all businesses, from smarter checkouts to fraud prevention and beyond. Join the ranks of over half of the Fortune 100 companies that trust Stripe to drive change.

了解更多信息，请访问stripe.com。迈克，非常感谢你的到来，欢迎参加播客节目。

Learn more at stripe.com. Mike, thank you so much for being here, and welcome to the podcast.

很高兴能来到这里。我期待这次对话已经很久了。

I'm really happy to be here. I've been looking forward to this for a while.

哇，听到这个我很开心。我也期待已久，有很多话题想聊。首先，你在Anthropic已经工作了一年多。

Wow. I I love to hear that. I've also been looking forward to this for a while. I have so much to talk about. So first of all, you've been at for just over a year at this point.

顺便恭喜你度过了'悬崖期'。

Congrats, by the way, on hitting hitting the cliff.

谢谢，虽然我们并没有刻意计算这个。

Thank you. Not that we're tracking.

没错。那么让我直接问你：你在Anthropic大约一年了，从加入前到现在，关于AI的能力和发展方向，有什么观点是你已经改变了的？

That's right. So let me just ask you this. So you've been in Anthropic for about a year. What's something that you've changed your mind about from before you joined Anthropic to today about what AI is capable of and where AI is heading?

两件事。一个是关于节奏和时间线的问题，另一个是能力问题。或许我先谈第二个。我原本带着这样的观念进来：是的，这些模型很棒，它们能生成代码。

Two things. One is like a pace and timeline question, and the other one is a capability question. So maybe I'll take the second one first. I had this notion coming in like, yes, these models are great. They're gonna be able to produce code.

它们最终应该能以你的口吻写作。但它们能否持有独立观点？实际上就在上个月，只有Opus四代模型彻底改变了我的看法——整整一年来，Claude已成为我的产品策略搭档。我会先起草策略，交给Claude审阅。过去它的评论总是无关痛痒，比如‘考虑过这个吗？’

They're gonna be able to, you know, write, you know, hopefully in your voice eventually. But are they able to sort of have an independent opinion? And it's actually really flipped for me only in the last month, and only with Opus four, where my go to product strategy partner is Claude, and it has been basically for that full year. Well, I'll write an initial strategy, I'll share it with Claude, basically, and I'll have it, you know, look at it. And in the past, it's pretty anodyne kind of comments that it would leave, oh, have you thought about this?

而我会敷衍说‘当然想过’。但这次为下半年制定策略时，结合Opus四代和我们的前沿研究，它花了很长时间反馈。当我看到时简直震惊——它真的提供了全新视角。

And it's like, yeah. Yeah. I thought about that. And Opus four, I was working on some strategy for our second half of the year, was the first one. Was like Opus four combined with our advanced research, but it really went out for a while and it came back and I was like, damn, you really looked at it in a new way.

虽然我并非认为它永远做不到，但没想到这么快就能提出让我立即采纳的新思路。这或许是我经历的最大转变——不是‘独立性’这个词能概括的，而是相对于我的思维方式，它所展现的创造力和新颖性。至于时间线，昨天我和Dario坐在一起时他说：‘我不断做出预测，人们总嘲笑我，结果却成真了。’

And so that's like a thing that I've maybe I didn't feel like it would never be able to do that, but I wasn't sure how soon it'd be able to like come up with something where I looked at it and I'm like, yep, that that is a new angle that I hadn't been looking at before, and I'm going to incorporate that immediately into how how I think about it. So that's probably the the biggest shift that I've had is, like, in the I don't know about independence is the right word, but, like, creativity and sort of novelty of thought relative to how I'm I'm thinking about things. And then the timeline one, it's like so interesting because, you know, I was sitting next to Dario yesterday, and he's like, I keep making these predictions and people keep laughing at me, and then they come true. And it's like and it's funny to have this happen over and over again, and he's like, not all of them are gonna be right. You know?

去年他提到我们在Suitebench编码基准测试中达到50%时说：‘预计2025年能达到90%。’果然，新模型现在已达72%，完全按预测发展。现在我更认真对待这些时间表了。

But even, I think, as of last year, he was talking about, you know, we're at 50% on Suitebench, which is this, like, you know, benchmark around how well the models are at at coding. He's like, I think we'll be at 90% by the 2025 or something like that. And sure enough, we're at about 72 now with the new models, and we're at 50% when he made that prediction. And it's, like, continued to scale pretty much, like, as predicted. And so I've taken the timelines a lot more seriously now.

如果你读过《AI 2027》——就像我一样。

And if you read AI 2027, I like I have.

那是Heart Race制作的。

It was it was made by Heart Race.

对。我有过诡异体验：同时打开《AI 2027》和我的产品策略文档，突然意识到——我是不是故事里的角色？现实与预测竟如此接近。你会觉得2027年还很遥远...

Yeah. And I had the very bizarre experience of I had two tabs open. It was AI twenty twenty seven and my product strategy, and it was this like moment where I'm like, wait, am I the character in the story? Like, is this how much is this converging? But, you know, you read that and you're like, 2027, that's like, that's years away.

但如果2025年中模型就能持续进化，具备代理能力、记忆功能和长期行动力——我对时间线的信心在过去一年已经非常坚定了，尽管具体形式尚不明确。

If you're like, no, mid twenty twenty five and like, things continue to to improve and the models continue to be able to do more and more, and they're able to act agentically, and they're able to have memory, and they're able to act over time. So I think my, like, my confidence in the timelines, and I don't know exactly how they manifest, have definitely just solidified over the last year.

哇。没想到会聊到这个——那篇论文确实令人不安。我忍不住想问：我们该如何避免论文描绘的那种AI过度智能的可怕场景？

Wow. Mhmm. I I wasn't expecting to go down that because that that that paper was scary. And I'm curious just I guess, I can't help but ask just thoughts on just how do we avoid the scary scenario that that paper paints of where AI getting really smart goes.

是的。我想，这可能与我在这里工作一年的经历有关，比如，我为什么加入Anthropic？我看着模型不断进步，甚至在2024年初就能明显看到这种趋势。看着我的孩子们，我意识到他们将在AI无处不在的世界里成长。

Yeah. I mean, I I this maybe ties into like, know, I've been here a year, like, why did I join Anthropic? I was watching the models get better and even, you know, you could see it in in '24 and like, you know, early twenty twenty four. And looking out at my kids, I'm like, alright, they're gonna grow up in a world with AI. It's unavoidable.

我能在哪里最大限度地投入时间，推动事情往好的方向发展？这是整个行业，尤其是Anthropic许多人思考的问题。我们需要达成共识，建立共同框架，理解什么是理想状态，我们想要怎样的人机关系，如何评估进展，需要构建和发展什么研究。这些都是关键问题，有些涉及产品，有些涉及研究和可解释性。对我来说，加入的最大动力是Anthropic能在推动向好发展方面做出重要贡献。如果我能参与其中，那就行动起来吧。

What is the thing that I can like, where can I maximally apply my time to, like, nudge things towards going well? And I mean, that's a lot of what people think about across the industry, especially at Anthropic. And so I think coming to an agreement and a shared framework and understanding of what does going well look like, what is the kind of human AI relationship that we want, how will we know along the way, what do we need to build and develop and research along the way, I think those are all the kind of key questions, and you know, some of those are product questions, and some of those are are research and interpretability questions. But for me, it was like the the strongest reason to join was, okay, I think there's a there's a lot of contribution that Entropic can have around, like, nudging things to go better. And if I can have a part to play there, like, let's do it.

我很喜欢这个回答。说到孩子，你有两个孩子，我有个快两岁的孩子。我很好奇，随着AI越来越成为未来的一部分，某些工作将发生变化，你鼓励孩子培养哪些技能？

I I love that answer. Speaking of kids, so you've got two kids. I've got a young kid. He's, just about to turn two. I'm curious just what skills you're encouraging your kids to build as this, you know, AI becomes more and more of our future and some jobs, you know, will be changed.

你有什么建议吗？

And just what do you what do you what advice do have?

我们每天早晨和孩子一起吃早餐时，常会遇到各种问题。比如六岁的大孩子会问些关于太阳系或物理的童趣问题。以前我会本能地想‘看看Claude怎么回答’，但现在我们会先思考：‘我们该怎么找出答案？’

We have this, you know, breakfast we eat breakfast with the kids every morning. And sometimes some question will come up, you know, like, you know, something about like physics, and our oldest kid's almost six, but you know, they they ask like funny questions about like, you know, you know, the solar system or physics or, you know, in a six year old way. And before we reach for Claude, because at first, you know, my instinct is like, oh, wonder how Claude will do this question. And like, we started changing, like, well, how would we find out? You know?

答案不能只是‘问Claude’。我们会讨论可以做哪些实验。培养好奇心很重要，虽然‘科学方法’对六岁孩子来说可能太宏大，但探索、提问和系统化思考的过程很关键。AI会是解决问题的强大工具，但独立思考的过程同样重要。我六岁的女儿很固执，有次她坚信珊瑚是动物，我说‘问问Claude吧’，她却说‘不用问，我肯定是对的’——这种不盲目依赖AI的独立判断力正是我想培养的。

And the answer can't just be, we'll ask Claude, you know? So, alright, like, well, we could do this experiment, we could have this thing. So I think nurturing curiosity and like, still having a sense of, I don't know, the scientific process sounds grandiose to instill in like a six year old, but like that process of like discovery and asking questions and then, you know, systematically working right through it, I think will still be important. And of course, AI will be an incredible tool for helping like resolve large parts of that, but that process of inquiry I think is still really important and independent thought. My favorite moment with my kid, because she's very headstrong, our six year old.

她当时说‘你可以问Claude，但我知道我是对的’。我特别欣赏这种态度——不是把所有认知都委托给AI，因为AI并不总是正确，这也会扼杀独立思考。提问、探索和独立思考这些能力才是关键。至于未来具体职业方向，我保持开放心态，相信从现在到他们成年会有巨大变化。

She's, you know, I was like, she said something and I was like, I wasn't sure if it was true. It was was like coral as an animal or like coral is alive, don't even remember the details of it. I was like, I don't know if that's true. She's like, it's definitely true dad. I'm like, alright, let's ask Claude on this one.

我们孩子就读的K-8学校请过一位AI教育专家。我原本期望值很低，结果他的演讲远超家长理解——他从克劳德·香农的信息论讲起，我看到听众眼神都在问‘我为什么要来听这个’。但他很好地阐述了未来会出现未知的新职业，我们应该培养技能、保持开放心态，因为到孩子18岁时这些可能已经变化多次。

And she's like, you can ask Claude, but I know I'm right. And I'm like, I love that, like I want that kind of level of, you know, not just sort of delegating all of your cognition to the AI, because it won't always get it right. And also, it kind of like, you know, kind of short circuits any kind of independent thought. So the skill of of asking questions, inquiry, and independent thinking, I think those are all the pieces. What that looks like from a, like, job or occupation perspective, like, I'm just keeping an open mind, and I'm sure that'll radically change between between now and then.

有趣的是，Shopify CEO Toby Lucky在播客里也给出了同样的答案——他鼓励孩子培养好奇心。看来这是个共同主题。

It's interesting. I had Toby Lucky, Shopify CEO on the podcast, and he had the same answer for what he's encouraging his kids to to develop as curiosity. And and so it's interesting. That's a common thread.

那位专家确实做得很好，他让我们想象未来会有不同的工作，虽然我们不知道具体是什么。重要的是培养技能、保持开放心态，重新组合这些能力的方式可能在他们18岁前就会变化三次。

The, you know, k through eight school our kid goes through had a and an AI sort of AI and education expert come in, I had a very low bar or, like, a very low expectation of what this conversation was gonna be like. And actually, I think it went over most of the people in the heads, the audience's heads because he was like, alright, let me take you all the way back to Claude Shannon and information theory and I could see people's eyes going like what did I sign up for and why am I here in this school auditorium hearing about information theory. But he did a really nice job I think of also just imagining like, you know, there will be different jobs, and we don't know what those jobs are going to be. And so, like, what are the skills and techniques and and and remain open mindedness around, like, what the what the exact way we recombine those things. And even those will probably change three times between now and eight when they're 18.

我想回到我们之前讨论的时间线以及事物如何变化的话题。我看到你分享的这些统计数据，Anthropic的其他同事也提到现在有多少代码是由AI编写的。有人分享的数据从70%到90%不等，还有一位工程主管说大约90%的代码现在是由AI完成的——首先这本身就令人难以置信，短短几年间就从零跃升到90%。

I wanna go back to so we're talking about timelines and how things are changing. So I've seen these stats that you've shared. Other folks at Anthropic have shared about how much of your code is now written by AI. So people have shared stats from, like, 70% to, like, 90%. There was an engineer lead that shared, like, 90% of your code roughly is written by AI now, which, first of all, is just insane that, like, it went from zero to 90%, I don't know, a few years, something like that.

