迈克尔·特鲁尔：23岁创立Cursor，挑战GitHub Copilot，给工程学学生的建议

我们意识到自己内心对编程的未来充满期待。经过深思熟虑后，我们认为如果真正坚持自己的信念，未来五年整个编程领域将迎来变革，软件开发流程都将由模型驱动。当时感觉这个领域的工作者都没认真对待这个趋势——他们虽有优秀产品并持续优化，但并未真正瞄准'彻底自动化现有编程模式，重塑软件开发形态'的远景。正是基于这个认知，我们决定投身于此。

We realized we were really inherently excited about the future of coding. And I think we took a step back and realized that if we were being really consistent with our beliefs, there was gonna be an opportunity for all of coding to change in the next five years and for all of software development to flow through models. It felt like no one working on the space at the time was really taking that seriously. It felt like they had great products and they were making them a bit better, but they weren't really aiming for a world where, you know, all of coding as we know it today gets automated and building software ends up looking very, very different. Then with that in mind, we set out to to work on that.

让我们从你的创始人历程起源开始聊起。这要追溯到中学时期你阅读PG（保罗·格雷厄姆）文章的经历对吧？

Let's start this talk with sort of the origin story of your journey as a founder. You kinda have to go way back to middle school when you were reading the essays from PG. Right?

其实我很早就对创业产生兴趣，当然也涉猎过其他领域。最初接触编程的契机，是某个寒假我和哥哥想开发爆款手游。当时完全不懂技术，只能谷歌搜索'如何制作游戏'，结果发现需要下载Xcode。打开后满屏都是晦涩难懂的Objective C彩色符号——这门语言现在虽仍存在但确实没落了。面对这堵'Objective C高墙'，我哥哥立刻放弃了编程，如今他走上了截然不同的职业道路。

So early on, I think, you know, had been interested in in starting a company for a long time, had been interested in a bunch of a bunch of other things too. I think, actually, I originally got into programming, being interested in in starting something kind of commercial, where, the first time that I ever saw code, it was over some winter break. And my brother and I, we wanted to create a hit mobile game. We didn't really know how to do that. We looked on Google.

他现在尝试往绘画方向发展。而我买了本Objective C教材坚持学习，最终开始制作手游，这就是我的编程启蒙。期间我也深受PG和Sam的创业文章影响，YC社区成员的思想从高中早期就给了我巨大启发。

How do you create a game? We heard that you need to download this application called Xcode. We did that, and, we were hit with these weird colorful esoteric symbols, which were Objective C, which, you know, is still around, but maybe a little bit less popular than it was then for good reasons, and, stared at, this kind of impenetrable wall of objective c. And my brother promptly ejected, didn't move on with programming. He now is on a very different career path.

他目前在尝试往绘画方向发展。而我坚持了下来，买了本Objective C教材，后来真的开始开发手游——这就是我的编程启蒙。过程中我也确实是PG和Sam文章的忠实读者，YC社区许多人的思想从高中早期就给了我巨大启发。

He's kind of trying to paint for something like that. But, I, yeah, kept going and, bought a book on objective c and then eventually, started working on on mobile games. That was the genesis of me getting into programming. And then along the way, also, yes, was a big fan of, PG's essays and, Sam's essays too also and a bunch of the folks in YC. And that was definitely a big inspiration even from the very early stages of of high school.

Cursor最令人惊叹的是你年仅24岁就在短时间内打造出这个巨头企业。外人看来可能觉得是横空出世，但实际上这是你十余年积累的成果。你一直在推进各种项目对吧？高中时就开始接触AI了？

I think the wildest thing about Cursor is that right now you're just 24 and build this monster of a company in a really short amount of time. To a lot of people, it could seem that it's a bit of out of out of nowhere, but this was really in the making for more than a decade. You've been working and shipping a lot of different projects. Right? And you were working in AI even when you were in high school.

能具体说说那些项目经历吗？你是如何入门的？

Right? Tell us a bit about the projects and how you got started with that.

我很幸运早年就接触了编程和AI，还有优秀的合作者。手游尝试失败后，我做了个技术简单却爆火的应用——可以伪造《钢琴块》《Flappy Bird》等游戏高分记录发给朋友。这个病毒式传播的项目让我明白创业不一定要技术艰深。

I was lucky enough to find programming early on. I was also lucky enough to be interested in AI early on, and have some great collaborators to work on AI projects with. Soon after kind of the foray into mobile games, which also turned into, I wasn't very good at mobile games. So one of the things that I built and actually one of the things that got most popular, which was kind of the technically easiest thing to build, which was maybe a lesson in startups of, you know, the code isn't everything, was this mobile game or this mobile app where you could spoof high scores in things like piano tiles and Flappy Bird, and then send them to your friends. And that was kind of the thing that went viral.

后来我和朋友想开发能用'训狗方式'教育的机器狗：通过奖惩机制而非编程来训练它。虽然毫无头绪，我们又开始谷歌搜索，由此接触遗传算法——当时有人尝试用其进化神经网络（如Neat项目），最终我们发现了强化学习（RL）。尽管RL在2015年就已发展多年，我们最终还是造出了几台机器人。

It wasn't the, you know, painstakingly handcrafting the game engine yourself type thing. But, yeah, no. Soon after that, got interested with a friend in the idea of building a robotic dog where we thought it would be really great to have a robot that you could teach to do things without programming it. Instead, you could give it positive and negative feedback like you give a dog, so you could give it a treat if it does some you know, quote, treat if it does something good. You would say bad if it does something bad, and then maybe it would you know, you could teach it to play fashion and things like that.

