迈克尔·特鲁尔：Cursor如何以AI的速度进行构建

我们正处于一个经历过iPod时刻的市场，而且，它即将迎来iPhone时刻。

We are in a market that's had a iPod moment, and, like, it's gonna have an iPhone moment.

我认为未来肯定会有更多这样的时刻，我们试图打造一家能够持续创造这些产品的公司。

And I think that there are definitely more in the future, and we've tried to build a company that can continually build those things.

我不认为API提供商真的知道该如何看待我们，今年四月，我们现在占据了他们API收入的相当高的两位数百分比，现在他们将不得不做出容量规划决策，甚至可能是融资决策。

I don't think the API providers really knew what to make of us, these April, and their thing now comprises, like, a really high double digit percent of their API revenue, and now they're gonna have to make capacity planning decisions, maybe financing decisions.

我认为在我们的领域存在一个巨大的多产品机会，可以构建一整套AI编程工具包，我们希望成为许多客户的AI编程服务提供商。

I think that there's a big multi product opportunity in our space where there's a whole AI coding bundle to be built, and we wanna be, for many of our customers, like, the AI coding provider for them.

今天，你将听到Cursor公司CEO迈克尔·特鲁尔的分享，关于打造我们见过增长最快的开发者工具——从因其规模导致主要云服务商宕机，到占据API提供商收入的双位数百分比，而团队却依然只是四个二十多岁的年轻人。

Today, you'll hear from Michael Truel, CEO of Cursor, on building the fastest growing developer tool we've ever seen, from taking down major cloud providers with their scale to becoming double digit percentages of API providers' revenue while still just being four twenty somethings.

我们讨论了为什么Focus能击败科幻小说和AI编程大战，他们如何在200多人规模时仍坚持那著名的两天工作试用期，以及猎取全球所有SONNET代币的战略艺术。

We discussed why Focus beat science fiction and the AI coding wars, how they maintained their infamous two day work trials even at 200 plus people, and the strategic art of hunting all the SONNET tokens in the world.

还有奥罗波洛斯问题：当颠覆软件的工具本身也是由软件构成时会发生什么？

Plus, the Oroboros question: What happens when the tool disrupting software is itself made of software?

让我们开始吧。

Let's get into it.

感谢你的到来，迈克尔。

Thanks for being here, Michael.

很高兴来到这里。

Glad to be here.

是的。

Yeah.

非常感谢。

Appreciate it.

他极少、极少做这类事，我不得不恳求他。

He very, very rarely does these things I had to beg.

所以我真的很感谢你能来。

So I really appreciate you coming up.

不。

No.

不会错过这个的。

Wouldn't miss this.

是啊。

Yeah.

好的。

Okay.

众所周知，Cursor公司的CEO迈克尔，这绝对是我们见过增长最快的公司之一。

So as everybody knows, Michael, CEO of Cursor, it's one of the fastest growing companies certainly we've ever seen.

它无处不在。

It's everywhere.

太疯狂了。

It's crazy.

为此你得雇佣操作员。

You have to hire operators for that.

实际上，我想探讨的不是典型的创始人成长历程，而是什么把你带到了这里。

So actually, what I wanna do is dig into not the typical kind of founder journey, what brought you here.

会稍微涉及一些，但主要是，你是如何处理这种混乱局面的？

There'll be a little bit of that, but like, how are you handling the mayhem?

这样很酷吗？

Is that cool?

当然。

Sure.

是啊。

Yeah.

不。

No.

听起来很棒。

That sounds great.

好的。

Okay.

很好。

Good.

那么首先，我们先简单回顾一下历史。

So to start off with that, we'll just do a little bit of history.

最近我见了一家公司，他们进来就说，我们是Cursor的3D团队。

So I met with a company recently and they came in and they said, we are the three d of Cursor.

我说，这故事真有趣，因为Cursor曾经就是一家3D公司。

And I said, funny story because Cursor was once a three d company.

是这样吗？

Is that right?

是的。

Yes.

你介意聊聊关于

Do you mind talking about kind of a bit of

起源故事吗？

the origin story?

当然可以。

Of course.

其实可以从多个不同的时间点来追溯公司的创立。

So there's a bunch of different ways you could actually peg the start date.

但本质上，公司的创立源于我和联合创始人们——我们曾是学校和其他地方的亲密同事，有两个时刻让我们对创业感到无比兴奋。

But effectively, the way the company got started was my co founders and I, we were close colleagues from school and some other places, and two moments got us really excited about building a company.

其一是尝试首批真正实用的AI产品，特别是体验了GitHub Copilot这个我们领域的先行者。

One was trying some of the first useful AI products, and in particular trying GitHub Copilot, the incumbent in our space.

这让我们萌生创业冲动的关键在于，这些产品确实能解决实际问题。

And the reason this got us excited about starting a company is these products were actually useful.