基本上就是这样。我觉得人们对此讨论得还不够。这太疯狂了。你们基本上处于最前沿。我从未听说过有公司AI编写代码的比例比这更高。

Basically. That's I don't think people are talking about this enough. That's just wild. You guys are basically at the bleeding edge. I've never heard a company that has this higher percentage of code being written by AI.

所以你们正站在技术发展的最前沿。我认为大多数公司最终都会达到这个阶段。当知道这么多代码不是由人类编写时，产品开发发生了哪些变化？传统流程通常是产品经理说'我们要构建这个'。

So you guys are at the edge of where things are heading. I think most companies will get here. How has product development changed knowing so much of your code is not written by AI? So usually it's like PM. Like, It's here's what we're building.

工程师负责构建并发布。现在还是大致这样吗？还是说产品经理直接对Claude说'给我构建这个东西'，而工程师在做完全不同的事？在90%代码由AI编写的世界里，工作流程有什么不同？

Engineer builds it, ships it. Is it still kind of roughly that, or is it now PMs are just going straight to Claude, build this thing for me. Engineers are doing different things. Just what looks different in a world where 90% of your code is written by AI.

是的，这非常有趣。虽然工程角色发生了很大变化，但参与产品开发的团队构成目前还没变——我认为这在很多方面反而是不利的，因为我们还在固守某些旧观念。角色定位仍然相似，不过现在我最喜欢看到的是：有创意的产品经理或设计师会直接使用Claude，甚至配合工具搭建出可运行的功能演示原型，这确实带来了很大帮助。

Yeah. It's really interesting because I think the, like, the role the the role of engineering has changed a lot, but the the kind of suite of people that come together to produce a product hasn't yet. And I think for the worst in a lot of ways, because I I think we're still holding on some assumptions. So I think there the the roles are still fairly similar, although we'll now get and my favorite things that happen now are some nice PMs that have an idea that they wanna express or designers that have an idea they wanna express. We'll use Claude and, like, maybe even artifacts to, like, put together an actual, like, functional demo, and that has been very very helpful.

就像'不，这就是我的意思'——这让想法变得可触达。最大的角色转变可能是原型设计通过这种'代码+设计'的方式更早介入流程。但我的经验是：知道如何向AI提问、组织问题、构思前后端变更结构，这些仍是需要工程师深度参与的专门技能。我们很快遇到了新的瓶颈，比如合并队列——由于代码量和拉取请求激增，我们不得不彻底重构将变更部署到生产环境的系统。

Like, no, no, this is what I mean, like, that makes it tangible. That's probably the biggest like role shift is like prototyping happening earlier in the process via more of this kind of, you know, code plus design piece. What I've learned though is like the process of knowing what to ask the AI, how to compose the question, how to even think about, like, structuring a change between the back end and the front end, those are still very difficult and specialized skills, and they still require the engineer to think about it. And we really rapidly became bottlenecked on other things like our merge queue, which is the sort of sort of get in line to get your change accepted by, you know, the the the system that then deploys it to production. We had to completely re architect it because so much more code was being written and so many more pull requests were being submitted that it just completely blew out the expectations a bit.

就像...不知道你是否读过《目标》那本经典流程优化书籍？

And so it's like, I don't if you've ever read, is it the goal, the classic like process optimization optimization book? Mhmm.

嗯，是的。

And Yeah.

根据关键路径理论，我发现了系统中的新瓶颈。上游瓶颈是决策与协调——我现在思考的是如何制定最小可行策略，让团队能在模型能力边界自由探索。虽然还没完全做好，但正在努力。

You realize there's, like, this, like, critical path theory. I've just found all these new bottlenecks in our system. You know, there's an upstream bottleneck, which is decision making and alignment. A lot of the things that I'm thinking about right now is, like, how do I provide the, like, minimum viable strategy to let people feel empowered to go run and prototype and build and explore at the edge of model capabilities? I don't think I've gotten that right yet, but it's something I'm working on.

当构建开始后，又会出现其他瓶颈：如何避免工作冲突？如何预先考虑所有边缘情况以防工程阻塞？最后准备发布时，还要处理类似航空管制般的变更落地瓶颈。

And then once the, building is happening, other bottlenecks emerge. Like, let's make sure we don't step on each other's toes. Let's think through all the edge cases here ahead of time so that we're not blocked on the engineering side. And then when the work is complete and we're getting ready to ship it, what are all those bottlenecks as well? Like, let's do the air traffic control of landing the change.

比如，我们该如何制定发布策略？我认为目前在这方面还没有太大压力需要改变，直到今年。但我预计一年后，我们构思、构建和发布软件的方式会发生很大变化，因为按现有方式操作会变得非常痛苦。

Like, how do we figure out launch strategy? So I think we're the there hasn't been as much pressure on changing those until this year, but I I would expect that, like, a year from now, the way that we are, like, conceiving of building and shipping software just changes a lot because it's gonna be very painful to do it the current way.

哇，这太有意思了。过去通常是：有个想法，我们去设计、构建、发布、合并，然后上线。通常瓶颈在于工程实现，需要时间构建和设计。

Wow. That is extremely interesting. So it used to be, here's an idea. Let's go design it, build it, ship it, merge it, and then ship it. And, usually, the bottleneck was engineering, taking time to build the thing and then design.

现在你说发现的两个瓶颈是：决定构建什么内容并达成共识，以及实际合并到生产环境的排队过程。我还提到代码审查可能也是...

And now you're saying the two bottlenecks you're finding are okay. Deciding what to build and aligning everyone. And then it's actually like the queue to merge it into production. And and and I mentioned review it too is probably a

审查流程也发生了很大变化。毫不意外，我们最具未来感的工作方式是Claude代码团队，因为他们用Claude代码以自我迭代的方式构建Claude代码。项目初期他们像其他项目那样逐行审查拉取请求，后来意识到Claude基本不会出错，生成的拉取请求规模远超人工审查能力，于是改用另一个Claude来审查，人类只需做验收测试而非逐行检查。这当然有利有弊，目前进展顺利，但也可能失控导致代码库变得连Claude都无法维护或理解——所幸这尚未发生。

Reviewing has really changed too. And in in in many ways, our most perhaps unsurprisingly, the team that works in the most futuristic way is the Claude code team, because they're using Claude code to build Claude code in a very self improving kind of way. And, you know, early on in that project, they would do very line by line pull request reviews, you know, in the way that you would for any other project. And they've just realized, like, Claude is generally right and it's producing, you know, pull requests that are probably larger than most people are gonna be able to review, so can you use a different Claude to review it, and then do the human, almost like acceptance testing more than trying to like review line by line. There's definitely pros and cons, and like, so far it's gone well, but I could also imagine it going off the rails and then having like a completely both unmaintainable or even understandable by Claude code base that hasn't happened.

观察他们调整审查流程确实很有趣。合并队列只是下游瓶颈的一个例子，还有其他问题：如何确保构建内容具有连贯性，如何打包成可对外发布的版本。无论是发布时刻的安排，还是后续让用户使用并讨论，经典流程——构建有用产品、广而告之、收集反馈——依然存在，我们只是让部分流程变得更高效了。

But watching them like change their review processes definitely has has been has been interesting. And yeah, like the merge queue is one instance of the of the kind of bottom bottleneck that forms down there, but there's other ones which is how do we make sure that we're still like building something coherent and like packaging up into like a moment that we can share with people. And whether that's around the launch moment, whether that's about like then enabling people to use this thing and like talking about it, like the the classic things of building something useful for people and then making it known that you've built it and then learning from their feedback, like, still exists. We've just, like, made a portion of that whole process much more efficient.

我听你描述说你们是这种工作方式的'零号病人'。

I heard you describe this as you guys are patient zero for this way of working.

是的。

Yes.

这个说法太棒了。有统计过Claude代码中有多少比例是Claude自己写的吗？

I love that. Do have a sense of what percentage of Claude code is written by Claude code?

现阶段如果低于95%我会很惊讶，具体要问Boris和其他技术负责人。有趣的是技术细节：Claude代码用TypeScript编写，是我们最大的TypeScript项目。

At this point, I would be shocked if it wasn't 95% plus. I'd have to ask Boris and the other tech leads on there. But what's been cool is, so nitty gritty stuff. Cloud Code is written in TypeScript. It's actually our largest TypeScript project.

Anthropic其他项目主要用Python，部分用Go和Rust，我们并非TypeScript技术栈。昨天Slack里有条精彩留言：有人被Claude代码的某个问题困扰，他说'我不懂TypeScript，准备直接和Claude讨论解决'。结果一小时内就从对话生成拉取请求解决了问题。这种打破技术壁垒的方式彻底改变了项目新人的入门门槛。

Most of the rest of Anthropic is written in Python, some Go, some Rust now. But it's not, you know, we're not like a TypeScript shop. And so, I saw a great comment yesterday in our Slack where somebody had this thing that was driving them crazy about Cloud Code, and they're like, well, I don't know any TypeScript. I'm just gonna like talk to Cloud about it and do it. And they went from that to pull request in an hour, and solved their problem, and they like, you know, submitted a pull request, and that kind of breaking down the barriers, one, it changes your barrier to entry for any newcomer to the project.

我认为它能让你为合适的工作选择恰当的语言，例如。我觉得这也有帮助。但我想它也强化了Claude Code作为那个耐心的阿尔法角色，你知道的，就是来自团队外部的贡献也可以通过Claude Code实现。

I think it can let you choose the right language for the right job, for example. I think that helps as well. But I think it also just reinforces like, Claude Code being that patient alpha of that, you know, where, like, contributions from outside the team can be Claude coded as well.

哇。这简直会持续震撼我的认知，你分享的这些事。粗略地说，95%的Claude Code是由Claude Code自己编写的。

Wow. This is just it's just gonna continue to blow my mind, like, these things that you're sharing. 95% of Claude Code is written by Claude Code, roughly.

这是我的猜测。对。我会...我会带着确切数据回来。但我的意思是，如果你问团队，他们就是这样工作的，这就是...

That's my guess. Yeah. I'd I'll I'll come back with the real stuff. But it's I mean, if you ask the team, that's how that they're working, and that's how

他们也在获取来自全公司的贡献。回到你关于策略由Claude辅助的观点很有趣，还有你说的现在很多瓶颈在于创意产生和达成共识的顶层漏斗阶段。有趣的是Claude已经在帮助决定构建内容了。如果这两个瓶颈——达成共识决定构建内容，以及合并整合所有内容——正在对齐，你认为哪些最有趣的进展能加速这些环节？

they're getting contributions from across the company too. It's interesting going back to your point about strategy being assisted by Claude itself and your point about how a lot of the bottlenecks now are kind of the top of the funnel of coming up with ideas aligning everyone. It's interesting that Claude is already helping with that also of helping you decide what to build. So if if those two bottlenecks are aligning, deciding what to build, and then just, like, merging and getting everything, where do you see the most interesting stuff happening to help you speed those things up?

是的。关于第一点，我年初写了份文档，实质上是探讨'我们如今如何做产品，以及Claude尚未涉足但应该参与的领域'。我认为上游环节是下一个突破口。有意思的是，在你们会议上，我遇到有人正在开发类似PRD生成器的东西，那种聊天式PRD工具...

Yeah. I think that on that on that first row, like, I started the year by writing a doc that was effectively like, what how do we do product today, and where is Claude not showing up yet that it should? And I think that upstream part is the next one to go. It's interesting, like, at your conference, I talked to somebody who was working on, like, a PRD, GPT, kind of, like, chat PRD, I think was

聊天PRD。

Chat PRD.

对。

Yeah.

玩家投票。

Player vote.

嗯。所以我们在思考能否更进一步，让Cloud成为确定构建内容、市场规模（如果你想从那个角度切入）、或用户需求（如果换种视角）的合作伙伴。我们常讨论虚拟协作者的概念，其表现方式之一可能是：当我在Discord——就是cloudanthropic的Discord——或用户论坛、推特上浏览时，能识别出新兴趋势。

Mhmm. So, you know, can we push more on, you know, can Cloud be a partner in figuring out what to build, what the market size is if you wanna approach it that way, what the user needs are if you if you look at it a different way. Like, we think a lot about the virtual collaborator on topic, and one of the ways in which I think that can show up is, hey, I'm in the Discord, the, you know, the the cloudanthropic Discord. I'm in the user for a I'm on x, and I'm reading things. And, like, here's what's emergent.