这个构想让我们无比兴奋。完全不懂实现方法的我们再次求助于谷歌，在探索过程中接触到遗传算法——当时有人正尝试用其进化神经网络（比如Neat项目），最终我们发现了强化学习（RL）。即便在2015年RL也已是多年研究领域，但我和朋友最终确实造出了几台机器人。

That idea really animated us. We had no idea how to build it, And so, again, you know, started the place where one would start, which is Google, and kind of went down a lot of rabbit holes and, you know, took us into a place of, learning about genetic algorithms, and maybe that was gonna be helpful for building this robot dog that we wanted to build. And then we eventually learned about this neural network stuff because some people were playing with taking genetic algorithms and using them to evolve neural networks at the time, you know, with work like Neat. And then, eventually, it took us to RL, reinforcement learning, which was you know, even back in 2015, people had been working on it for a long time. In the end, my friend and I, we did eventually build a couple of robots.

我们当时确实没有做任何真正持久性的实质性工作，但我们做了一些在当时很有趣的事情——改进强化学习算法，使其数据效率更高，也就是让它们能从极少量的数据点（比如几十个）以及人类提供的嘈杂数据中更好地学习。虽然那不算真正的机器狗，但我们造了几台机器人，其中一个是多轴机械臂，能挥动球拍打乒乓球。装上合适的传感器并给予正负反馈后，就能教会它在看到球时挥拍。我们还教一个基维驱动机器人沿线追踪。这段经历堪称绝佳的机器学习教育，部分源于我们天真的无知——当时根本不知道有Torch、TensorFlow这些现成的工具库可用。

We did we didn't do any sort of substantial work that really lasted, but we did work that was interesting at the time in taking reinforcement learning algorithms and making them more data efficient, you know, making them better at learning from very, very few data points, you know, order of tens of data points, and also from noisy data, you know, data that a human's giving. It wasn't exactly a dog, but we built a a couple of robots where one of them was this many axis robot arm that could kind of swing a paddle and play ping pong. And if you put the right sensor on it and then you gave it the right sort of positive and negative feedback, you could teach it to swing when it sees a ball. And then, we had this Kiwi drive robot that we would teach to follow a line. To do that, it was actually kind of this great education in ML, partially because of our dumb naivete, where we didn't really know that there were things like Torch and TensorFlow and kind other lots of building blocks we could use from.

可能我们谷歌搜索的水平不够高吧。

Maybe we weren't good enough at googling.

你们是从零开始自己实现了神经网络？

You implemented your own neural network from scratch

没错。所以当时...

Yeah. So when

你们那时才十六七岁？

you were, like, I don't know, 16, 17?

问题的限制在于我们用的是微控制器机器人，内存极小，根本装不下标准机器学习库。在试图造机器狗的过程中，我们实现了一个微型神经网络库。我记得我们当时连内部原理和微积分都不太懂，就跌跌撞撞地复现了神经网络的一些核心思想——这段经历让我们受益匪浅。

The constraints of the problem were we were dealing with robots, and so we were dealing with microcontrollers. And so microcontrollers have very little memory, and they couldn't really fit any of the normal standard ML libraries. So as part of our bike shedding trying to build a robot dog, we implemented our own tiny neural network library. And I have memories of us not really understanding any of the internals of how these things worked or not really understanding calculus, but kind of fumbling our way through reimplementing some important ideas from neural networks. I, you know, I think it it taught us a lot.

不过基础知识的空缺后来花了好些年才补上。

I think that there were lot of gaps in the fundamentals that it took many years to fill in later.

快进到创立nnSphere（这名字挺有意思，毕竟现在叫Cursor）。你们创始团队刚从MIT毕业对吧？那是2022年的事。

Then fast forward to the founding of n nSphere. It's a interesting name because Cursor is not what it is. When you guys started, you had just graduated MIT. Right? That was back in 2022.

2022年你们四人最初着手开发的想法是什么？

What were the first idea that all four of you started working on back in 2022?

Cursor的雏形始于2021年。我们几位联合创始人长期关注AI领域——每人都有类似'机器狗'的启蒙时刻：有位合伙人在2021年尝试用语言模型打造谷歌竞品并自训练对比模型；另一位在学术界研究计算机视觉；还有人曾在谷歌这类公司做推荐系统。

Yeah. So the the genesis of Cursor was in 2021. My cofounders and I, we had been interested in AI for a long time. Each of us kind of had our own little robot dog moment where one of my cofounders, he worked on trying to build a competitor, Google actually, using LMs in in 2021 and and training his own, and training his own contrastive models. One of my cofounders, worked on computer vision in academia, and, you know, some of us also worked on recommendation systems at at companies like Google.