这首次证明了我们不该只在实验室里研究AI，是时候构建现实世界的系统了，确实有很多实用价值可以创造。

And this was the first existence proof of we shouldn't be working on AI in a lab, it's time to actually build systems out in the real world and there's really useful things that you could be doing.

第二个让我们兴奋的因素是规模定律。

The second thing that got us excited was scaling laws too.

我们着迷于这样一个现象：即便这个领域暂时没有新突破，模型性能仍会持续提升。

We got excited about how it seemed like even if the field ran out of ideas, the models would get better.

这大概发生在2021到2022年间。

And so this was around 2021, 2022.

后来Cursor这个产品其实源于一次头脑风暴，当时我们对开发多领域适用的Cursor X版本充满热情。

And then Cursor sort of came out of kind of a whiteboard exercise where we were very excited about a Cursor for x for many different spaces.

那是什么意思？

And what does that mean?

我们当时认为，在知识工作的多个不同垂直领域，都会出现一家公司来自动化该领域的知识工作，每个细分领域都会有一家公司。

We thought at the time that there would be, for a bunch of different verticals of knowledge work, the company that automates that area of knowledge work, a company for each space.

而这家公司会做几件事。

And that company, it would do a couple of things.

首先，它会为该领域打造最佳产品，并随着AI技术的成熟与进步，重新定义该知识工作的实际形态。

The first thing it would do is it would build the best product for that space, and it would define what the actual act of that knowledge work looks like as AI matures and gets better.

然后凭借这个产品，它将赢得市场渠道，建立大型业务，获取数据和资本等资源，进而转型为更接近实验室的形态（尽管不是基础模型实验室），开始利用获得的数据来优化底层模型，推动该领域的自主性发展。

And then with that product, it would win distribution, it would win a big business, and it would get resources like data and capital, and then it would back into being something that looks a little bit more lab like, though not a foundation model lab, where it would start to use the data it gets access to, to actually work on the underlying models and kind of push the autonomy in the space.

而这又会反过来推动产品进步，改变最佳产品的形态，形成良性循环。

And then that would then in turn push forward the product and change what the best product looks like, you get this flywheel going.

所以我们对此非常、非常、非常感兴趣。

So And we were really, really, really interested in that.

我们认为微软会在编程领域实现这一愿景。

And we thought that Microsoft would do that for coding.

而我们想选择一个相对冷门、竞争较少的领域。

And we wanted to work on a sleepy, more or less competitive space.

我们有些同事是机械工程背景，对CAD系统很熟悉。

And we had some colleagues who did mechanical engineering and we were familiar with CAD systems.

于是就有了这个——没错，最初我们其实是从机械工程起步的，开发帮助人们在CAD系统中提高效率的模型，同时也构建自己的CAD系统。

And so there was this, yeah, initial full start of working on mechanical engineering actually, and working on models to help people be more productive within CAD systems and also building our own sort of CAD system.

这就是我们的创业始末。

So that was how we got started.

这是个糟糕的主意，创始人市场契合度极差。

It was a bad idea, the founder market fit was horrible.

我们遇到了盲人摸象的问题，与Mackees通话时询问他们日常做什么，但我们始终无法真正凭直觉理解。

There was this blind man and the elephant problem where we would hop on calls with Mackees and ask them what they do during their days and we only we never really had an intuitive sense for it.

我几乎希望在那六七个月的工作期间，我们能直接去公司实习，真正了解这个领域。

I almost wish that in that kind of six, seven months where we were working on that, we had just gone and been interns at a company to really learn the space.

但最终我们搁置了那个想法，重新回到我们真正最感兴趣的编程工作上。

But eventually we put that idea aside and kind of came back to the thing that we were really most interested in, is working on programming.

我有个理论，想听听你的看法——关于Cursor早期为何表现优异，其实原因很平庸：当时我们调研这个领域时有很多公司，它们做各种不同的事，很多都像科幻小说。

So I've got this theory, I would actually like to hear your views on it, on why Cursor did so well early on, and it's actually pretty banal, which is at the time we were surveying this space and there's a lot of companies, and they were doing a lot of different things, and a lot of it was pretty science fiction.

我们要创造一个软件工程师代理，要用新技术训练模型，要重写编辑器等等。

We're gonna create an agent that will be a software engineer, we're gonna create a model using this new technique, we're gonna do all the things, we're gonna rewrite the editor, etcetera.

我认为Cursor早期成功的原因之一是你们极度专注，选择了VS Code，当时Copilot已培育市场多年，正是这种极致专注和远超同行的产品力造就了成功。

And one of my theories why early on Cursor did well is you were incredibly focused, you chose Versus Code, Copilot had matured the market for a few years at the time, and it was this narrow focus and just a way, way, way better product that did it.

所以有两个问题。

And so two questions.