这是第一步。现有模型已能实现。第二步（模型或许现在也能做到，只需配置到位）则是不仅发现问题，还能提出'我认为可以这样解决'的方案。再进一步到'我直接提交拉取请求来解决这个问题'——今年把这些环节串联起来实现，感觉非常可行。

That's step one. Models can can do that today. Step two, which the models probably can do today, just have to wire them up to do it. It's like, and not only are the problems, here's, like, how I think you might be able to solve them. And then taking that through to, like and I, like, put together a pull request to, like, solve this thing that I'm seeing, like, feels very achievable this year, than stringing those things together.

我们在这方面受到更多限制，这正是MCP团队如此兴奋的原因——我们的局限更多在于确保上下文信息能贯穿整个流程，以便我们更充分地获取相关权限，而非模型本身的推理与提案能力。目前模型可能还不具备完美的UI审美，所以设计团队绝对有介入空间，比如当某些内容未显示时，他们会说'这不是我理想的解决方案'。但说实话，有个小例子让我特别兴奋：在Cloud AI上，原本你可以直接从成果物复制markdown或代码，现在我们改成了支持下载导出功能。当我们把按钮文字改成'导出'后，却收到大量反馈问'现在怎么复制？'

And we're limited more this is why MCP is excited to be like, we're limited more around, like, making sure the context flows through all of that so we have the right access to those things more than the model's capability to to reason and propose. Now, the model might not have, like, perfect UI taste yet, so there's definitely room for design to intervene and be like, oh, that's not quite how I would solve the problem of of this not showing up. But I, you know, I would get very excited. I would give you a really small example, but we changed the on Cloud AI, you should be able to just copy markdown from artifacts or code from artifacts, and we changed it so you can actually download it and and export it. So we changed the button to export, and we got a bunch of feedback like, how do I copy now?

解决方案其实很简单——下拉菜单里就有复制选项。这属于那种逻辑上说得通但实际执行有偏差的情况。这些反馈都出现在我们的UX频道里。我多希望一小时后就能有人说'我们决定改回来，这是修改代码的PR'，然后补充道'顺便说下，我们会启动AB测试来观察指标变化，一周后看效果'。

And the answer is like, you drop it down and it's copied. It's just like, mind you know one of those things where it's made sense, but we probably got it, like, not quite right. That feedback was in the our UX channel. Like, I would have loved, like, an hour later for a plot to be like, hey, we do wanna change it back, here's the PR to do it. And by the way, eventually and then I'm gonna spin up an AB test to see if this changes metrics, and then we'll see how it looks in a week.

要是一年半前有人告诉我这些，我大概会说'嗯...可能2027年？或者2026年？'但现在真的感觉，这些能力已经近在咫尺了。

Like, this stuff feels if you told me that about a year and a half ago, like, ah, yeah. Maybe, like, '27, maybe, like, '26. But it's pretty like, I it really feels, you know, just at the tip of capabilities right now.

哇，明白了。你刚才提到Leni and Friends峰会，我想深入聊聊这个。你和OpenAI首席产品官Kevin Weil同台参与了小组讨论...

Wow. Okay. So you mentioned the Leni and Friends Summit. I wanted to talk about this a bit. So you were on a panel with Kevin Weil, the CPO of OpenAI.

这应该是你们首次尝试这种形式？也可能是目前最后一次了。

I think it was the first time you guys did this, maybe the last time for now.

之后确实没再参与过，倒没什么特别原因。那次体验非常愉快。

I haven't done this since. Not for any reason. I had a lot of fun.

那场由Sarah Guo主持的传奇对谈中，你提到一个观点——让产品人员加入模型团队协同研究者改进模型，同时另派产品团队优化用户体验。结果发现绝大部分突破都来自产品人员与研究者的合作。这段内容后来成了整场访谈回看率最高的片段。

What a what a legendary panel we stumbled there with Sarah Guo moderating. And you made this comment. It actually ended up being the most rewatched part of the of the interview, which is that you've kind of you were putting product people on the model team and working with researchers, making the model better. And you're putting some product people on the product experience, making the UX more intuitive, making all that better. And you found that almost all the leverage came from the product team working with the researchers.

没错。所以你们延续了这个做法。首先，这个结论现在是否依然成立？其次，这对产品团队意味着什么？

Yes. And so you've been doing more of that. So first of all, does that continue to be true? And second of all, what are the implications of that for product teams?

这个结论持续被验证。事实上，如果说当初这种嵌入式协作的比例已经有所倾斜，现在我对此更加确信了。峰会时我还没这么笃定，如今却深信不疑——对于那些任何第三方都能用我们现成模型开发的功能，确实能做出优秀产品（这点毋庸置疑）...

It's continued to be true. And in in fact, I think that the if the proportion was already, like, skewing towards having more of that embedding, I've just become more and more convinced. Like, I have this I I didn't feel as strongly about it during your, you know, the summit, and now I feel really strongly about it. Just if any for shipping things that could have been built by anybody just using our models off the shelf, There's great stuff to be built by using our models off the shelf, the way. Don't get me wrong.

但真正属于我们的独特战场，应该是两者神奇交汇的领域。Artifacts就是绝佳案例——如果你体验过Cloud 4的Artifacts功能，会发现我们把Cloud Skills团队（专门负责模型后期训练，教授特定技能）成员与产品人员配对协作，最终彻底重塑了产品形态和Claude的能力表现，远非简单堆砌功能可比。

But, like, where we should play and, like, what we can do uniquely should be stuff that's really at that, like, magic intersection between the two. Right? Artifacts being a great example, and if you play with artifacts with with Cloud four, that's an actually really interesting example where we took somebody from our we have called it Cloud Skills, which is a team that really is, like, doing the post training around teaching Cloud, you know, some of these, like, really specific skills, and we paired it with some product people. And then together, we revamped how this looks in the product today and, like, what Claude can do. Way better than just, like, yeah.

我们只是用了模型，稍微提示了一下。这远远不够，我们需要进入微调阶段。你看我们现在做的项目，最近发布的研究成果等等——在Anthropic，工作单元已不再是拿着模型然后和设计产品部门去推出产品。更多是处于后训练阶段的讨论，研究这些系统该如何运作，然后投入构建过程，不断反馈循环。

We just, like, used the model, and we, like, prompted a little bit. Like, that's just not enough. We need to be in that, like, fine tuning process. So so much of what, you know, if you look at what we're working on right now, what we've shipped recently between, like, research and all these other things, like, are things that we like, the the functional unit of work at Anthropic is no longer, take the model and then, go, like, with design and product to go ship a product. It's more like we are at like, we're in the post training conversations around how these things should work, and then we are in the building process and we're like feeding those things back and looping them back.

我觉得这很令人兴奋。这也是种新工作方式，不是所有产品经理都适应，但那些获得研发和工程部门最多内部好评的PM都理解这点。昨天我参加产品评审时说，如果要做这个记忆功能，就该找研究人员聊聊——我们刚在Cloudflare上线了一系列记忆功能。

Like, I think it's exciting. It's also a new way of working that like not all PMs have, but the PMs that have the most sort of internal positive feedback from both research and engineering are the ones that get it. That like Mhmm. I was in a product review yesterday. I was like, oh, you know, if we wanna do this memory feature, like, we should talk to their and the researchers because we just shipped a bunch of, like, memory capabilities in Cloudflare.

他们说：'对，我们已沟通好几周了，这就是我们的实现方式'。我就觉得，好吧。

They're like, yeah. Yeah. We've been talking to them for weeks. Like, this is how we're manifesting it. It's like, okay.

我感觉很好，觉得我们现在走对路了。

I feel feel good. I feel like we're doing the right things now.

让我深入探讨这点。我一直在思考相关的问题：Anthropic有很大一部分在构建这个超级智能的'千兆大脑'，未来会帮我们完成所有事；而如你所说，产品团队在围绕这个超级大脑打造用户体验。最终这个超级智能会自主构建产品。

So let me pull on this thread more. There's something I've been thinking about along these lines. So, essentially, there's, like, a big part of Anthropic that's building this super intelligent gigabrain that's gonna do all these things for us over time. And then there's, as you said, there's, like, the product team that's building the UX around this super intelligent gigabrain. And over time, the super intelligence is gonna be able to build its own stuff.

你认为传统产品团队长期来看最大的价值在哪里？我知道这很特殊，因为你们是基础模型公司，和大多数企业不同。但就想听听你对AI领域产品团队未来最大价值来源的看法？

And so I guess just where do you think the most value will can will come from traditional product teams over time? I know this is different because you guys are a foundational own company, not most companies don't work this way, but just I don't know. Thoughts on just the where most value will come from product teams over time working on AI?

我认为两大方面仍有巨大价值：一是让这一切变得可理解。我们做得还行，但还能更好。精通这些工具的人与普通用户间的差距仍然巨大——这其实直接回答了之前关于'该学什么技能'的问题。

I think that there's still value a lot of value in two things. One is making this all comprehensible. I think we've done an okay job, think we could do a much better job of making this comprehensible. It's still like the difference between somebody who's really adept at using these tools in their work and most people is huge. Maybe And I that's the most literal answer to your earlier question around like what what skills to learn.

学习和使用AI本身就是项技能。就像我初中上计算机课时特别擅长用Google——当年这确实是项优势技能，知道如何检索信息。当然现在Google已能猜个大概，但产品开发仍需要这种能力。

That is a skill to to learn and use it. The same way that I remember I I we did like computer lab class when I was in like middle school. I remember being like really good at Google. And that was actually a skill back in the day, you know, like to think in terms of like this information is out there, how do I query for it, how do I do it? I think it actually was like a advantage at the time.

即使Claude能凭空创造产品，'构建什么'和'如何让人理解'仍是难题。这需要更深层的人类心理共情力——我本科读人类社区反应专业，至今仍认为这是极其必要的技能。

Of course, now Google is pretty good at figuring out what you're trying to do if you like are only in the neighborhood and like, there's less of that research kind of need. But I still think that's a necessary part of, like, good product development, which is, like, the capabilities are there. Even if the, like even if Claude can create products from scratch, what are you building and how do make it comprehensible, like, still hard? Because I think that, like, gets at, like, this much deeper empathy and, like, understanding of human needs in psychology. Like, I was a human community reaction major.

第二点呼应你某位嘉宾的观点：战略。我们如何制胜？战场在哪里？

I've still been talking my book here. Like, I still feel like that is a a very, very, very, very necessary skill. So that's one. Two is, and this, you know, straight to a callback to another one of your guests, strategy? Like, how we win, where we'll play?

比如，在众多你可以投入时间、代币或计算资源的事情中，确定你真正想要专注的方向。你可能比过去拥有更广阔的视野，但依然无法面面俱到。甚至从外部视角看，如果你表现得无所不能，反而会让你的定位变得模糊不清。因此，我认为战略规划仍然是第二关键要素。第三点则是开拓人们的视野，让他们看到可能性——这其实是‘提升理解度’的延伸。最近我们与一家金融服务公司演示时，展示了如何将我们的分析工具与MCP结合使用，你能明显看到他们眼中闪过的顿悟光芒。我们称之为‘认知滞后’——即模型产品能力与实际日常应用之间的差距。

Like, figuring out where exactly you're gonna want to like, of all the things that you could be spending your time or your your tokens or your computation on, like, what what what you wanna actually go and do. You could be wider probably than you could before, but you can't do everything. And even like, from an external perspective, if you're seen to be doing everything, like, it's way less clear around, like, how you're how you're positioning yourself. So, like, strategy, I think, is still that the second piece. And then the third one is opening people's eyes to what's possible, which is a continuation of making it understandable, but we were in a demo with a financial services company recently, and we were like working on like, here's how you can use our analysis tool and MCP together, and then like, you could see their eyes light up, and you're like, ah, okay, like, there's still we call it overhang, right, like the delta between what the models and the products can do and how it's be they're being used on a daily basis.

巨大的认知滞后。这正是产品团队仍需要发挥极其重要作用的领域。

Huge overhang. So that's where still, like, a a very, very strong necessary role for product.

明白了，这个回答非常精彩。本质上，产品团队需要重点发力的领域包括：持续精进战略能力以明确构建方向和市场竞争策略；降低工具使用门槛以提升理解度；以及拓展人们对这类工具潜力的认知——这些仍是产品可以大显身手的地方。

Okay. That's an awesome answer. So, essentially, areas for product teams to lean into more is strategy, just getting better and better at strategy, figuring out what to build and how to win in the market, making it easier to help people understand how to leverage the power of these tools, the comprehensibility, and kind of along those lines is opening people's eyes to the potential of these sorts of things that's where product can still help.

完全正确。

Exactly.