但是，我们对AI确实非常感兴趣。2021年时，我们正试图决定如何将这份兴趣付诸实践——是投身学术界的AI研究，加入某个现有的AI巨头团队，还是创立自己的事业？有两个关键瞬间让我们热血沸腾：一是目睹首批AI产品的诞生。

But, we were really interested in AI. In 2021, we were trying to figure out what we'd do with that interest. Do we go and work on AI in academia, or, you know, do we go join, you know, a big existing AI effort, or do we start our own thing? And there were two moments that really got us excited. One was seeing the first AI product start to come out.

GitHub Copilot对我们而言就是标杆案例。另一个是看到研究表明，随着模型规模扩大，AI能力将呈现可预测的提升。2022年初，我和联合创始人进行了为期一个月的编程马拉松，开始探索如何选择知识工作领域，并构建随着AI日益成熟的应用形态。

You know, GitHub Copilot was really the canonical example for us. The other was seeing work about how it looks like AI was gonna predictably get better in the future as you scale up these models. At the very beginning of 2022, me and my cofounders, we went on a, like, month long hackathon, basically. And we started hacking on ideas related to kind of picking an area of knowledge work and building what it looks like as AI gets, more and more mature.

你们为第一个构想收集了大量数据对吧？

You guys have collected a lot of data for that first idea. Right?

没错。我们长期深耕的首个创意是机械工程领域的AI助手——试图训练模型预测用户在SolidWorks或Fusion 360等CAD系统中的操作。选择这个领域是因为我们认为它冷门且缺乏竞争，尽管从开始就是个糟糕决定：我们没人真正是机械工程师，而且该领域的技术条件也尚未成熟。

Yeah. So the first real idea that we worked on for a long time was in mechanical engineering. It was trying to build a a copilot for mechanical engineers and trying to train models to kinda predict what you would do in a CAD system like SolidWorks or Fusion three sixty, which is where, Mechis model out parts in three d on a computer. We picked it because we thought it would be boring and sleepy and uncompetitive, and, we were kind of doing an armchair MBA thing even though it was a horrible choice from the get go because none of us were really mechanical engineers, and also the science wasn't really ready for that area.

但你们坚持研发了好几个月？爬取了大量CAD文件，最终实现了自动补全功能的初级版本？

But you guys kept working at it for a number of months. Right? And you crawled and got all these CAD files and actually got something working with, auto completion. Right? That was, like, the first version of it working?

是的。说实话大量工作耗费在数据抓取上——我们需要全网CAD模型，还要处理各种文件格式转换。CAD软件市场非常碎片化，存在众多流行但互不兼容的系统。

Yes. We, a bunch of the work was in data scraping, honestly. It was trying to get all the CAD CAD models in the Internet. There are also all these different file formats and trying to convert them all into something that's canonical because CAD is this weird software market where there are all these different systems that are pretty popular. It's very fragmented.

云端CAD系统更棘手，它们既不支持文件导出，也禁止爬取。当时的模型训练基础设施相当原始，我们在架构搭建上投入巨大，还要反复试验如何在这些根本不具备扩展性的CAD软件中强行植入插件。

There are also Cloud CAD systems that don't have easily exportable files, and they don't want you to scrape their stuff. And so there was a bunch of work there. Also, the training infrastructure for doing any kind of modeling work back then was pretty rudimentary. And so there was a lot of work on the on the infra side there and just a lot of experimenting with models and a lot of experimenting with how you even jerry rig an extension into these CAD systems because they're the you know, we were building an extension. These applications aren't really extensible at all.

当时我们还有并行项目：两位联合创始人在开发端到端加密通讯系统（其中一位有安全研究背景）。不同于Signal和WhatsApp仅加密消息内容，这个系统会隐藏通讯元数据——对需要保护联系人关系的用户（如记者与线人）至关重要。

There were actually also other projects that we were working on at the time. So two of my cofounders, they were working on an end to end encrypted messaging system, because one of them has a background in security research. And the idea there was apps like Signal and WhatsApp. They encrypt the body of the messages, but they don't hide who's talking to who at at what time, which is actually really crucial information if you don't wanna trust the messaging app provider. So, you know, if a journalist is talking to, you know, talking to some informant in the government, just knowing that they're communicating at all is a is actually, you know, a really big piece of information.

所以到2022年中，这个项目已经推进了半年左右？

So that was in the middle of, 2022. So you guys were working for about good six months on this idea?

对，没错。

Yes. Yes.

当时你们获得了多少用户？产品发布后。

And how many users were you getting at that did you get at that point? So you shipped the product.

所有这些项目都命运多舛，基本上没有用户。

All all of these projects were ill fated and had basically no users. Users.

你们是什么时候意识到这个想法行不通的？就像突然发现，哦不，我们所有人都在为此努力，试图创业，但就是不见成效。

What point did you realize that the idea was not working? It's like, oh, no. We we're all working on this. We we're trying to do a start up. It's not working.

那一刻是什么感觉？

And what what was that moment like?

我觉得每个项目的情况都有所不同。对于那个消息系统，我的两位联合创始人投入了很多，技术层面非常出色，但存在一些难以扩展的弊端。他们尝试向个人用户推广未果，又转向企业市场依然受挫，大概在努力争取市场关注几个月后。

I think it was a bit different for each of the projects. I think for the messaging system that, two of my cofounders worked on, it was really technically impressive, but it had these bad trade offs where it wasn't very scalable. And I think they tried to give people, it and it didn't really work. And then they tried to sell it b to b, and then it didn't really work. And I think it was after a couple of months of trying to get traction.