嘿。

Hey.

你觉得这个观点靠谱吗？

Do you think this is legit view on it?

第二个问题是：当其他人都在四面出击时，你们是如何保持专注的？

And then the second question is, like, how did you decide to maintain focus when everybody else was doing everything else?

因为那正是该开发代理或模型的时期。

Because it's it was the time to build the agents or build the model.

是的。

Yes.

我确实认为这其中有很多道理。

I definitely think that there's a lot of truth to that.

我认为还有一个重要的附加说明，那就是这家公司的故事尚未完结，还有太多事情要做。

I think that also there's an important asterisk in that the story of this company is still yet to be written, and there's so much more to do too.

哦，当然。

Oh, sure.

当然。

Of course.

是啊。

Yeah.

是啊。

Yeah.

不。

No.

如果你

If you

能达到这个程度，我是说目前的成功，

get to this point, the success to this point, I mean,

曾有

there was

一股强劲的上升气流。

just such an updraft.

是的。

Yeah.

是的。

Yeah.

对。

Yes.

回到我们研究CAD相关工作时，那个领域的冷启动问题比我们当前领域困难得多。在帮助人们提高生产力、构建机械工程模型用于现实制造方面，现成的模型都表现不佳。

So going back to when we were working on the CAD stuff, the cold start problem in that space was much harder than in our space, where to get started on helping people be more productive and building MechE models of stuff that they were gonna make in the real world, none of the out of the box models were good at that stuff.

当时实际上缺乏优质的三维开源模型资源，更不用说具有迁移学习能力的模型了。

Like there was actually like no good three d representations to like open source three d models that had transfer.

如果你直接使用现有的基于文本的大语言模型来尝试处理CAD任务，它们根本达不到要求。

If you took the existing text based LLMs and you tried to get the LLMs to be good at CAD, they weren't really.

因此我们在CAD项目上投入了大量时间，除了与机械工程师开那些我们其实不太懂他们日常工作的会议（这显然是个大问题），主要精力都花在了建模工作和数据抓取上。

And so much of our time spent working on the CAD idea, in addition to calls with Mechies where we didn't really understand what they did in their day job, which was obviously a big problem, it was spent doing a lot of modeling work and a lot of data scraping work.

以至于当我们决定暂停这个项目转向编程领域时，团队都产生了某种创伤后应激障碍。

And we were we kind of had PTSD from that when we decided to put it aside and work on programming.

所以初期阶段，我们确实高度专注、行动极其迅速，通过不断叠加技术方案，以最快速度推出产品并积累市场动能。

And so initially, yes, we were super focused, we were super expedient, and we did hack on hack on hack to just get something out into the world as fast as possible and start to get some momentum.

部分原因在于我们虽然有些资金支持，但远不如现在的种子轮融资规模。

And part of that was we didn't have we had some funding, but nothing like the seed rounds of today.

要知道我们有四位联合创始人，虽然讨论过招聘和团队扩张，但说实话我们当时还在学习如何有效管理团队。

And, you know, we had four cofounders and still you know, we talked about hiring and expanding the team, but I think we were still really fully learning how to do that.

当时的竞争格局确实如此：微软和数十家创业公司同场竞技，这些初创企业分属不同赛道——有的直接尝试构建大基础模型，有的提出各种改变工作流程的宏大产品构想，而我们只求以最快速度推出产品。

And so yes, the competitive landscape at the time, it was Microsoft, it was dozens of startups, these startups fit into a bunch of different buckets, there were some that were immediately trying to build big foundation models, there were some that had highfalutin product ideas of like very different changes in people's workflows, and we just tried to get something out as fast as possible.

我记得当时，对我们来说，承诺机制实际上是每月的投资者更新，虽然那时可能根本没人看。

And I remember at the time, the like commitment device for us was actually the monthly investor update, which which probably no one read at the time.

但从决定开发Cursor的第一天起，大概只用了两周时间，我们就有了一个自己使用的IDE。

But it was, I think from day one deciding to work on Cursor, it was a couple weeks to actually have an IDE that we used ourselves.

最初我们甚至没有基于VS Code进行分叉，而是完全从零开始构建。

And initially, we didn't even fork Versus Code, we actually built from scratch.

所以我们从头打造了一个日常自用的IDE，又花了几周时间才让其他人用上。

So we built IDE from scratch that we used ourselves as a daily driver, Couple more weeks to actually get into other people's hands.

总共大概几个月内，我们就向互联网发布了首个测试版，立刻引起了人们的兴趣。

And then in the span of total, I think a couple of months, we had launched our first beta out to the Internet, and immediately it started to get interest from people.

这顺势就形成了发展势头。

And then that kind of set off the momentum.

具体来说，在势头形成的同时，同领域的许多同行正在快速扩展业务范围。

Specifically, while the momentum was building, a number of the people in the same space were broadening very quickly.