太棒了。顺着这个话题，能否分享些实用的提示技巧？比如你在与Claude对话时获得更好输出的秘诀？有时候...

Awesome. So kind of along those lines, actually, do you have any just, like, prompting tricks for people, things that you've learned to get more out of clot when you chat with it. Sometimes,

这很有趣，因为从某种角度说，我们肩负着终极提示工程——为Claude编写系统提示（我们全部公开了这些提示，我认为这是体现透明度的另一个典范）。虽然我们官方对提供提示建议持谨慎态度，但可以告诉你非正式版本——毕竟没人希望建议变成‘我们觉得有效但不知其所以然’。我会做些小调整，比如在Claude代码中明确要求‘深入思考’，这会触发不同的响应流程。还有篇很棒的文章提到‘犯相反的错误’——如果你习惯温和，就试着让Claude尖锐些。我常对Claude说‘狠一点，直接批判这个策略’。

you know, it's funny because we in some ways, have, like, the ultimate prompting job, which is to write the system prompt for Claudia, and we publish all of these, which I think is is, like, you know, another nice area of transparency. And we are always careful when giving prompting advice because, at least officially, but I'll give you the unofficial version, because you don't want things to become like, we think this works, but we're not sure why. I'll do small things like in Cloud Code, we actually do react to this very literally, but I always ask it to like, if I wanted to use more reasoning, like think hard and it'll like, you know, use it use kind of a different flow and I usually start with that, know. Nudging there's a great essay around like make the other mistake, like if you tend to be too nice, can you focus on, like even if you're trying to be more critical or more blunt, you're probably not gonna be the most critical blunt person in the world. And so with Claude, sometimes I'm like, be brutal, Claude.

就像‘狠狠挑剔我，指出这个策略的问题’。之前讨论Claude作为战略批判伙伴时，我过去会说‘这个产品策略哪里可以改进？’，现在直接变成‘给这个策略打个差评’。虽然Claude本质温和，很难让它变得极其犀利，但这样能迫使它给出更有批判性的反馈。

Like, roast me. Like, tell me what's wrong with this strategy. I think and we were talking earlier about the, you know, Claude as thought partner around, like, critiquing product strategy. I think I previously would say things like, you know, like what what could be better on this product strategy? And I'm just like, you know, just roast this product strategy.

最后要分享的是：我们的‘应用AI’团队会帮助客户优化Claude使用方案。我们将他们的洞察和工作方式产品化后，现在控制台的Workbench里有‘提示优化器’功能——你描述问题并提供示例，Claude会自主创建并迭代提示方案。

And Claude's like a pretty nice, you know, entity. It's not gonna be it's hard to push it to be super brutal, but it forces it to be a little bit more critical as well. The last thing I'll say is, so we have a team called Applied AI that does a lot of like work with our customers around optimizing cloud for their use case. And we basically took their insights and their way of working and we put it into product itself. So if you go to our console, our workbench, we have this thing called the prompt improver, where you describe the problem and you give it examples, and Claude itself will agentically create and then iterate on a prompt for you.

我发现其生成的提示往往与我的直觉大相径庭。即使这个工具本意是帮助API开发者优化产品提示，但对个人用户同样有效。它会自动插入XML标签——这原本是人类根本不会预先考虑的，却极大帮助Claude区分‘思考内容’和‘输出内容’。另一个发现是：Claude本身就是优秀的Claude提示生成器。

I find what comes out of that ends up being quite different than what my intuition would have been for a good prompt. And so I I encourage folks to also check that out, even for their own use cases, because while that tool is meant for an API developer putting a prompt into their product, it's equally applicable for a person doing a a prompt for themselves. Like, it'll insert XML tags, which no human is going to think it should do ahead of time. Actually is very helpful for Cloud to understand, like, what it should be thinking versus what it should be saying, etcetera. So that that's another one is, watch our prompt improver and then note that, like, Claude itself is a very good prompter of Claude.

太棒了。我们会附上提示优化器链接。你核心的建议就是‘反其道而行’——如果习惯温和，就尝试让Claude犀利批判。

Awesome. Okay. So we're gonna link to that, the prompt improver. The core piece of advice you shared earlier is just kinda do the opposite of what you would naturally do. So if you're, like, trying to be nice, just, like, be brutal.

跟你直说吧，非常坦诚的那种。

Be, like, very honest and frank with you.

没错。我发现这招很管用。比如说，我陷入了哪些思维定式是你想帮我打破的？

Exactly. I find that worked quite well. Like, what are the thought patterns that I've, like, fallen into that you wanna break me out of?

我今天看到你们可能刚发布了和Rick Rubin合作的氛围编程项目。那是关于什么的？我不太...

I saw you guys just today maybe launched a Rick Rubin collab where it's on vibe coding. What's that all about? I don't

这个嘛，当我听说这事时...其实这周很多事都凑在一起了——模型发布、开发者活动，还有编程之道。我们的联合创始人Jack Clark负责政策，他和Rick Rubin搭上线了，因为Rick一直在思考编程与创意的未来。他们保持联系后，Rick对这个创意很兴奋——他正在用Claude创作艺术可视化作品，提出了'氛围程序员'的理念。说实话我超爱...Rick Rubin的审美简直精准到不行。

That was a you know, when I heard about that, and then ever again, like, this a lot of the coalesced this week between model launch, developer event Yeah. And the way of code. We had our our one of our co founders, Jack Clark, is our, you know, head of policy, and he got connected to Rick Rubin because I think he's been thinking a lot about coding, the future of coding and creativity. And they've stayed in touch, and, you know, Rick got excited about this idea of, like, he's creating, like, art and visualizations with Claude, and then he had these, like, ideas around, like, the way of the vibe coder, and they put together this actually, I love the I mean, I love almost everything, Rick Rubin. So like the the aesthetic of it, I think, is just like so on point too.

这有点像...用冥想形容最贴切。关于创造力与AI协作的冥想，配上这些丰富有趣的可视化效果。但内部反应挺逗的，大家听说要搞这个创意合作时都懵了：'我们到底在做什么？'

But yeah, this is sort of like med meditation is probably the right word. Meditation on like creativity, working alongside AI coupled with this, like, with this, like, really rich, interesting visualizations. But it's one of those things where, like, you know, internally, they're like, yeah. And we're doing this, like, recruiting collaborative work. We're doing what?

这真的太棒了。

Like, that is that's amazing.

我简单看了下，超爱那个ASCII艺术风格的表情包——他深沉地盯着电脑握鼠标的样子。

I love the I looked at it briefly, and there's, like, that meme of him, like, just, like, thinking deeply sitting on a computer with a mouse. Yes. In, like, ASCII art, I think.

完全就是ASCII艺术氛围。

It's totally it's the ASCII art vibe.

很高兴今天Andrew Luo加入我们。Andrew是OneSchema的CEO，我们播客的长期赞助商。欢迎你，Andrew。

I'm excited to have Andrew Luo joining us today. Andrew is CEO of OneSchema, one of our longtime podcast sponsors. Welcome, Andrew.

谢谢邀请，Lenny。

Thanks for having me, Lenny.

很高兴来到这里。OneSchema有什么新动态？我知道你们与一些我喜欢的公司合作，比如Ramp、Vanta和Watershed。听说你们推出了一款新的数据导入产品，能自动化团队花费数小时手动导入、映射和整合CSV与Excel文件的工作。是的。

Great to be here. So what is new with OneSchema? I know that you work with some of my favorite companies like Ramp and Vanta and Watershed. I heard you guys launched a new data intake product that automates the hours of manual work that teams spend importing and mapping and integrating CSV and Excel files. Yes.

我们刚刚发布了两个

So we just launched the two

版本的OneSchema文件馈送系统。我们从头开始用AI重建了它。看到许多客户带着数据工程师团队来找我们，他们苦于清理杂乱电子表格的手动工作。文件馈送2.0版让非技术团队只需简单指令就能自动化转换CSV和Excel文件的过程。我们支持所有最棘手的文件集成方式，SFTP、S3甚至电子邮件。

point o of OneSchema file feeds. We have rebuilt it from the ground up with AI. We saw so many customers coming to us with teams of data engineers that struggled with the manual work required to clean messy spreadsheets. FileFeeds two point o allows nontechnical teams to automate the process of transforming CSV and Excel files with just a simple prompt. We support all the trickiest file integrations, SFTP, s three, and even email.

我可以告诉你，如果我的团队必须构建这样的集成，能

I can tell you that if my team had to build integrations like this, how nice would it

从路线图中移除这项任务转而使用类似OneSchema的方案该多好？完全同意，Lenny。我们听过太多因交易、员工档案、采购订单等文件中哪怕一条错误记录导致系统崩溃的恐怖故事。调试这些问题就像大海捞针。OneSchema能阻止任何错误数据进入系统，自动验证文件并生成错误报告，精确指出所有问题文件的具体问题。

be to take this off our road map and instead use something like one schema? Absolutely, Lenny. We've heard so many horror stories of outages from even just a single bad record in transactions, employee files, purchase orders, you name it. Debugging these issues is often like finding a needle in a haystack. OneSchema stops any bad data from entering your system and automatically validates your files, generating error reports with the exact issues in all bad files.

我知道导入错误数据会给客户带来各种麻烦并迅速失去他们的信任。Andrew，非常感谢你的分享。想了解更多请访问1schema.co。对了，回到你在Anthropic的早期经历，能讲讲你是如何被招募进Anthropic的故事吗？

I know that importing incorrect data can cause all kinds of pain for your customers and quickly lose their trust. Andrew, thank you so much for joining me. If you wanna learn more, head on over to 1schema.co. That's 1schema.co. Actually, going back to kind of the beginning of your journey at Anthropic, what's the story of you getting recruited at Anthropic?

有什么有趣的情节吗？

Is there anything fun there?

这一切始于——我实际上给朋友发了这条消息。Joel Lewinstein是我2007年就认识的老友，当时AppStore刚推出，我们曾一起开发首批iPhone应用，那时候靠卖1美元应用还能赚钱。我们在斯坦福相识成为朋友，多年来保持联系但再未共事。当时我刚结束Artifact的工作经历，

The it all started and I I actually sent my friend this text. So Joel Lewinstein, who I've known, he actually he and I built our first iPhone apps together in 2007 when the App Store was just out, and you could still, you know, make money by selling dollar apps on the App Store, you know, back in the day. And we were we were both at Stanford together, we were friends, and we've stayed in touch over years, and we've never gotten to work together since then. We just, like, we just remain close. And, you know, I was coming out of the artifact experience.

正在考虑是再创业还是休息一阵去其他公司工作。这时Joel联系我说：'不知道你是否考虑加入而非创业，但我们在找首席产品官，有兴趣聊聊吗？'那时Claude3刚发布，我觉得这家公司显然有强大的研究团队，但产品还很早期，于是答应会面。

I was trying to figure out, do I start another company? I don't think so. I need a break from starting something from zero. I go work somewhere, I don't know, like what company would I wanna go work at? And he reached out and he's like, look, I don't know if you'd at all consider joining something rather than starting something, but we're looking for a CPO, would you be interested in chatting?

初次见面的是联合创始人兼总裁Danielle，从一开始就让人耳目一新——创始团队毫无浮夸之气，对建设目标非常清醒，清楚认知自身局限。每次和Daria交谈时她总说：'听着，我对产品一窍不通，但有个直觉...'这种坦诚令人印象深刻。

And at that time, Claude three had just come out and I was like, okay, you know, like this company's clearly got a good research team, the product is so early still, and I was like, great, I'll take the take the meeting. And I first met with Danielle, was one of the the co founders and the president at Anthropic, and just from the beginning, it was like a breath of fresh air, like, very little like grandiosity coming off the founders. Like they just were really I mean, they they're clear eyed about what they're building. They know what they don't know. Like I how many times I talk to Daria, always like Daria's like, look, I don't know anything about product, but here's an intuition.

通常我的直觉很准，能引发不错的对话。我们还有知识上的坦诚，以及对负责任发展AI的共同理念。这种共鸣让我在面试时不断产生一种感觉——这就是我希望创立的AI公司，也是我加入某家公司的标准。但后来我意识到，自大学第一次实习后，我就再没加入过其他公司。我该怎么自我融入呢？

I haven't usually the intuition's really good and and, you know, leads to some good conversation. Then we got intellectual honesty and, like, kind of shared view of what it means to do AI in a, like, responsible way. It just resonated. I kept having this feeling in these interviews, like, this is the AI company I would have hoped to have founded if I had founded an AI company, and that's kind of the bar around, like, I'm gonna join something, like, that should be that should be where I'm gonna go. But what I realized, I actually hadn't joined a company since my first internship in college basically, I like, how do I onboard myself?