至于CAD相关创意，我们花了数月时间试图让模型真正对终端用户有用，同时也在思考：我们是否真的对这些领域感兴趣，还是说有其他让我们更兴奋的方向？

For the CAD ideas, it was, yeah, many months of trying to get the models to really be useful for end users. And then also reckoning around, are we really interested in these areas, or is there something else that we're inherently much more excited about?

所以你们有过一个决定性的时刻，认为这些想法行不通，必须再次转型。

So there was a moment that you decided, okay. These ideas are not working. We have to pivot again.

是的。

Yes.

你们在确定代码补足方向前，已经尝试了三四个甚至五个不同的想法？

You you you churn through a three ideas, three, four, five ideas before landing into into a code completion?

没错。我们早期就受到Copilot这类工具的启发，但因为觉得AI编程领域竞争太激烈而回避了——现在想想挺疯狂的，毕竟现在竞争依然激烈。那么变化是如何发生的呢？

Yeah. I think that, so we had been inspired by, tools like Copilot really early on, and we had avoided working on, AI and coding because we thought it was too competitive. Which is crazy. Then, still is competitive now. So how can changes?

2022年时，GitHub Copilot已经创造了约1亿美元的收入。

'22, GitHub Copilot was making already about a 100,000,000 revenue.

我觉得可能更多。很可能更多。是的。

I think Or more. Potentially more. Yeah.

而你们当时的反应是，'我们还能做得比GitHub Copilot更好'，因为人们以为这场竞赛已经结束了。

And you guys are like, oh, we could still do a better job than GitHub Copilot because people thought the game was done. It's

就像，嘿。GitHub...其实我们最初并不认为自己能做到。但后来，经过长时间尝试各种想法却始终缺乏激情，加上屡屡受挫的经历，这些塑造了我们的关注点和目标。我们意识到自己本质上对编程的未来充满热情。

like, hey. GitHub Well, I mean, we we didn't think we could at at the start. And then I think, you know, it was the desperation of having worked on ideas for a while and not really being excited about them after a while and then not really working out. And that kind of shapes, I think, what you care about and what you're aiming for. And we realized we were really inherently excited about the future of coding.

我认为还因为我们观察到同行开发产品的方式，见证了技术发展的轨迹。退一步思考后，我们确信如果坚持自己的信念，未来五年整个编程领域将迎来变革，软件开发都将基于模型运作。而当时业内似乎没人认真对待这个趋势——他们只是在优化现有产品，而非真正瞄准'彻底自动化现有编程模式，重塑软件开发形态'的远景。

I think also we got to see how some of the other people in the space were, you know, working on their products. We got to see how the tech was developing. And I think we took a step back and realized that if we were being really consistent with our beliefs, you know, there was gonna be an opportunity for all of coding to change in the next five years and for all of software development to flow through models. And it felt like no one working on the space at the time was really taking that seriously. Like, it felt like they had great products and they were making them a bit better, But they weren't really aiming for a world where, you know, all of coding as we know it today gets automated, and, building software ends up looking very, very different.

带着这个认知，我们开始了探索之旅。

Then with that in mind, we set out to to work on that.

这是个大胆的决定——你们放弃了那些不够熟悉的领域，尽管面对GitHub Copilot这个巨人，依然选择专注于热爱的编程领域来解决问题。

That was a bold move because you said, okay. We're gonna stop working on all these other ideas that we didn't have as much of a background, and you were excited about programming. Even though you had this big Goliath in the room with GitHub Copilot, you decided to go, and let's just solve this problem.

当时并不觉得多大胆或冒险，毕竟只是几个人窝在客厅对着笔记本，又不是大公司转型。我们最初试探性地想过做安全审查工具来检测代码漏洞，或者针对量化研究员开发专用工具——确实还做过量化研究原型。

It didn't really feel bold or, like, a gagging move at the time because it's, like, you know, a bunch of people sitting around in their living room, like, on laptops. It's not, like, you know, like, pivoting some giant company. But, yeah, no, we did. And, you know, initially, we kind of waded into it where we were thinking, well, you know, maybe we do this kind of very a niche tool for, basically, security reviews, you know, trying to detect future CVEs in your code, or maybe we build something that's just for this one niche area of software. You know, we we thought about building for quants and actually, kind of prototype some things just for quantitative researchers.

但在过程中，关于Cursor如何成为AI编程最佳解决方案的灵感不断涌现。我们对这个方向充满信念与激情，最终决定全力投入。

But, yeah, in in doing that, we were just brimming with ideas for what Cursor could be if it were just about trying to be the best way to code with AI, in general. And then I think that we just we had a ton of conviction about that, and we had a ton of excitement about that. And so at some point, we just decided to to go for it. Yeah.

这个决定是在2022年底做出的对吧？首款产品开发了多久？最初版本是什么形态？听说几周后就上线了，具体是怎样的？

And that was end of, 2022, right, when you decided to make that move? And how quickly did you ship the first product, and what did the first product look like? And that was around you shipped it a couple weeks later, and what was what was that look like?