比如他们很快转向命令行工具，或是集成IntelliJ之类的产品，而你们选择不这么做。

Like, they'd go to CLI very quickly or they would, like, integrate with IntelliJ or whatever it was, and you decided not to.

这是经过深思熟虑的决定，还是单纯因为你们已经忙得不可开交？

Was this, like, super intentional or was it just, you know, you're getting pulled on the right one, like, you had enough work to do?

是的。

Yeah.

这些想法确实是经过考量的，因为我们基本上全天候都在工作。

The ideas were intentional in that we kind of just worked all the time.

四位联合创始人每天一起吃早饭、午饭和晚饭。

And so the four cofounders every day would be breakfast, lunch, and dinner.

我们要讨论什么？

What are gonna talk about?

你们要无休止地辩论下去吗？

You're gonna debate endlessly?

这些核心战略问题是：你们要开发编辑器还是扩展程序？

These core strategic questions of do you build an editor or an extension?

你们在模型方面有什么动作吗？

Do you do anything on the model side of things?

是的。

Yeah.

还有最初的产品代理。

And other the initial product agents.

构建一个新想法。

Build a new idea.

对。

Yeah.

没错。

Yeah.

是的。

Yeah.

我认为我们当时非常、非常明确地想要掌控这个平台。

I think that we were really, really intentional about wanting to own the Surface.

所以当时（虽然现在不算很有争议），但那时人们觉得做编辑器非常奇怪。

So at the time, it's not super controversial now, but at the time, people just thought it was very weird to do an editor.

无论是否分叉，他们都说你无法让人们更换他们的信用编辑器。

Whether it was a fork or not a fork, they said you can't get people to to switch their credit editor.

人们对此太过依赖，我们知道这是错误的。

They're too tied to it, which we knew was wrong.

因为我们实际上因为Copilot而切换了Versus Code生态系统。

Because we had actually switched the Versus Code ecosystem because of Copilot.

没错。

Right.

我们当时都像卢德分子一样使用命令行VIM，所以我们知道如果能打造更好的解决方案，就能让人们转换。

We were all like Luddites using command line VIM, and so we knew that if you built a better mousetrap, you could get people to switch.

门槛会很高。

The bar would be high.

是的。

Yep.

然后，是的，我们非常有意在未来最终触及模型层面，这背后有一整套故事，实际上这对我们来说是一个非常重要的产品杠杆。

And then, yeah, we were very, very intentional about eventually in the future we wanna touch the model side of things and there's been a whole story of backing into that and that's actually been a really important product lever for us.

但我们不想从那里开始。

But we didn't wanna start there.

我们只是想先向世界推出一些东西，不涉及任何建模内容。

We wanted to just get something out to the world, not touch any of the modeling stuff.

太棒了。

Awesome.

好的。

Okay.

所以我跟你讲过这个轶事，你说你不记得了，但我记得很清楚，早期都是关于规模扩张的事。

So I I told you this anecdote and you said you didn't remember it, but I remember it very well, which is the early days were about scale.

这些年来我见过很多公司。

And I've seen a lot of companies over the years.

我在这个行业已经三十年了。

I've been the industry for thirty years.

我从未见过如此快速的规模扩张，尤其是一个小团队做到的。

I've never seen scale like this quickly, especially with a small team.

我记得有天晚上接到你电话，你说我们搞垮了一个大云平台，因为他们处理不了我们的流量。

And I remember one night I got a call from you and you're like, listen, we've taken down one of the big clouds because like they can't handle our scale.

其实那次服务中断相对轻微。

And there was this actually relatively minor service disruption.

不过你们团队很快就修复了问题。

And then you guys fixed it actually pretty quick and it was fine.

但据Oscar说，就在那段时间有人跑到Cursor办公室，在窗户上贴了个iPad写着'Cursor挂了'。

But apparently in that time or so Oscar tells me someone showed up at the cursor office and put an iPad on the window and says cursor's down.

显然当时影响已经大到引起用户注意了。

So like definitely it was like to a point where people are noticing.

对我来说这很震惊，因为他们居然找到了那栋不起眼的办公楼。

And it was for me, it was kind of a shock because it's kind of this nondescript building like that they found out.

是啊。

Yeah.

所以很想听听你和团队是如何应对这种规模压力的，毕竟你们现在连依赖的顶级平台都开始不堪重负了。

So it would be great to hear how you think and the team thinks about handling this much scale, especially because, I mean, you're really at the point that you're, like, you're even stressing the platforms you rely on even though they're some of the largest platforms.

我是说，已经无路可走了。

I mean, there's nowhere to go.

是啊。

Yeah.

是啊。

Yeah.

那个轶事已经湮没在

That anecdote is lost within the

众多之中的众多。

Of the many of the many.

是啊。

Yeah.