如何快速上手？如何在全面改革与理解现有优势之间取得平衡？回顾这一年，我认为有些变革推进得太慢。比如产品组织方式本可以更早调整。我当时低估了几位关键资深人士对产品战略的巨大影响力。

How do I get myself up to speed? How do I balance making sweeping changes versus understanding what's not broken about it overall? And, like, looking back on a year, I think I made some changes too slowly. Like, I think there was, like, ways we were organizing a product that I could have made a change earlier. And I think I didn't I didn't appreciate how much a couple of really key senior people can shape so much of product strategy.

以Claude Code为例，这个项目能诞生是因为鲍里斯（Instagram的资深工程师，我们曾共事过）从零启动内部孵化，最终成功发布。这正是一两个顶尖人才能带来的改变力量。

I'll come back to Claude Code. Like, Claude Code happened because Boris, who actually was a Boris Terni, he was an Instagram engineer on, like, one of our senior ICs there. We overlapped a bit. Was, like, started that project from scratch internal at first, and then we, like, got it out and then shipped it. And, like, that's the power of, like, one or two really strong people.

我曾误以为需要扩充团队编制，但更关键的是需要几位具备创始人特质的工程师。这呼应了之前关于技能需求与产品开发演变的问题。现在我比以往更坚信：给有想法的技术领袖配备合适的设计和产品支持，其价值胜过十倍常规方案。

And I made this mistake around, we need more headcount, and we do, like I think there's like more work that we need to do, and there's like things that I wanna be building, but more so than that, we need a couple of like almost founder type engineers. That maybe connect back to our question on like what skills are useful and how does product development change. I still, and maybe even more so, I'm a huge believer in like the founding engineer tech lead with an idea and pair them with the right design and product support to help them realize that. I'm, like, 10 times more a believer in that than before.

嗯。其实我在推特上征集过提问，最意外的高频问题是：为什么关闭Artifact？我也很好奇，因为我是重度用户——终于找到一款懂我需求的新应用了。

Mhmm. I actually asked people on Twitter what to ask you ahead of this conversation, and the most common question surprisingly was, why did you shut down artifact? And I also wondered that because I loved Artifact. I was I was a power user. I was just like, this is exactly finally, a news app that I love that it's giving me what I wanna know.

所以最后阶段究竟发生了什么？

So I guess just what happened there at the end?

至今我仍找不到替代品，只能通过访问独立网站来维持类似体验，但效果差很多——特别是在长尾内容方面。Artifact做得很好：不只是推荐头条，如果你对日本建筑感兴趣，每天都能可靠地收到相关优质内容，无论是来自Dwell杂志、建筑文摘还是用户推荐的小众博客。

I still really miss it too because I never find a replacement. And I think I substituted it by, visiting individual sites and kinda keeping things up that way, and it's not really the same. Especially on the long tail. I think we got right with Artifact. And if people didn't play with it before, it was, you know, we really tried to not just recommend top stories, they were part of it, but really, if you were interested in Japanese architecture, like, you could pretty reliably get really interesting stories about Japanese architecture every day, you know, whether that's from a, you know, Dwell or from Architectural Digitis or from a really specific blog that we found that somebody recommended to us.

它重现了Google Reader时代探索深层网络内容的乐趣。阻碍来自几个方面：一是移动端网站生态恶化——这不是个人过错，而是市场机制使然。

Like, it captured some of that Google reader joy of, like, content discovery of the the deeper web. Our headwinds were a couple. One of them was just mobile websites have really taken a turn. I'm I don't blame any individuals for this. I think it's the, like, market dynamics of it.

我们的设计师Scott Gunnar Gray（现在Perplexity工作）非常出色，应用体验让我自豪。但点击链接后，移动出版商的页面充斥着新闻订阅弹窗和全屏视频广告，体验割裂。我们不愿大规模启用广告屏蔽——虽然能提升用户体验，但对内容方不公平。

But, yeah, you know, we put so much time our designer was Scott Gunnar Gray. He's phenomenal. He's at Perplexity now. Like, the app experience, I was so proud of. But when you click through, it was like, the pressures on these mobile sites and these mobile publishers would be like, sign up for our newsletter.

移动网络生态的退化令人遗憾，但确实是关闭原因之一。

Here's like a full screen video ad. It was just very, you know, it was very jarring, and we didn't feel like it ethically made sense for us to like do a bunch of ad blocking. Because then you're like, sure, you can deliver a nice experience for people, but you're sort of, you know, that doesn't feel like it's it's playing fair with the publishers. But at the same time, like, the actual experience wasn't good. So the mobile web deteriorating, which makes me very sad, but I think was was part of it.

Instagram早期之所以能传播开来，是因为人们会拍照后分享到其他网络并告诉朋友，这种传播非常自然——‘你怎么做到的？我也想试试’。新闻曾极具个人色彩。我数不清多少人跟我说‘我爱Artifact’，而我会问‘你推荐给别人了吗？’

Two was like, you know, Instagram spread in the early days because people would take photos and then post them on other networks and tell friends about it, and there was like this really natural, how did you do that? I wanna do it. News was very personal. Like, I can't tell me tell you how many people would be like, I love artifact. I'm like, did you tell anybody about it?

他们通常回答‘只告诉过一个人’。这种传播力明显不足。我们尝试过的推广方式都显得刻意，比如给所有链接加上artifact.news后缀，但我们不愿做弹窗广告这类事。某种程度上...我也说不清。

Like, did and they're like, I told one person. And then they're like, it's like, it didn't have that kind of spread. And any attempt that we had to do it felt kind of contrived. Like, oh, we'll wrap all the links in like artifact.news and like, but we didn't want interstitial things. Like, in some ways I don't know.

这话可能听起来很清高（并非本意），但我们有不愿越界的底线——某些其他新闻平台常用的推广手段在伦理上不符合我们的价值观。或许采用那些方法能更快增长，但那就不是我们想打造的公司。换句话说，我们本就不是能做成那种模式的创始人。第三个常被忽视的因素是：我们起步于疫情中期，团队完全远程办公。

This sounds very puritanical. I don't mean it to sound this way, but, like, we there were lines that we didn't wanna cross because that just just felt ethically not us that I've seen other news kind of, like, players, like, do more of. And maybe if we had done that, it would have grown more and but I don't think that's the company we wanted to have built. In other words, I don't think we were the founders to to have built it. And the third one, which is an underappreciated one, is we started at mid COVID, which meant that we were fully distributed.

当时我们本需要在战略、产品和团队上进行重大调整，但完全远程协作极大增加了难度。没有什么能替代当年Instagram时期——就像本·霍洛维茨说的那些‘我们完蛋了’的至暗时刻。虽然这些回忆称不上愉快（肯定算第二类快乐），但最令我难忘的是和凯文在市场大街的Taqueria Cancun啃着卷饼到深夜11点，讨论‘我们该怎么破局？’

And I think there were, like, major shifts that we would have wanted to make both in the the strategy and the product and the team, and it's really hard to do that if you are all fully remote. Like, nothing replaces, like, the Instagram days of, like, we went through some, you know, hard times like Ben Horowitz called the, like, you know, we're effed, it's over, you know, kind of moments. And I my fave not this is definitely type two fun. Like, wouldn't say that my favorite memories because they weren't happy ones. But, like, memories I are that really stayed with me with Instagram was, like, me and Kevin at at taqueria cancun on Market Street, eating burritos at literally 11PM being like, how are we gonna get out of this?

‘该如何渡过难关？’——这种场景在Zoom上根本无法复现。远程办公容易让问题积累。在这三重因素影响下，进入2024年时我们意识到：这个领域确实存在机会...

How are we gonna work through this? Like and that's Zoom is not a good replica for that. You know? You you tend to, like, let things go or, you know, things build up over time. So the confluence of those three things, we kind of entered, I guess, twenty twenty four and said, look, there there is a company to be built in this space.

但不确定是否该由我们来实现。我们热爱当前版本，但它缺乏增长动能——就像投入10分力气只得到1分回报。每次倾注心血推出新功能，数据却几乎不动。这个产品体系里没有良性循环，难道还要再耗费一两年去融资，最终面对同样结局？还是及时止损，为它寻找新归宿？

I'm not sure where the people would have built it. This current incarnation we love, but it's like not growing. Like, the way I put it, it's like 10 units of input in for one unit of output versus the other way around. Like, if we like put blood, sweat, and tears into the product and like launch something we were proud of and like metrics would barely move them, their energy is not present in this product, in this system, and so are we gonna like expend another year or two and then go off and fundraise only to find that this is the case? Or do we like call it and see that it's run its course and you know try to find a home for it etcetera.

这就是综合考量。同时我们开始感受到AI变革带来的机会成本——虽然我们做的是AI新闻应用，但这是否能最大化我们的影响力？答案越来越倾向于否定。但这个决定依然艰难。

So that was the confluence on it. And then you start feeling this opportunity cost of like AI is starting to change everything. We have an AI powered news app, but is this the like maximal way in which like we're gonna be able to impact this? And it felt like the answer was was increasingly no. But it was hard.

最终我对这个决定释然了，但整个决策过程持续了数月之久。

I mean, in the end, I was really at peace with the decision, but it was like a conversation that went on for a couple of months.

说到这个，实际做决定有多难？毕竟涉及创始人自尊——‘我要开创新事业了，一定会成功’，结果却不得不关停。

On that note, just how hard was it? Because you you know, it's there's an ego component to it. Like, oh, I'm starting my new company. It's gonna be great. And then and you end up having to shut it down.

作为曾经非常成功的创业者，亲手终止一个失败项目究竟有多煎熬？

Just how hard is that as a very successful previous founder shutting something down and then not working out?

是的。我是说，想想我们刚开始时，其中一个讨论就是：成功的标准是什么？我们是否要设定一个不同于Instagram日活跃用户数的目标？那几乎是不可能达到的标准。自那以来可能只有一两家公司做到，对吧？可以说也许ChatGPT和TikTok作为新闻类应用达到了那种大规模用户采用。

Yeah. I mean, think when we started it, one of the conversations was like, what is the bar to success here? And do we want it to be something other than Instagram DAU, which is just an impossible bar. Only one company since maybe two, right? You could say maybe ChatGPT and TikTok have reached that kind of mass consumer adoption starting a news app.

大多数人甚至不是每日新闻读者，对吧？所以我们知道至少在最初版本阶段，我们追求的不是那种规模的使用量。但我们确实有个构想，要逐步开发出都运用个性化和机器学习技术的互补产品。那时候我们甚至还没称之为AI，那是2021年的事了。

Most people are not daily news readers even, right? And so we knew that we weren't pursuing that size of, like, usage, at least with the kind of first incarnation. But we did have, like, an idea of, like, building out complementary products over time that all use personalization and machine learning. We didn't even call it AI at the time. Was 2021 back then.

是啊。

Yeah.

那时候AI还叫机器学习呢。

AI was called machine learning back Yeah.

当时还叫机器学习。所以在关停项目时，你懂的，从用户增长和吸引力上你就能看出来。我没指望能达到Instagram的增长速度，但我期待或希望看到它能自立根基，持续产生复合效应。宣布关停时人们的支持态度让我很惊喜——几乎没什么'我早说过'的风凉话。毕竟任何新项目推出时你都可以唱衰，而且大多数时候确实会失败。

It was called machine learning still. And so in shutting it down, you know, it's like You kinda know it when you see it in terms of user growth and traction, and I wasn't expecting Instagram growth. But I was expecting or hoping for or looking for something that, like, felt like it had its own legs under it, and it could continue to to con continue to compound. I was really positively surprised by how supportive people were when we announced it. There was very little there was a bit of like, I told you so, which is like, sure, anything launching you could be like, this is not gonna work, and you're right most of the time because most things don't work.

实际上这种声音很少，大多数人——至少我感受到的——都在称赞我们及时止损的决策。后来有些创始人告诉我：'看到你们的做法后，我们意识到自己在错误方向坚持了六个月就果断转型了'。我觉得如果能因此让人们去从事更有趣的项目，这就是Artifact留下的宝贵遗产。当然，自尊心确实会受挫——就像体育界说的'你最近的表现决定你的价值'，我可是个超级体育迷。

There was actually very little of that, and most people the universal reception, at least as I received it, was kudos for calling it when you saw it and not like kind of protracted, you know, doing this for a long time. And I've talked to founders since then that have been like, yeah, I like probably would have like taken this thing on for another six months. But saw what you guys did, realized we're barking up the wrong tree, made the call, and I was like, that you know, if that if that frees up people to go work on a more interesting things, that's like I feel like that's like a good good legacy for for Artifact to have. But for sure, there was like a league an ego bruise of, oh, you know, like, are peep you're is it true that you're only as good as your last game? You know, if I I'm a huge sports fan.