我们确实花了一些时间才公开推出产品。嗯。从写下第一行代码到开放并正式发布，大概用了三个月。最初，我们是从零开始构建了自己的编辑器，姑且这么称呼它。

It did take us a little bit of time to ship something publicly. Mhmm. It took us roughly, I think, three months from first line of code to open it up and GA it. Originally, what we did is we built our own editor, quote, unquote, from scratch.

天啊。

Oh my god.

尽管如此，它还是使用了许多开源构建模块。像CodeMirror这样的优秀基础组件，以及语言服务器等，有很多开源技术能帮助你构建编辑器。但没错，我们确实是从零开始拼凑起来的——当时还开发了自己的远程SSH版本、自己的Copilot集成功能，因为我们连自动补全这样的功能都没有。

Still it was still using a bunch of open source building blocks. There are a lot of great primitives like CodeMirror and, you know, the language servers, there's a lot of open source tech that can that can help you build an editor. But, yeah. No. It was cobbled together from scratch, and there was our own version of remote SSH, our own Copilot integration at the time because we didn't have anything like autocomplete.

你必须构建自己的PIN系统，必须完成所有语言服务器集成。要让产品在成熟的代码编辑器市场中具有竞争力，成为人们日常使用的主力工具，需要投入大量工作。但大约四周后，我们做出了能作为日常工具使用的版本，又过了四周交给首批测试者，再四周后便正式发布了。

You you have to build, you know, your own PIN system. You have to build all your own language server integrations. There's just a a lot that ends up going into, something as developed as, you know, the code editor market, you know, making something that can actually be competitive there and service someone's daily driver. But it was I think it was four weeks until we built something that we could use as our daily driver. It was maybe four weeks later where we gave it to the first beta testers, and then there was another four weeks, and then we GA ed it.

当时它仍然非常粗糙。向公众开放时并不觉得是件大事。你们...

And it was still very, very crude at the time. It didn't feel like a big thing to just open it up to the public. What did you

在第一个版本中学到了什么？毕竟你们从零构建了代码编辑器，那时还没考虑分叉方案。

learn in that first version? Because you you built a code editor from scratch. You guys haven't done the whole forking yet.

是啊。结果如何？我们被现实震慑了。团队已经很久没做出让人真正满意的产品了，所以这次我们全情投入，高度专注。

Yeah. And what do think? Fear of god in us. I mean, we had people hadn't hadn't really liked some of the things that we had built for a while. So I think that, you know, we were kind of all in on it and very focused.

收获是什么？我们初步掌握了AI功能的开发模式。最初只有一个快捷键命令，调出编辑器里的万能遥控器界面。你发出指令后，AI会自行判断意图——是要返回聊天响应？生成可采纳的代码建议？检索代码库回答问题？还是执行长时间或短时任务？当时几乎没有控制选项。

But what did we learn from that? I think that we learned kind of the first initial set of AI features where, you know, when we started, I think that there was just one key command, and it pulled up this, like, universal remote in the editor. And then you asked it to do something, and then entirely, the AI would just figure out, oh, do you what what what exactly do you want it to do? You know, do you want something back that's, a chat response, or do you want, like, a code suggestion that you can then take, or do you want it to go search around your code base and answer a question, or do you want it to go spin for a really long time or a short time? And there wasn't a lot of control.

通过2022年底的技术实践，我们认识到产品形态必须调整。这些早期AI功能后来成为Cursor的核心部分，这既来自我们自身迭代，也得益于用户反馈。另一个教训是：我们快速构建了理想编辑器的功能完整版，外加自认为出色的AI功能。但要让编辑器功能真正达到世界级水平，道路远比想象漫长——比如TypeScript项目Fiasco经过12年发展，有庞大团队维护。

And I think that we learned, you know, given the tech of the time, at the end of 2022, that you actually it has to the form factor has to look a bit different. And so we learned kind of the first early AI features that then became part of the core of Cursor from iterating both for ourselves and also giving it to people. I think another thing we learned was, you know, we were very rapidly building a feature complete version of what we want in a normal code editor plus then some AI stuff that we thought was great. But then, you know, a feature complete code editor for the world, is gonna be a way, way, way longer road. We thought that, you know, fiasco had been developed over the course of twelve years, was one of the earliest, TypeScript projects, had lots of people on it.

我们曾天真地认为几个月就能做出同等水平的产品。现实很快教育了我们，于是决定专注于AI功能开发。就像浏览器基于Chromium渲染引擎那样，我们最终转向基于VS Code的架构。

We thought, oh, yeah. Of course. You can kind of spin something up that's just equivalent for the world in in a few months. And I think that we learned very rapidly that that wasn't the reality, and our time was gonna be best spent just focused on the AI stuff. And so similar to how browsers often base themselves off of Chromium's rendering engine, we then switch to being based off of, Versus Code.

另外，你们当时也自主研发了一些模型对吧？比如那时候你们从Codex获得了不少灵感，是吗？

The other thing is you guys had also implemented your own models too. Like, back then, you got a lot of inspiration from, Codex. Right?