想当年。

Back in the day.

是啊。

Yeah.

是啊。

Yeah.

我觉得早期我们面对规模扩张时，团队实在太小了，却要运营一个快速发展的服务。

I think that well, early on, the way we encountered scale was just it was such a tiny team operating a service that that started to grow very fast.

我的联合创始人们都很优秀。

And my I my cofounders are are great.

不过你知道，我们并不是经验最丰富的团队——从从业年限就能看出来。

And but, you know, we're not the the most experienced group, if you can't tell, just in terms of of years of experience.

很快，就有大量用户开始使用这项服务。

And so, very quickly, you know, we had lots of people using the service.

特别是像我们自有的文件同步系统，你可以把它想象成Cursor内部有两三个不同的小型Dropbox，早期Cursor里还有个AI搜索引擎。

There's ways in which, especially with things like we have our own file sync system that you can think of it as like, there's kind of two or three different sort of mini drop boxes within Cursor where early on within Cursor there's kind of like a search engine for the AI.

表面上看这似乎不应该那么复杂，但实际构建起来却相当麻烦。

And it seems like a kind of on the surface it doesn't sound like it should be that complex, but it ends up being kind of annoying to build.

根据构建方式的不同，确实会开始对你依赖的系统造成压力。

And depending on how you build it, definitely can start stressing stressing the systems that you rely on.

但很快，我们在普通云服务业务方面就达到了相当大的规模。

But yeah, very quickly we were we're getting to a scale when it came to just normal boring cloud services stuff.

当时的情况是，我们运行着一个比许多公司都庞大的Kubernetes集群，全公司只有五个人，还要边做边摸索，期间遇到了不少问题和麻烦。

And so there was a whole story of, you know, we were running a very, very large Kubernetes cluster, larger than many other companies, and then trying to figure that out on the fly with five total people at the company and having things, you know, having hiccups and troubles with that.

后来我们通过做出正确的架构决策和扩大团队，总算掌握了局面。

Then we sort of just got a handle of that by making some of the right architecture decisions, growing the team.

接下来遇到的主要扩展问题其实是给API供应商造成了压力。

Then the next big scaling problem that came up was actually just stressing the API providers.

是啊。

Yeah.

这倒不是靠技术上的小聪明就能解决的，更多是关系处理问题。

And that was less a being very clever technically to get past that scale, and that was more a relationship thing Yeah.

这些API供应商可能真不知道该怎么看待我们，毕竟我们只是四个二十多岁的年轻人，但现在我们的业务却占了他们API收入的很大比例。

Where, you know, these I I don't think the API providers really knew what to make of us because it's, you know, these four twenty somethings, and their thing now comprises like a really high double digit percent of their API revenue.

现在他们不得不进行容量规划决策，甚至可能需要融资决策来应对这种幕后增长。

And now they're gonna have to make capacity planning decisions decisions, maybe financing decisions to, you know, handle the growth under the hood.

是的。

Yeah.

那更多只是...我觉得这是我们仍在学习的过程，你知道的，就是与人建立关系。

And that was more of just a and I think it's something we're still learning, you know, forging relationships with people.

我们还发现一个很聪明的做法——原来这些API令牌，你可以从多家供应商那里为同一个模型获取。

It was also getting very clever about turns out these tokens, these API tokens, you can get them for the same model for many providers.

市面上存在令牌转售商。实际上，战略性地将资源分散到签有承诺合同的多个供应商是有帮助的。

There are there are token resellers that exist out And it's strategically helpful actually to spread it across multiple providers which have committed contracts.

所以我们变得非常擅长搜寻世界上所有现存的SONNET令牌。

And so we got very good at hunting out all the all the SONNET tokens that exist in the world.

因此那个规模层级对我们来说确实很棘手。

And so that was that was a level scale that was tricky for us.

我应该说，目前我们自己做了相当多的训练工作。

I I'd say right now we do a decent bit of our own training.

我们也自行进行部分推理。

We do some of our own inference.

所以现在，可以说，规模问题又有了全新的维度。

And so there's, like, now a whole, you know, side of the scales.

那里存在一个全新的规模难题，需要做出相应决策。

There's there's a whole new scale problem there and and making decisions there.

你觉得

Do you think

这会最终形成对第三方服务的异构依赖吗？

that this converges on, you know, heterogeneous dependency on third parties?

或者你认为最终会主要依靠你们自己运行的基础设施？

Or do you think it converges on largely you're running your own infrastructure?

还是说，你们还没考虑到那么远？

Or like, have you not gotten that far?

是指底层模型推理吗？

For the underlying model inference?

对。

Yeah.

不。

No.

不。

No.

不。

No.

不。

No.

不只是基础设施这么简单。

Just not just just just infrastructure in general.