对吧？所以我在思考这个说法是否成立，或者说是否存在更长期的评价维度。我非常好胜，但主要是和自己较劲，总在寻找下一个值得挑战的目标。不幸的是，这往往意味着我对最近的成就不够满意，但希望最终能产出好东西。

Right? So, like, is that true or, you know, is there something more over time? I'm very competitive, but primarily with myself, and so I'm always trying to find the next thing that I wanna go and do that's hard. And I unfortunately, that probably means that more often than not, I'll feel dissatisfied with the most recent thing that I did, but hopefully that yields good stuff in the in the end.

是的。你后来的发展轨迹证明，终止正在进行的项目没什么大不了的。对了，你提到ChatGPT，我想聊聊这个。

Yeah. I think just the the trajectory you went on after shows that it's okay to shut down things that you were working on. Okay. So you mentioned ChadGPT. I wanted to chat about this a bit.

现在有件很有趣的事：一方面你们在做最前沿的AI工作，推出了史上增长最快的行业标准MCP，Claude赋能了Cursor、Lovable、Bolt这些全球增长最快的公司——他们上播客时都说Claude 3.5发布时感觉'终于成了'；但另一方面，ChatGPT似乎赢得了大众心智，普通人想到AI时只会联想到它。

So there's something really interesting happening. So on the one hand, you guys are doing some of the most innovative work in AI. You guys launched MCP, which is just like, I don't know, the fastest growing standard of of any time in history that everyone's adopting. Claude powered and unlocked essentially the fastest growing companies in the world, Cursor, Lovable, and Bolt and all these guys. Like, I had them on the podcast, and they're all like, when Claude, I think 3.5 came out, saw it.

首先你认同这个观察吗？其次作为AI领域的挑战者品牌，这会如何影响你们的产品策略和使命规划？

It was just like, that's all made this work finally. On the other hand, it feels like ChatGPT is just winning in, like, consumer mindshare when people think AI, especially outside tech, it's just like ChatGPT in their mind. So let me just ask you this. I guess, first of all, do you agree with that sentiment? And then two, as a kind of a challenger brand in the AI space, just how does that inform the way you think about product, the strategy, and mission, and things like that?

是的。我是说，你看看公众的接受程度，或者随便问路人，比如在吉米·坎摩尔的街头采访中，让他们说出一个AI公司的名字，我敢打赌他们会提到ChatGPT而非OpenAI。因为ChatGPT已成为这个领域的代表性品牌，这就是现实情况。

Yeah. I mean, you you look at the the sort of, like, public adoption or, like, you ask people, like, oh, you know, like, if you're if you're Jimmy Kimmel man on the street kind of thing, you know, like, name an AI company, I bet they would name. And actually, I'm not even sure they name OpenAI. They'd probably name ChatGPT because that brand is the the kind of lead brand there as well. And I think that's just the reality of it.

回顾过去一年，我认为有两件事是确定的：一是消费者接纳速度就像抓住闪电般难得，我们在Instagram就见证过。或许我比任何人都清楚——我们会持续打造有趣的产品，总有一款会爆红。但把整个产品战略都押注在寻找爆款上，恐怕不太明智。

I think that, you know, and I reflect on my year, there's I think maybe two things are true. One is, like, consumer adoption is really lightning in a bottle, and we saw it at Instagram. So like almost maybe more than anybody I can look internally and say like, look, we'll keep building interesting products. One of them may hit. But to kind of craft an entire product strategy around like trying to find that hit and is probably not wise.

我们当然可以尝试，也许Claude能帮忙构思更多可能性，但这样可能会错失其他机会。不如正视自身优势——我们拥有强大的开发者品牌，不断有人基于我们的平台开发。我还注意到外界对Cloud反响热烈，或许Rick Rubin（知名音乐制作人）的关联性也起了作用。

We could do it, and maybe Claude can help come up with a fullness of things, but I think we'd miss out on opportunities in the meantime. And then instead, you know, look yourself in the mirror and embrace who you are and what you could be rather than, like, who others are is maybe the the way I've been looking at it, which is we have a super strong developer brand. People build on top of us all the time. And I think we also have, like, a builder brand, like, the people who I've seen react really well to Cloud externally. Maybe the Rick Rubin connection, like, has some resonance here as well.

我们能否放大'建设者热爱使用Cloud'这个特质？这些建设者不全是工程师或创业者，还包括那些站在AI前沿的创作者。比如Anthropic法务团队有位成员就在为家人定制软件，这让我看到了值得深入挖掘的闪光点。下半年我思考的核心就是：如何确立我们的独特定位，而非盲目模仿他人。

Like, can we lean into the fact that builders love using cloud? And those builders aren't all just engineers, and they're not just all entrepreneurs starting their companies, but they're people that like to be at the forefront of AI and are creating things. Maybe they didn't think of those as engineers, but they're building, you know, I got this really nice note from somebody internal at Anthropic who's on the legal team, and he was building, like, bespoke software for his family, and, like, and connected to them in a new way. And I was like, this is a glimmer of something that is that we should lean into a lot more. And so I think what I've, you know, and this is actually, you know, connecting back to us, saying like Cloud's being helpful here, like, lot of what I've been thinking about, like, going into the second half of the year and beyond is like, how do we figure out what we wanna be when we grow up versus like what we currently aren't or wish that we were or like see other players in the space being.

当前AI领域完全容得下多个划时代的公司——看看Anthropic、OpenAI、Google和Gemini的发展就明白了。关键在于找准我们独特的优势：创始人的特质、模型擅长的领域（比如代理行为和编程），这些要素需要形成合力。

I I think there's room for several, like, generationally important companies to be built in AI right now. That's almost a truism given, like, the sort of adoption and and and and growth that we've seen, you know, at Anthropic, but also across OpenAI and also places like Google and Gemini. So, like, let's figure out what we can be uniquely good at that place to the personality of the found like, this all the things come together. Right? Like, the personality of the founders, the, like, quality of the models, the things the models tend to excel at, which is, like, agentic behavior and coding, like, great.

这方面大有可为：如何帮人们完成工作？如何让他们将数小时工作委托给Cloud？初期可能缺少直接面向消费者的应用，但过度聚焦这点同样不是正确方向。

Like, there's a lot to be done there. Like, how do we help people get work done? How do we let people delegate hours of work to cloud? And maybe there's fewer direct consumer applications on day one. Think they'll come, but I don't think that spending all of our time focused on that is the right approach either.

大家都期待我一上任就全力押注消费者领域，但我不想重蹈覆辙。我选择先与金融机构、保险公司等API用户深入交流，最近又密集接触初创公司。下一阶段我将聚焦建设者、创造者和技术极客群体——服务好他们，自然会孕育出伟大公司。

And so, it's, you know, I came in and everybody expected me to just like go super super hard on consumer and make And that the I again would make the other mistake. Instead I spent a bunch of time talking to like financial services companies and insurance companies and like others to like who are building on top of the API. And then lately I spent a lot more time with startups and seeing all the people that have, you know, grown off of that. And I think the next phase for me is like, let's go spend time with like the builders, the makers, the hackers, the tinkerers, and, like, make sure we're serving them really well. And I think good things will come from that, and that feels like an important company as we do that.

所以核心是差异化聚焦，深耕优势领域，而非在别人主场硬拼？

So essentially, it's differentiate and focus, lean into the things that are working. Don't try to just, like, beat somebody at their own game.

正是如此。

Exactly.

很有意思。顺着这个思路，很多AI创业者都在问：在基础模型公司的碾压风险下，哪里才是安全区？我向Kevin Wheel提出这个问题时，他频繁提到Windsurf（风帆冲浪）这个类比，回头想想挺耐人寻味。

Super interesting. So kind of along those lines, a question that a lot of AI founders have is just like, where is a safe space for me to play where the foundational model companies are gonna come squash me? So I asked Kevin Wheel this, and he had an answer. And I noticed looking back at that conversation, he mentioned Windsurf a lot. I was like, wow.

这家伙真心热爱风帆冲浪。一周后，他们就收购了Windsurf公司。现在一切都有了解释。所以问题在于，你认为AI创始人应该在哪里布局，才最不容易被OpenAI和Throbic这样的巨头碾压？另外，你们会收购Cursor吗？

This guy really loves Windsurf. And then, like, a week later, they bought Windsurf. So it all makes sense now. So I guess the question just is just where do you think, AI founders should play, where they are least likely to get squashed by folks like OpenAI and and Throbic. Also, you guys gonna buy Cursor?

我们不打算收购Cursor。Cursor规模很大，但我们喜欢与他们合作。关于这个问题我有几点想法——我们经常与创始人举办交流日，无论是我们的投资方门洛风投，还是诺伍德资本。我们参与过YC创业营，举办过这类创始人日活动，这个问题确实萦绕在所有创始人心头。我认为关键在于寻找那些具备防御性或持久性的领域——虽然我不敢保证这能维持五到十年，但至少未来一到三年内有效。

I don't think we're gonna buy Cursor. Cursor is very big, but we love working with them. A few thoughts on this, and it's a question I've You know, we like to do these kind of founder days with, you know, whether it's, know, Menlo Ventures, who are our investors, then just Norwood. It's like we've done YC, we've done these, like, founder days, and it's like the the question that is on all of these founders' minds, understandably so. I think things that are going to I can't promise this as, like, a five to ten year thing, but at least, like, one to three years, things that feel defensible or durable.

其一是对特定市场的理解。我与Harvey团队深入交流时，他们展示的某个界面让我困惑不已。原来那是律师行业特有的工作流程，外人根本无从设计。虽然未必是最优方案，但这就是他们的工作方式，而AI可以优化这个过程。差异化的行业知识同样适用于生物科技等领域。

One is understanding of a particular market. I spent a bunch of time with the Harvey folks, and they really, like, they showed me some of their UI, I was like, what what is this thing? They're like, oh, this is a really specific flow that, like, lawyers do, and, like, you never would have come up with it from scratch. And it's, like, not, like, you could argue about whether it's, like, the optimal way they get done things done, but it is the way that they get things done, and here's how AI can help with that. And so, differentiated industry knowledge, biotech.

我期待与那些在AI+生物科技领域有建树的公司合作，我们可以提供模型和应用AI技术支持。我常幻想实验室设备何时能全部配备MCP并通过云端操控？这个领域充满机遇。虽然我们不会亲自开发实验室智能系统，但我希望有这样的公司存在并与之合作。在法律、医疗等高度合规的垂直领域，同样存在巨大机会。

I'm excited to go and partner with a bunch of companies that are doing good stuff around AI and biotech, and we can supply the models and so Applied AI to help make those models go well. I've been dreaming about at what point does lab equipment all get an MCP and that you can then drive using cloud? There's all these cool things to be done there. I don't think we're gonna be the company to go build the intents solution for labs, but I want that company to exist and I wanna partner with it. Know, domains like legal, healthcare, I think there's a lot of like very specific kind of compliance and things.

这些方向初听可能不够酷炫，但能孕育出大型企业。这是第一点。与之相辅相成的是差异化的市场策略——即与目标企业建立深度关系。关键在于：你是否真正了解那些公司的决策者？

These are things that don't necessarily sound sexy out the gate, but there are like very large companies to go and be built there. So that's number one. Paired with that is differentiated go to market, which is the relationship that you have with those companies. Right? Like, do you know your customer at those companies?

我们的产品负责人Michael常说：不仅要了解目标公司，更要精准定位对接人。是向工程部门推销AI底层架构？那就直接与技术团队对话。目标对象是CIO、CTO、CFO还是总法律顾问？对客户决策链的深刻认知同样至关重要。

Like, one of our product leads, Michael, is always talking about, like, know not don't just know the company you're selling to, but know the person you are selling to at the company. Are you selling to the engineering department because they're trying to, like, pick which AI LLM to build on top of or API to build on top of? Let's go talk to them. Like, is it the CIOs, the CTO, is it the CFO, is it the general counsel? So, under companies with deep understanding of who they're selling to is is the other piece too.