嗯，是的。当我们最初着手开发第一个投入大量精力的项目——即尝试用AI提升机械工程师的工作效率时，由于市面上的现成模型无法满足需求，我们在首轮融资时就意识到需要资金进行模型训练。我们重点参考的论文之一正是Codex的原版论文，因为根据我们的测算，这个为GitHub Copilot提供支持的初代自动补全模型，其训练成本其实并不高——尽管在2022年上半叶，人们普遍认为AI模型训练极其昂贵。

Mhmm. Yes. So when we were setting out to work on, you know, our our first idea that we really spent a bunch of time on, which was trying to help mechanical engineers be more productive using AI, One of the things when we raised our first round of funding because we we actually kind of needed money from the get go to do a little bit of model training because you couldn't bootstrap it with the models that existed off the shelf if they weren't good enough for that task. One of the papers that we would tap around is actually the original codex paper because by our calculations, Codex, which was the first this was the first autocomplete model behind GitHub Copilot. It didn't really cost that much money to train even though even back then at kind of the beginning and middle of 2022, people were talking about how expensive AI models were to train.

我记得训练成本大约是10万美元（可能记忆有误）。在机械工程领域的探索中我们进行了自主训练，但转向Cursor项目时我们吸取了教训，决定尽可能务实、避免重复造轮子，所以初期完全没有涉足模型训练。直到2023年打磨产品时，这反而成为了关键的产品杠杆——尤其在规模扩张用户激增后。

I think it cost, my math might be wrong, but I think it was about a 100 k in training costs. And then, you know, during this foray into mechanical engineering, we had done our our own training. And then, when we set off on Cursor, I think we were a little bit burned by that. And so we wanted to be as pragmatic as possible, not reinvent the wheel, and so we started by doing none of that. And then over the course of 2023, you know, in dialing in the product, that ended up being a really important product lever, especially as we got to scale and got a bunch of people using the product.

这还赋予了我们利用产品数据优化产品的能力。培养这种能力对公司发展至关重要。

And then that also gives you the ability to use product data to make the product better. And so that actually has been a really important, muscle to build in the company.

YC下一期孵化项目正在招募中。心怀创业梦想？立即申请ycombinator.com/apply。行动永远不嫌早，填写申请的过程就能提升你的创意。好。

YC's next batch is now taking applications. Got a startup in you? Apply at ycombinator.com/apply. It's never too early, and filling out the app will level up your idea. Okay.

回到正片。

Back to the video.

2023年你们还在质疑Cursor能否成功对吧？当时你们创始团队仍在争论是否应该转型，比如'这个创意真的可行吗'，同时还要努力推动增长。

What happened then in 2023 was when you were still not sure about whether cursor was gonna be a thing. Right? You were still debating with your cofounders whether you should still pivot. It's like, oh, is this idea still gonna work? And you're still trying to grow it.

对吧？因为实现营收花了很长时间。

Right? Because it took a long time to to get to revenue. Right?

是的。整个2023年业务确实在增长，但规模有限。我们面临的挑战在于这个领域没有明确的路线图——有些市场可以立即开展用户调研，系统梳理需求并针对性开发解决方案，但我们的情况不同。

Yeah. I think that over 2023, it, it was growing. The numbers were kind of small. And I think that also we were working on something where there wasn't always a clear next step. I think that there are probably some markets where you're really well served by going immediately, talking to a bunch of people, listing down their problems, really rigorously or, you know, really kind of systematically and exhaustively thinking through each problem, what would kind of be the direct solution, and then prioritizing them and then going from there.

作为一款面向终端用户、复杂度受限的应用，我们致力于打造最佳的AI编程体验。这需要基于现有工具探索可能性——很多设想理论上很有价值，但具体实现路径往往充满不确定性。

But I think that we were and are in a space that's that's a bit bit different than that. You know, I we're this end user application that doesn't have much of a complexity budget. We are trying to, build the best way to code with AI. And so a lot of that is figuring out, you know, given the tools that you have today, what can you actually do? There's a lot of things that you could write down that that would be useful if you could build them, but then, you know, figuring out how to build them and all the details, it's not entirely clear how to move forward on that.

确实，2023年期间有很多这样的时刻。而且你知道吗，实际上还要补充一点，如果我们早期用户群体的需求走向完全决定了产品方向，那我们可能会被引向与现在不同的道路。我们有一批非常活跃的完全不懂编程的用户，当时我们讨论过是否该专注于他们。还有另一批强烈要求我们做技术栈专项功能的用户，希望我们只针对单一技术做垂直工具而非横向平台，这些我们都顶住了压力没做。2023年我们做了大量原型验证，像在沙漠中摸索前行，同时还要思考哪些领域不仅需要开发软件，还需要自建模型来改进或替代API模型——比如我们的标签页下一个编辑预测功能，以及具体如何实现这些。

And so, yeah, there were a lot of times over the course of 2023. And then, you know, actually, also to add to this, of our early user base, if you just kind of followed the gradient of exactly what they wanted, you would get pulled in slightly different directions than we ended up in. You know, we had a really loud segment of users that didn't know how to code at all, and we talked about, you know, should we focus on those folks? We had a really loud segment of users that wanted us to do things that were very, tech stack specific, you know, just building for one technology and making it much less of a horizontal tool, and we resisted doing that too. There was a lot of early prototyping and, kind of wandering the desert in 2023 and then, you know, figuring out things around, you know, where does it make sense to not just build the software, but also build our own models to improve the API models or or to replace them in places, like, you know, for instance, with our our tab, you know, our next edit prediction, and then how exactly to do that.