就像，越来越多情况下你们会把东西收归内部，就是为了能掌控它。

Like, more and more, you're pulling stuff in house just so that you have control of it.

仅仅是为了运营我们的

Just for operating our our

网站、桌面应用、后端这些吗？

website, desktop apps, back end, things like that?

是的。

Yeah.

我认为我们从一开始就采用了多云策略，所以我们显然正走在依赖多家供应商的异构化默认路径上。

I think we've been pretty multi cloud from the start, and so I think we're definitely on a default path to heterogeneous, rely on multiple providers.

我们使用Databricks和Snowflake，同时在AWS、GCP和Azure上都部署了自建或网络服务。

We have Databricks, Snowflake, we have we're on AWS, GCP, and Azure for self or web stuff.

我们采用PlanetScale作为数据库，整个系统——你知道，那些关于扩展的无趣云服务问题——其实都高度依赖数据库，比如整个Kubernetes集群出问题，核心DNS服务宕机之类的。

We use PlanetScale for databases and have had our whole, you know, one of the scaling the kind of boring cloud services stuff was was really reliant on our DB where there was a whole Kubernetes side of things, things like core DNS going down.

然后还遇到一连串数据库问题，因为我们很多业务都是数据库密集型的。

Then there was a whole series of DB stories where some of the things we're doing are, like, pretty DB heavy.

最终我们意识到，通常最简单的解决方案就是直接扩展RDS实例。

And eventually we got to a point where, well, usually you should just scale the RDS instance.

这个方案长期效果不错。

That works well for a long time.

但终究会遇到瓶颈，这时候就要考虑是否该做数据库分片了。

Eventually you run out of that, and then it's like, do you shard the database?

没错。

Yeah.

后来我们转用了AWS的服务，他们宣称可以避免数据库分片。

And then we switched to AWS's service, which claims to not let you shard the database.

结果证明这种说法是错误的。

Turns out that's wrong.

并不尽然。

Not so much.

是的。

Yeah.

你以为这些公有云平台已经解决了一切问题，但实际上只有极少数客户能达到那种顶级规模，他们也是在实践中不断摸索。

You think of these public clouds as they have it all together, but really it's a very small set of customers for the highest highest level of scale, and they're figuring it out on the fly.

因此PlanetScale在这方面表现惊人，我们从无限制过渡到了PlanetScale。

And so PlanetScale has been amazing there, where we went from limitless to PlanetScale.

Sam Sam，你在吗？

Sam Sam, are you here?

非常感谢你，Sam。

Thank you very much, Sam.

我们非常感激。

We appreciate it.

我们所有开发者。

All us developers.

为了Sam。

For Sam.

不过确实。

But yeah.

不。

No.

对我们来说，我认为多家供应商各有所长，这就是我们的策略。

For us, I think multiple providers multiple providers are are great at different things, and so that's our plan.

在转向人才话题前简单说一句。

Just quickly before going towards talent.

所以你必须在专注力上取得平衡，而这正是你的强项。

So you've had to balance focus, which you're very good at.

自那以后，你们做了很多多产品相关的工作。

Since then, you've done a lot of multiproduct stuff.

对吧？

Right?

你们做了Bug机器人。

You did bug bot.

你们做了CLI。

You did CLI.

你们正在进行基础设施改进。

You're doing infrastructure improvements.

在多大程度上，做这个决定是自然而然且显而易见的？

To what extent is the decision to do this is pretty organic and just obvious?

又在多大程度上，你们会以更审慎的方式进行优先级排序？

And to what extent do you kind of do prioritization in a more deliberate way?

或者能否简单讲讲，考虑到你们当前处理的所有事项，你们是如何决定将研发资源投入哪些领域的？

Or maybe just walk through how you think about kind of where to expend R and D resources given everything that you're you're dealing with?

这是经过深思熟虑的。

It's pretty deliberate.

我们会对很多事情说不。

We try to say no to lots of things.

但我确实认为未来我们需要成为一家多产品公司。

But I do think we're gonna need to be a multi product company going into the future.

是的。

Yeah.

我认为在我们的领域存在一个巨大的多产品机遇，可以构建一整套AI编程工具包。

I think that there's a big multi product opportunity in our space where there's a whole AI coding bundle to be built.

对我们许多客户而言，我们希望成为他们首选的AI编程服务提供商。

And we kind of wanna be the, for many of our customers, the AI coding provider for them.

目前我们主要聚焦于编辑器这个切入点——也就是工程师日常开发软件时使用的操作界面。

And so far that's really focused on this wedge in, which is the surface that you sit in, the pane of glass that you sit in when you're an engineer going about your day, building software, which is the editor.

我们相信在这方面仍有大量工作要做，这也是我们的核心重点。

We think that there's still so much more to do there, and that's the main focus.

这正是我们投入资源的方向。

That's where we spend resources.