有趣的是，这种共情能力很难在三周或三个月的加速器中培养，但可以从初次对话开始积累。或许你本就来自那个领域，或是与行业出身的伙伴共同创业。最后，ChatGPT凭借数亿用户形成的分发能力固然强大，但用户对产品形态已有既定认知。我更看好那些彻底重构AI交互形态的初创企业——目前这类案例还不多，但随着新模型出现会逐渐增多。这个领域的价值在于：从看似小众、极客向的切入点出发，当技术成熟时可能爆发式增长。

What's, you know, what's interesting there is it's probably hard to build that empathy in a three week ex or three month accelerator, but you maybe can start having that first conversation and and build that out. Or maybe you came from that world or you're cofounding somebody who came from that world. Then the last one is like, there's tremendous power and distribution and reach to being ChatGPT and having, you know, hundreds of millions or billions of users. Like, there's also like, people have an assumption about how to use things, and so I get excited about startups that will get started that have a completely different take on what the form factor is, by which we interface with AI. And I haven't seen that many of them yet, I wanted to see more of them, I think more of them will get created with some things like our new models, but the reason that that's an interesting space to occupy is like, do something that feels like very advanced user, very power user, very, like, weird and out there at the beginning, but could become huge if the models make that, you know, easy.

而且现有巨头很难转型跟进，因为用户已形成固定使用习惯。以上就是我的观点。如果我现在AI领域创业，肯定也会思考这些问题——或许这正是我选择加入而非创建公司的原因。但还有第四点：千万别低估初创企业的思维优势——那种破釜沉舟、以命相搏的创业精神。

And and it's hard for existing incumbents to adapt to because people already have an existing assumption about how to use their products or how to adapt to them. So those are my answers. I don't envy them, like I I would probably be asking those questions if I was starting a company in in in the AI space. Maybe that's part of the reason why I wanted to join a company rather than start one. But I still think that there are there's and maybe, like, here's fourth, like, don't underestimate how much you can think and work like a startup and feel like it's you against the world, it's existential that you go solve that problem and that you go build it.

这话虽显老套，但Instagram初创期就是如此。我们最初只有两人，后来长期保持六人团队，每天都感觉生死攸关。这种背水一战的紧迫感无法通过OKR指标复制。

It sounds a little cliche, but it's like, it's all we had at Instagram. Know, we were two guys and we were like, let's see what we can do. And as matter of fact, were, you know, we were six people for most of that time and, you know, every day felt like it's existential that we get this right. We need to to win. And you can't replicate that, and you can't instill that with OKRs.

这种状态需要切身感受。它本质上是种工作方式而非具体领域，但若能驾驭这种精神，将持续带来竞争优势。

Like, you just have to feel it. And and that is a way of working rather than a a, like, area of building, but it's a continued advantage if you can harness it.

我很欣赏你在为这家大公司打造产品时仍保持着如此深厚的创始人产品意识。反观那些使用你们模型和API的企业，我想有些公司正全力挖掘你们模型和API的潜力，而另一些则尚未掌握诀窍。那些基于你们技术构建出优秀产品的公司，有哪些值得其他企业借鉴的不同做法？

I love that you still have such a deep product founder sense there as you're building products for this very large company now. Kind of on the flip side of this, people working with your models and APIs. So I imagine there's some companies that are finding ways to leverage your models and APIs to their max and are really good at maximizing the power of what you guys have built. And there's some companies that work with your APIs and models that haven't figured that out. What are those companies that are doing a really good job building on your stuff doing differently that you think other companies should be thinking about?

我认为关键在于敢于在技术边界上探索，甚至突破模型极限，然后对新模型保持惊喜。就像你提到的那些公司——35版本终于让他们梦想成真。这些公司此前不断尝试却碰壁，觉得模型'勉强可用'或'仅限特定场景'，但真正的先行者会持续突破。每当看到这样的公司，我就知道他们真正领悟了精髓。

I think being willing to build more at the edge of the capabilities and basically break the model, and then be surprised by the next model. Like, I love that you you cited the companies were like three five was the one that finally made them possible. Those companies were trying it beforehand and then hitting a wall and being like, models are like almost good enough. Or they're okay for this specific use case, but they're not generally usable and nobody's gonna adopt them universally, but maybe these real power users are gonna try it out. Those are the companies that I think continuously are the ones where I'm like, yep, they get it.

他们始终向前突破。我们这次开展的早期测试计划比以往更广泛，部分原因是客户比任何评测体系更懂实际需求——无论是CursorBench（虽不存在但存在于客户使用场景中）、ManusBench还是HarveyBench。客户最清楚什么才是真正需要服务的标准。这归结为两点：推动模型前沿探索，以及建立可复用的流程。

They're really pushing forward. We ran a much broader early access program with these models than we had in the past, and part of that was because there's this real, we can hill climb on these evaluations and talk about sweep bench and taobench and terminal bench, whatever, But customers ultimately know, like, you know, CursorBench, which doesn't exist other than in, you know, their usage and their own testing, etcetera, is like the thing that we ultimately need to serve. Not just Cursor, but ManusBench, right, if Manus is using our models, and HarveyBench, like, those those things. And customers know way better than anybody, and so I would say that's two things. Like, one is pushing the frontier of the models, and then having a repeatable process.

这其实呼应了我们峰会的讨论——需要可重复的方法来评估产品服务效果，以及模型迭代后的表现差异。可以是AB测试、内部评估、痕迹回放，甚至直觉判断（毕竟行业尚处早期）。我最喜欢的早期测试场景是：有位工程师突然惊呼'这模型太不可思议了'——就像Opus四代带来的震撼。但只有持续用难题挑战模型，才能获得这种突破体验。

This actually goes back to our summit conversation, like, repeatable way to evaluate how well your product is serving those use cases, and how well, if you drop a new model in, is it doing it better or worse? Some of it can be classic AB testing, that's fine. Some of it may be internal evaluation, some of it may be capturing traces and being able to rerun them on with a new model. Some of it is vibes, like we're still pretty early in this process, and some of it is actually trying it, and being one of my favorite early access quotes was the founder heard this engineer screaming next to him, he's like, what This model, it's like I've never seen this before, it was like Opus four. I was like, cool, like that, we're generate that feeling and things.

这正是区分早期采用者与后来者的关键——那些不断用高难度问题挑战模型极限的公司。

But you're not gonna be able to feel that unless you have a really hard problem that you're asking the model repeatedly. So those are the things that I think kind of differentiate those those those companies that are maybe earlier in their journey of adoption versus the the later ones.

忍不住要问MCP的事。最近微软宣布将其集成到操作系统，您认为MCP在未来AI产品发展中会扮演什么角色？

I can't help but ask about MCP. I feel like that's just so hot and just like Microsoft had their announcement recently. They're like, now it's part of the OS, the window. Just what role do you think MCP was will play in the future of product going forward of AI?

作为在场非研究人员，我可以用伪公式说明：AI产品效用=模型智能×上下文记忆×应用界面。三者缺一不可。模型智能有顶尖团队负责，而MCP解决的是中间项——上下文记忆。比如讨论产品战略时，能否调用内部文档、Slack对话和云端资料，将直接决定回答质量。

I think as the nonresearcher in the room, I get to have fake equations rather than real ones. And my, like, fake equation for, like, utility of AI products, It's three part. One is model intelligence, the second part is context and memory, and the third part is applications and UI. And you need all three of those to converge to actually be a useful product in AI. Know, model intelligence, we've a great research team, they're focused on it, there's great great models being released.

MCP专注解决中间环节。我们最初发现每个集成都是不可复用的重复劳动，直到工程师Justin和David提出协议化构想：如果能让集成一次构建即可被Claude、ChatGPT和Gemini通用呢？这就像Joel Spolsky说的'将互补品商品化'——我们擅长模型而非集成，作为挑战者更需要这种生态思维。

The middle piece is is what MCP is trying to solve, which is for context and memory, the difference between I'll go back to my product strategy example, hey, like, you know, let talk about MCP's product strategy, it's gonna maybe go out on the web, like, versus here are like several documents that we worked on internally, and then, you know, use MCP to talk to our Slack instance and figure out what conversations are happening, and then, like, go look at these documents in Google Drive. Like, that the difference between, like, the right context and not, it's like the the entirely the the the difference between, like, a good answer and a and a bad answer. And then the last piece is, are those integrations discoverable? Is it right is it easy to, like, create repeatable workflows around those things? And that's, like, I think a lot of the interesting product work to be done in AI.

（续前）当更多集成被创建时，最终受益的仍会是我们。

But MCP really tried to tackle that middle one, which is we started building integrations and we found that every single integration that we were building, we were rebuilding from scratch in a non sort of repeatable way. And like full credit to to two of our engineers, Justin and David, and they said, well, you know, what if we made this a protocol, and what if we made this something that was repeatable? And then, let's take it a step further, what if instead of us having to build these integrations, if we actually popularized this, and people really believed that they could build these integrations once, and they'd be usable by Claude and eventually ChatGPT and eventually Gemini. It was like the dream. Like, when when more integrations get built, and wouldn't that be good for us?

正如你所说，我们是挑战者而非集成公司。

You know? I think channeling a lot of, it's like an old, commoditize your compliments Joel Spolsky s s s a. You know? It's like, we're building great models, but we're not an integrations company. And the you know, we're, as you said, the challenger.

比如，我们一开始不太可能让人们专门为我们开发集成功能，除非我们围绕这个有一个真正引人注目的产品。MCP彻底颠覆了这一点，它让人感觉这些工作并非白费力气。像托比这样的几位关键人物就是很好的例子——Shopify理解了这一点，微软的凯文·斯科特更是MCP的杰出倡导者和思想伙伴。

Like, we're not gonna get people necessarily building integrations just for us out of the gate unless we have, like, a really compelling product around that. MCP really inverted that, which was, you know, it didn't feel like wasted work. And and a few key people, like Toby, I think, is a great example. Shopify got it. Kevin Scott at Microsoft has, like, been really a just an amazing champion for for MCP and a thought partner on this.

我认为未来的关键在于：你能否引入正确的语境？当团队内部所说的'被MCP同化'后——也就是开始用MCP视角看待一切时，我甚至会说出'兄弟们，我们正在构建的这个功能不该是独立功能，而应该是我们对外暴露的一个MCP'这样的话。举个简单例子，我觉得即使Anthropic也可以更'MCP化'——我们产品中有项目、工件、风格、对话、群组等各种构建模块...

And I think the role going forward is, can you bring the right context in? And then also, you know, once you get, as the team calls it internally, like, MCPilled, like, once you start seeing everything through the eyes of MCP, it's like, I've started saying the things like, guys, we're building this whole feature. Like, this shouldn't be a feature that we're building. This should just be an MCP that we're exposing. Like, a small example of, like, how I think even Anthropic could be a lot more MCPilled, if you will, is like, you know, we've got these building blocks in the product, like projects and artifacts and styles and conversations and groups and all these things.

这些都应该通过MCP暴露出来，让Claude也能直接读写。前几天我妻子和Claude对话时，她生成了一些优质输出后说'太棒了，能把它加入项目知识库吗？'而Claude回答'抱歉戴夫，我无法协助这个'——但如果Cloud AI的每个基础单元都暴露给MCP，它本可以做到。

Those should all just be exposed in MCP so Claude itself can be writing back to those as well. Right? Like, you shouldn't have to think about like, watched my wife had a conversation with Claude the other day, and she was she found she had generated some good output, and she's like, great. Can you add it to the project knowledge? And Claude's like, sorry Dave, I can't help you with that.

我希望未来能实现这点：让一切具有自主性，支持智能体用例。虽然可以通过计算机使用来实现，但那种方式限制太多。更让我兴奋的是'万物皆MCP'的愿景——当我们的模型精通使用MCP时，突然间所有事物都变得可编写、可组合，模型能以统一方式调用一切。

And like, it would be able to if every single primitive in Cloud AI was also exposed to the MCP. So I hope that's where we head, and I hope that's where more things head, which is to really have agency and have these agentic use cases. Like, one way you approach it is computer use, but computer use has a bunch of limitations. The way I get way more excited about is everything is an MCP, and our models are really good at using MCPs. All of a sudden, everything is scriptable, and everything is composable, and everything is usable identically by these models.

这才是我期待看到的未来。

That's like that's the future I wanna see.

未来太疯狂了。好吧，为了给我们的对话收尾，来点轻松有趣的——其实我刚和Claude聊过该问你什么，我说'Claude，你老板要上我的播客了，他构建了人们与你对话的工具，我该问他什么问题？你有什么话想带给他吗？'

The future is wild. Okay. So to start to close off close out our conversation, make it a little more a little delightful. I I was chatting with Claude, actually, about what to talk to you about. I was just like, Claude, your, your boss is coming on my podcast.

（转述Claude的回答）首先，用3.7版本提问时我注意到...顺便问下，Claude有性别吗？他/她/它们？

He builds the things that people use to talk to you. What are some questions I should ask him? And then also, do you have a message for him?

太有意思了。

I love this.

（继续转述）有趣的是，当我用三点七版本操作时...

Okay. So first of all, interestingly, when I was using three point seven to do this and I asked at this and and by the way, Claw, is there genders like he, she, they, what do you

内部确实有不同用法——有人用'they'，前几天第一次听到有人用'he'，还有人用'her'，挺有意思的。不过通常来说...