你们在2023年左右实现了从零到百万的突破对吧？这个过程肯定付出了巨大努力？

You went from zero to 1,000,000 around 2023. Right? And it was, it took it took a lot to get there. Right?

是的，实际数字比这还要高些，不过大体是这个量级。

Yeah. It was a a bit more than that, but sort of roughly that that ballpark.

然后2024年简直疯狂，你们一年内从1增长到1亿。聊聊这个指数级增长奇迹吧——你们如何保持每周10%的复合增长？这怎么做到的？

And then 2024 was a crazy year. You guys went from 1 to 100,000,000 in one year. Tell us about this, loss of, compounding power because you kept that growing 10% week over week. How did that happen?

早期数字看起来很小，但复利效应持续发力。我认为有几个关键因素推动增长：我们所在的市场有个特点——产品改进能立即反映在数据上。当我们首次让Cursor感知代码库时，当我们开始预测用户下一步操作时，当这个功能变得更精准、更快速、更强大（能预测连续修改序列）时，尤其当AI模型开始在你的代码库快速执行更多操作时——每次突破都带来明显增长。

So the numbers felt small early on, then the compounding kind of kept going. I think that there were a couple of things that really drove our growth. We're in this market where if you make the product better, you kind of see it in the numbers immediately where, you know, things start to grow more. And so we felt it around, you know, when we first started to make Cursor code base aware, when we first started to, you know, be able to predict your next action, when we made that then more accurate, then when we made that faster, then when we made that more ambitious, you know, it could predict sequences of changes. And then when we let the AI model start to take more action within your code base and then do that really fast, you know, speeding that up.

因此我们始终专注于产品优化，复利增长从未中断。并非所有市场都如此，但我们恰好处于终端用户偏好至关重要的领域。只要你做出最好的产品，人们就会关注并口口相传，这种效应持续了很长时间。

And so all along the way, you know, I we kind of just focused on making the product better. The compounding continued. And I don't think that this is true of all markets, but I think we're we're in a market where end user preferences matter a lot. And if you make the best thing, people care about it and talk about it, and that kept going for, you know, a long time.

有意思的是，同期YC孵化器里的公司也出现明显转向。我们常问他们用什么技术栈开发应用，前后两批公司的选择简直天壤之别。记得2023年用Cursor的可能只有个位数百分比，到2024年就暴涨到80%左右，像野火般蔓延。

I think one of the funny things that a lot of that that's happened around that time, we did see a big shift in the YC companies as they were going through the batch because we would ask what what kind of tech stack you use to build your applications. And it was night and day from one batch to the other. I remember in 2023, I think it was maybe single digit percentage of the batch we used cursor. Then 2024, it was, like, 80%. It's just, spread, like, by wildfires.

最优秀的开发者都在用你们的产品。

Like, the best builders were using you.

攻占了他们的推特时间线。没错。

Got onto their Twitter feed. Yeah.

所以推特是主要增长渠道吗？这些用户爆发式增长究竟从何而来？

You're on Twitter feed. Is that where a lot of, adoption? How how did all the growth came from?

在最初推出编辑器时，我们尝试通过社交媒体进行推广。实际上，我的一位联合创始人在2022年研究那些最终失败的想法时，通过在网络上持续发布AI相关内容获得多巴胺激励——他刻意不采用常规社交媒体策略，而是专注讨论AI。令人惊讶的是，仅通过研读论文、深度思考行业动态并公开分享，他就获得了领域内权威人士的关注。比如当时有个名为Flan t five的开源模型，多个AI项目正是通过他在Twitter的持续分享才了解到该模型的优势。

So the the very early stages when we were first launching the editor, we, tried to kind of evangelize it on on social networks. And, actually, my, one of my cofounders, kind of the the dopamine hit keeping him going in 2022 when we were working on some of these ill fated ideas, he started posting on the Internet and kind of explicitly set out to try to gain a lot of followers, not by doing kind of normal social media things, but by talking about AI, actually. It was kind of surprising, you know, the degree to which someone could actually just read kind of all the papers, think kind of deeply about what was going on at the time, talk about that publicly, and then get recognized by influential people in the space. And so there was, like, this particular, open source model, Flan t five at the time, that, multiple AI efforts that ended up using that model. They found out about, you know, kind of the benefits of that model directly from my cofounder just because he was posting on Twitter and, doing that kind of consistently.

他逐渐成为旧金山科技圈的小范围网红。早期他确实为产品带来了宣传效应，我们首轮用户候补名单的发布就像电影魔法般顺利，这对项目启动很有帮助。但之后我们便停止这种策略，2023年像苦行僧般专注于产品开发。

But he became, like, sort of niche, very niche, like, sort of niche, niche, niche of SF, micro celebrity. He would actually kind of evangelize the product early on. And so we had this kind of very movie magic demo, you know, when we when we first launched and when we first did a wait list to just get our initial batch of users. I think that that was helpful getting the you know, us kick started. But then after that, we kind of stepped away from that, and we kind of lived like monks 2023 and just focused on the product.