我们认为编辑器中工作方式的变革，也将逐渐影响团队的协作模式。

We do think that the ways in which work is changing within the editor start to affect how teams work together too.

因此我们相信这既蕴含着重大战略机遇。

And so we think that that presents both a big strategic opportunity.

就像要打造最佳编辑器，就必须配备能辅助团队评审和协作的配套功能。

It's also just like necessary to have the best editor thing is to also have this compliment that's, you know, helping teams review and collaborate a little bit more.

所以我们对此有着明确规划。

And so we're intentional about it.

我们仍在探索最佳实践——比如如何为这类项目提供发展空间，如何把握我们领域中巨大的交叉销售机会，无论是从增长工程、产品导向增长的角度，还是从赋能销售团队的角度。

It's been we're we're still, I think, learning how to how to do it well, like how to give projects like that air cover, how to do cross sell where there's really, really big cross sell opportunities in our space, both from a, like, growth engineering PLG, show them the button side of thing, and then, you know, enabling the sales team.

哦，我想说的是，很多创始人都低估了从单一产品转向多产品时在市场推广方面面临的挑战。

Oh, I will say there, like, many founders underappreciate how tough it is to go from single product to multiproduct when it comes to actually go to market.

我是说，这非常非常复杂。

I mean, it's it's very, very complex.

是啊。

Yeah.

而且我们在这方面还有很多要学习的。

And a lot we're still learning there.

但是，你知道，我们对初步成果感到非常非常兴奋

But, you know, very, very excited by kind

。

of the early results.

太棒了。

Awesome.

好的。

Okay.

那么我想转向人才话题。

Then I'd like to shift towards talent.

我认为你们拥有我所见过的最严格、最周到的人才招聘流程之一。

So I think you have one of the more rigorous and thoughtful talent hiring processes I've ever seen.

比如，我会特意留出晚上和周末的部分时间来帮你们沟通和招募人才。

Like, I try to, like, reserve a part of my evening and weekends to help you to talk to and recruit people.

而且每次参加这些电话会议前，我都会收到一份考虑极其周全的资料，说明当前进展和已完成的工作。

And every and before I hop on every one of these calls, I get this incredibly well thought out, like, here's where it is, here's what we've done, you know.

我是说，我觉得这个流程背后蕴含了太多东西。

I mean, I I just think there's so much behind this process.

你不介意的话，能否详细介绍一下你对招聘的看法，以及你如何执行流程，发现了哪些有效和可能无效的方法？

So you wouldn't mind, can you just kind of walk through how you think about recruiting and kind of how you run your process and what you found out what works and maybe what doesn't work?

好的。

Yeah.

让你的董事会成员打无数电话，直到他们受不了求饶为止。

Have your board members do lots of calls until they cry, uncle.

准备

Prepare

他们出局。

them out.

是的。

Yeah.

充分利用他们的时间。

Take advantage of their time.

对。

Yeah.

没错。

Yeah.

我们是怎么考虑招聘的？

How have we thought about recruiting?

我认为我们的流程在某些方面相当正统。

I think that there are ways in which our process is pretty orthodox.

我觉得一些独特之处在于，通常在小公司时，你会对首批工程师采取这样的做法：基本就是让人以合同形式合作，可能不会进行常规的代码主导面试流程。

I think that some of the things that might be unique, one is normally when you're a small company, there's this thing that you do with the first set of engineers where you basically just have people contract with you and you probably don't do a normal lead code style thing, a normal interview loop.

我们就是这么做的。

That's what we did.

这感觉最自然，因为你能真正了解自己与对方是否合拍。

It felt the most natural because you're kinda getting to the ground truth of do you work well with the person?

通常人们在招了几个人之后就会停止这种做法。

And then usually people stop it after a couple of a couple of hires.

我们尝试过多次在内部取消这个制度。

We have and we try to kill it many times internally.

我也曾试图取消它。

I've tried to kill it too.

我们现在某种程度上还在这样做——每个加入工程团队和设计团队的新人都会在办公室待两天，做一个项目。

We still kind of do that where everyone who gets hired on the eng team and the design team spends two days in the office, and they work on a project.

形式非常自由。

And it's very free form.

不是那种排满两天的白板面试接一个又一个。

It's not like, you know, you have this whiteboard interview and then that whiteboard interview and your days two days are packed.

就是给你一张桌子、一台笔记本，有三个可选项目，还有一份冻结的老版本代码库和开发环境配置。

It's here's a desk, here's a laptop, you know, here's three projects you could work on, here's a frozen version like of an you know, a frozen older version of the code base with the dev x setup.

直接上手做就行。

Just go do it.

这个机制其实有两个作用。

And then you this functions this has kind of two functions.

首先我认为这是个很棒的测试，它能检验那些常规编码面试测不到的维度——看候选人能否在代码库中完成端到端开发。

So one function is I think it's a really great test that tests for orthogonal things to the normal coding style interviews that we we ask before people get on-site, where you're seeing, you know, can they go end to end in the code base?