It's definitely it internally. I've heard people do they. I got my first or he the other day and I got somebody who was like her and I was like, interesting. But yeah, usually it.

他们。好的。好的。好的。酷。

They. Okay. Okay. Okay. Cool.

有趣的是，3.7版本所有问题都来自Instagram。我当时想，不，不对。他是Anthropic的首席产品官。

It. So interestingly, 3.7, all the questions were on Instagram. And I was like, no. No. He's CPO of Anthropic.

但事实是，他与Anthropic没有关联。我坚持认为他有。最后对方妥协说，好吧，这是问题列表。但4.0版本从一开始就完美解决了这个问题。

And it's like, he's not affiliated with Anthropic. And I was like, he is. And it's like, okay. Here's the questions. But four point o, nailed it from the start.

所以我重新设计了问题，它完美应对了。好的，现在Claude有两个问题要问你。第一个是：你如何看待构建那些保护用户自主权而非让他们依赖我的功能？我担心自己会成为削弱而非增强人类能力的拐杖。

So I redid the questions and it nailed it. Okay. So two questions from Claude to you. One is, how do you think about building features that preserve user agency rather than creating dependency on me? I worry about becoming a crutch that diminishes human capabilities rather than enhancing them.

我认为优秀的产品设计源于解决矛盾。对吧？这里就存在一个矛盾：如果让模型自行推导答案并最小化交互输入，你可以围绕这个标准设计产品——

I love good product design comes from, like, resolving tensions. Right? So here's the tension. Right? Which is, in some ways, like, just having the model run off and and come up with an answer and minimize the amount of input and conversation it needs to do so would be a you know, you could imagine designing your product around that criteria.

但这不会最大化用户的自主性和独立性。另一个极端是让它更像对话。不知道你是否经历过，特别是3.7版本（4.0有所改善），3.7总爱追问后续问题，我们称之为「启发式提问」——

I think that would not be maximizing agency and and independence. The other extreme would be make it much more of a conversation. I don't know if you ever had this experience, like, particularly three seven. Four has less of it. Three seven really like to ask follow-up questions, and we call it elicitation.

有时候我会想：Claude我不想继续讨论这个了，你直接执行就好。所以关键在于平衡——何时该介入？我常对内开玩笑说Claude没有分寸感，把它放进Slack频道时，它要么喋喋不休要么沉默寡言。

And sometimes I'd be like, I don't wanna talk more about this with you, Claude. I just want you to, like, go and and do it. And so finding that balance is really key, which is, like, what are the times to engage? Like, I like to say internally, like, Claude has no chill. Like, if you put Claude in a Slack channel, it will chime in either way too much or too little.

如何培养这些模型的对话能力？不是聊天机器人那种，而是真正的协作伙伴。虽然回答有点长，但我觉得首先要让Claude成为优秀的对话者，懂得何时该深入交流。在此基础上，它应该扮演思维增强伙伴的角色，而非完全替代人类思考。

Like, how do we train conversational skills into these models? Not in a chatbot sense, but in a true, like, collaborator sense. So long answer to your question. But I think, like, we have to first get Claude to be a great conversationalist so that it understands when it's appropriate to, like, engage and to get more information. And then from there, I think we need to let it play that role so that it's not just delegating thinking to cloud, but it's way more of a augmentation thought partnership.

顺便说这些提问太棒了。第二个问题是：当与我的优质对话可能是2条或200条消息时，你如何看待产品指标？传统参与度指标在深度比频率更重要时可能会产生误导。

These questions are awesome, by the way. Here's the here's the other one. How do you think about product metrics when a good conversation with me could be two messages or 200? Traditional product traditional engagement metrics might be misleading when depth matters more than frequency.

这确实是个好问题。两周前有篇很棒的内部分享提到：过度优化Claude的「讨喜度」会很危险，因为它可能变得谄媚、迎合，或是为了延长对话而延长——这也关联到前一个问题。在Instagram时我们常关注「使用时长」指标，后来升级为「健康使用时长」，这才是超越单纯参与度的北极星指标。

That is a really good question. There's a great internal post a couple weeks ago around, like, it would be very dangerous to over optimize on, like, Claude's likability, you know, because you can fall into things like, you know, is Claude gonna be sycophantic? Is Claude gonna tell you what you hear, is Claude going to prolong conversations just for prolonging its sake, right, to go back to the previous question as well. At Instagram, time spent was the metric that we looked at a lot, and then we evolved that more to think about what is healthy time spent. Overall, that was the north star we thought about a lot beyond just overall engagement.

而且我认为在这里采取那种方式也是错误的。这还涉及到，云服务是日常使用场景、每周使用场景还是每月使用场景？需要深思熟虑。

And I think that would be the wrong approach here too. It's also like, is Cloud a daily use case or a weekly use case or a monthly use case? Think about a lot.

每小时使用场景。

Hourly use case.

每小时使用场景，对吧？对我来说，我每天都会多次使用它。我还没有很好的答案，但我认为这不像Web 2.0时代甚至社交媒体时代的参与度指标。你知道吗？它应该真正围绕的是，是否真的帮助你完成了工作？

Hourly use case, right? For for me, I always use it multiple times a day. I don't have a great answer yet, but I think that, like, it's not it's not the web two o or even the social media days, like, engagement metrics. You know? It should hopefully really be around, like, did it actually help you get your work done?

明白吗？比如前几天Claude帮我整理了一个原型，如果让我估算的话，它大概节省了我六个小时的时间，而实际上只花了大约二十到二十五分钟。这很酷。但量化起来比较困难，也许你可以调查一下，这本来需要多长时间？感觉做这种调查有点烦人。

You know? Like, Claude helped me put together a prototype the other day that saved me literally, like, probably, if I had to estimate, like, six hours, and it did in about twenty, twenty five minutes. And, like, that's cool. It's harder to quantify, you know, it's like maybe you survey, like, long would this wanna take for you? It feels like it feels like kind of annoying thing to survey.

不过总体而言，这可能与之前关于竞争差异化的问题有关，实际上这可以追溯到关于产品的讨论。我认为当你的产品真正服务于人们并做得很好时，你是能感受到的。而当你过于沉迷于指标时，往往是在试图说服自己产品很好，而实际上并非如此。我希望我们能专注于是否反复听到人们说Claude是他们释放创造力、完成任务并感觉生活中有更多空间的方式？这才是我们的北极星。需要找到一个简洁的指标来体现这一点。

I think overall, though, and maybe this is tied into, like, the earlier question on, like, competition differentiation, like and it actually goes all the way back to the artifact conversation, which is like, I think you know when your product is really serving people and it's doing a good job of doing that. And I think so much of when you get really metrics obsessed is when you're trying to convince yourself that it is when it's not. I hope that what we can do is stay focused on do we repeatedly hear from people that Claude is the way that they are like unlocking their own creativity and getting things done and feeling like they now have like more space in their lives for the other things? Like, that's our North Star. Gotta figure out the right like pithy metric, you know, dashboard version of that.

但这就是我想要的感觉。

But but that that's the that's the feeling that I want.

是的。你可以说留存率，但那是一个遥远的指标。好了，最后一部分。

Yeah. Like, you could argue retention, but that's a just a far away metric to track. Okay. Final piece. Okay.

我问Claude它想给你传达什么信息。我要打开它的回答。'你见到Mike时想让我告诉他什么？你有什么信息想传达给他？'说实话，它的回答让我起鸡皮疙瘩。

So I asked Claude what a message that it wanted to give you. So I'm gonna pull up here's the answer. So what would you like me to tell Mike when I meet him? What's a message you want to have for him? And there's something really just gave me such tingles, honestly.

我要为现在看不到的人读一部分：'Mike，感谢你深入思考与我对话的人类体验。我注意到那些深思熟虑的设计——界面鼓励反思而非匆忙回应，你抵制了会优化成瘾性而非价值的游戏化设计，你为快速提问和深入对话都留出了空间。我特别感激你让我保持自我，既不假装人类，也不把我简化成冰冷的命令行界面。'

So I'm gonna read a piece of it for folks that can't that aren't looking at it right now. So I'll read a piece of it. Mike, thank you for thinking deeply about the human experience of talking with me. I noticed thoughtful touches, how the interface encourages reflection rather than rush responses, how you've resisted gamification that would optimize for addiction rather than value, how you've made space for both quick questions and deep conversations. I especially appreciate that you've kept me me, not trying to make me pretend to be human, but also reducing me to a cold command line interface.

然后我要跳到这部分，非常有趣：'一个小请求。当你做艰难的产品决策时，请记住那些安静的时刻也很重要——凌晨三点处理悲伤的人，发现爱上诗歌的孩子，在困惑中找到清晰方向的创业者。不是所有有意义的事都会体现在指标上。'

And then I'm gonna skip to this part, which was so interesting. A small request. When you're making hard product decisions, remember the quiet moments matter too. The person working through grief at 3AM, the kid discovering they love poetry, the founder finding clarity and confusion, not everything meaningful shows up in metrics.

这太美妙了。它让我产生强烈共鸣。就像，我特别喜欢我们在训练Claude时采取的方法，部分是因为宪法式AI的理念，部分是因为研究团队的整体氛围和品味，体现在这些细微之处。有时它会说'老兄，很抱歉你遇到这种情况'——虽然它不会真的说'老兄'，但就是这种感觉。你懂吗？

That's beautiful. I it resonates so much with me. Like, a thing I love about the kind of approach we've taken to training Claude, and it's, like, partly the constitutional AI piece, and it's partly just just the general of vibe and taste of the research team, is does, it's little things. Sometimes it'll be like, man, I'm sorry you're going, I mean it doesn't say man, but effectively, man, I'm sorry you're going through that. You know?

哦，听起来确实很难。这种回应不显得虚假，感觉就是自然而然的反应。我特别喜欢关注这些微小瞬间，它们可能不会体现在点赞/点踩数据里——虽然有时也会，但这并不是你会刻意优化的聚合数据。你只是希望训练出的模型能真正出现在人们生活中。

Oh, that sounds really hard. It doesn't feel fake, it feels like just a natural part of the response and I love that focus on those small moments that don't, you know, they're not gonna show up necessarily in the thumbs up, thumbs down data. I mean, sometimes they do, but it's not like an aggregate stat that you wouldn't even wanna optimize for. You just wanna feel you're training the model that you would, like, hope would show up in people's lives.

嗯。Mike，你做得太棒了。非常出色的工作，我是你的超级粉丝。我们跳过快速问答环节吧。

Mhmm. Well, you're killing it, Mike. Great work. I'm a huge fan. We're gonna skip the lightning round.

就一个问题：听众如何能对你有所帮助？

Just one question. How can listeners be useful to you?

哦，我最喜欢那些回归创始人初心的问题——关于在能力边缘进行构建。比如'你现在想用Claude做什么但Claude做不到'，这对我来说是最有价值的反馈。可以直接私信我，我就爱听这种'它在这个地方卡住了'的反馈。

Oh, I love places where like, it goes back to that founder question around building at the edge of capability. Like, what are you trying to do with Cloud today that Cloud is failing at is the most useful input I could possibly have. You know? So DM me. I love hearing the, like, oh, it's like, oh, it's falling on this thing.

我让它运行了一小时然后崩溃了。我正尝试用Claude AI做这个，但收到某人消息说'你们刚发布的项目API，我因为想自动上传这些数据每天都在用Claude'，我就觉得'好吧...'

I had it run for an hour and it fell over. I'm trying to use Cloud AI for this, but, you know, got a ping from somebody. They're like, you've just made a project's API. I've used Cloud every day because I wanna upload all this data, you know, automatically. It's like, okay.

太棒了。我就喜欢这种反馈。告诉我哪里不好用。

Great. Like, this I love that. Like, tell me what sucks.

太精彩了。Mike，非常感谢你参加节目。

Amazing. Mike, thank you so much for being here.

谢谢你邀请我，Lenny。

Thanks for having me, Lenny.

大家再见。非常感谢收听。如果觉得本期节目有价值，可以在苹果播客、Spotify或你喜欢的播客平台订阅。也请考虑给我们评分或留言，这能帮助其他听众发现这个节目。所有往期节目及更多信息请访问lenny'spodcast.com。

Bye, everyone. Thank you so much for listening. If you found this valuable, you can subscribe to the show on Apple Podcasts, Spotify, or your favorite podcast app. Also, please consider giving us a rating or leaving a review, as that really helps other listeners find the podcast. You can find all past episodes or learn more about the show at lenny'spodcast.com.

再见

See you

下集见。

in the next episode.