比如，他们是否具有主动性？

Like, are they agentic?

我们的引擎设计与产品紧密相连，因此我们倾向于招聘具备产品意识的产品工程师。

Our engine design and product are pretty tightly coupled, and so we try to hire product engineers who have product sense.

这让你能感受到，如果让他们独自在没有团队的环境下，他们会构建出什么？

This gives you a sense of that, you know, what would they build if left in a vacuum without a team?

所以我认为这确实为我们提供了大量关于在我们环境中取得成功所需的原始技术能力的信号。

And so I I think it really gives us a lot of signal on the raw technical skills needed to be successful in our environment.

另一个作用是它同时充当文化面试，你会和我们共进四到六餐。

The other thing that it does for us is it also functions as a culture interview where you have four to six meals with us.

你知道，这让我们能感受到我们是否愿意与你共事。

And, you know, that gives us a sense of would we wanna be around you?

你愿意和我们共事吗？

Do you wanna be around us?

如果要说其中一个好处，或许排在第三位的是它确实为候选人提供了大量关于公司及入职第一天会是怎样的信息。

If one of the benefits, you know, maybe at sub point sub point third benefit is it really gives the candidate a ton of information about the company and what it's gonna be like to show up on the first day.

我认为这导致了，如果他们同意加入，他们那边也会有极高的人岗匹配度。

And I think that that's led to, you know, really, really, really high chance of fit on their end too if they say yes.

所以这是我们做的比较非传统的事情之一——这个为期两天的现场面试，即使我们现在已超过200人，我们仍坚持这么做。

And so that's that's one of the more unorthodox things we do is we have this two day on-site, and we've clung to it even though we are over 200 people now.

但你们不会对市场推广或其他部门这样做吧？我们最初确实这样做过。

And but you don't do this with, like, go to market or other We did initially.

是啊。

So yeah.

招聘我们的第一位销售人员时，

To hire our first sales guy,

像是的。

like Yeah.

所以，是的。

So to yeah.

所以招聘第一批销售代表时，我们会给他们分配入站销售线索。

So to hire our first reps, we would give them we're like, here are inbound leads.

这很棒。

That's awesome.

你有配额指标。

You have a quota.

是的。

Yeah.

当时，是的，当时结构更明确些，你知道，他们会做产品演示。

It was, yeah, it was a little bit more structured where, you know, they would do a demo.

他们会做一些模拟客户沟通。

They would do some, like, mock customer communications.

对。

Yeah.

但我们会让他们接触真实数据并深入分析。

But we would give them access to the real the real data and have them dig in.

我记得最早最早最早那次，真的就是销售代表直接上岗。

I think the very, very, very first one we did was was literally, like, the rep came in.

我们向他们展示了一切。

We showed them everything.

我们说，教教我们该怎么搞销售。

We're like, teach us how we should do sales.

是啊。

Yeah.

但后来事情开始变得更有条理了。

But then it started to get more structured.

好的。

Okay.

太棒了。

Awesome.

很好。

Great.

听着，我认为这波浪潮总体上正在改变很多关于如何创建公司的传统观念。

So listen, I think this wave in general is changing a lot of orthodoxy on how you build companies.

我的意思是，这完全是一个新的超级周期。

I mean, it's just it's a new super cycle.

就像，你知道的，我们正在质疑一切。

Like, you know, we're questioning everything.

你绝对站在了这股浪潮的最前沿。

You're definitely on the forefront of that.

我是说，你们让相对资历尚浅的人管理庞大的组织，而且效果出奇地好。

I mean, you've got, you know, relatively, I would say, junior folks running very large orgs, and it's working out incredibly well.

还有一件事你们正在做的，我甚至可以说做到了近乎极致的程度，就是并购。

Another thing that you're you're you're doing, I mean, I would say almost to like an extreme amount is is is M and A.

对于一个成立仅两年的公司来说，你们非常擅长进行这种小型并购整合。

Like, you've been very, very good at doing these kind of tuck ins for a two year old company.

对吧？

Right?

我的意思是，显然很多私营企业都会收购公司，但你们在这方面做得非常出色。

Like, I mean, clearly a lot of private companies acquire companies, but you've done a great job about that.

你愿意分享一下你对这方面的思考吗？

Would you be open to sharing kind of how you think about this?

在人工智能时代之前有句老话：初创公司永远不该收购其他初创公司。

I mean, like, the adage pre AI was like, startups should never buy startups.

但实际上这种做法非常成功，不仅对Cursed如此，整体来看都是这样。

And it's actually been hugely successful, not just with Cursed, but across the board.

所以我很想听听这对你们是如何奏效的。

And so I think it'd be great to hear how how that's worked for you.