产品完全依靠口碑传播。我记得那年团队几次有人提议'产品已经够好了，该转向增长引擎了'，随后我们会进行两个月左右的增长冲刺。

And it it really just spread from word-of-mouth. I remember there were a couple of times during that year where there were members of the team that would say things like, guys, the product's already good enough. Like, let's put it aside. Let's just focus on growth engineering. And, then there would be, like, a two month sprint on, you know, doing some version of that.

但这些尝试最终都被当年其他更重要的工作冲淡了。

And it just never it kind of washed away compared to the to the other stuff that we worked on that year.

到2024年时Kershaw规模如何？当时公司有多大？

And by that time, in 2024, how big was Kershaw? How many how how big was the company at that point?

2023年规模很小——我的联合创始人们都是顶尖工程师，四人团队就能完成大量工作。我们在首批招聘上走过弯路，早期既过分谨慎又招聘不足，到2023年底仍是个位数团队。

It was pretty small in 2023 where my cofounders are are fantastic engineers. And so and there were four of us, and so we could go, pretty far without hiring anyone. We also had our own, you know, set of missteps in figuring out the first set of people to hire and how exactly to do that. And so we're both very patient early on and also, you know, focused on hiring a lot less than we probably should have early on. I think we ended 2023 at only single digits people.

对，当时还不足10人。

Like, we were less than 10 still. Oh. Yeah.

真了不起。现在换个话题——你对编程未来的发展怎么看？

Amazing. So Now I guess when cur curious, now shifting gears a little bit about what are your thoughts in terms of how the future's gonna look with, coding?

我们始终走中间路线：2022年筹备公司时，人们对我们专注AI投来异样眼光——那时ChatGPT尚未爆发，做AI被视为古怪选择。即便对AI有兴趣的人，也分两派：一派只想优化现有产品形态，另一派则认为'除了AGI其他都无意义'，觉得我们做的东西一两年后就会过时。

We were kind of this this maybe middle road bet from the start where when we set out to work on the company and we were kind of hiring our our first people, we would get these weird looks around, you know, why are you I mean, at the 2022, it wasn't really like this, right, because kind of Chatuche BT happened and then the whole world woke up to things in, you 2023. But especially during 2022, when we were working on the CAD stuff and then the early code stuff, people thought working on AI, was it was kind of weird to do. People were not entirely convinced that it was a good use of time and that there were gonna be lots of great applications to fall out of And then even the people who are interested in AI, there was, I in our space, you know, a bunch of people that were just focused on optimizing kind of the form factor that exists already, and just making those products a little bit better. And then at the same time, you know, in our social circles and professional circles, there's a bunch of people that, you know, were thinking, oh, why would you work on anything other than AGI?

我们始终认为：AI将是比近几个世纪任何技术革命更具颠覆性的变革，但需要数十年行业共同努力。在实现彻底改变软件开发或知识工作的终极状态前，需要逐个突破独立技术节点。就近期而言，对我们服务的专业工程师群体来说，代码仍至关重要。

And, you know, all of the work that you're doing right now in one or two years, you know, circa 2022 is gonna go away. And, yeah, I think that we've always had this view that there's going to be lots and lots of, incredibly valuable things to build over the next couple decades. AI is gonna be this transformative technology, maybe more so than, you know, any revolution in recent technological revolution in recent centuries. But it's gonna take a couple of decades, and it's gonna be this industry wide effort where there are all of these independent capabilities that each need to fall to really get to, you know, a place where you can entirely get to the end state of transforming building software on computers or kind of the other areas of knowledge work that might be transformed by AI. And, yeah, I think concretely kind of in the near term, we think that for professional engineers, which is the end user we serve, the market that we serve, you know, code is still really important.

而且，将会经历一段漫长而混乱的中间阶段，在此期间，你将与AI协同工作。渐渐地，它会越来越像一位同事。更进一步，它或许会演变成一种，你知道的，非常先进的编译器，能够为你隐藏部分代码。你需要阅读逻辑、审查并编辑它。所以，

And, there will be this long messy middle where, you will be working with the AI. More and more, it will become like a colleague. More and more, it may also become like a, you know, a very advanced compiler that can start to hide some of the code for you. You're gonna have to read the logic and, yeah, and review it and and edit it. So And

你认为哪些技能仍然重要？人们现在应该继续学习或停止学习什么？

What do you think are the skills that are still gonna matter? What should everyone still be studying or stop studying?

我认为编程就像数学一样，本质上是一种良好的通识教育。它不会消失，而且学习计算机科学现在还能带来许多实用技能。当人们进入充满活力的行业时，他们在学校学的具体知识并不特别关键，更重要的是他们在过程中获得的学习能力。我不认为AI会改变这一点。

I mean, I think that programming like math is kind of just a good general education. I don't think that that goes away, and I think that there's also lots of practical skills that comes from studying computer science right now. I mean, when people are kind of entering dynamic industries, the, like, specific stuff that they they study in school isn't super crucial. It's more the kind of learning that they get along the way. And I don't think that's changed with AI.

如果观众中有年轻的迈克尔·特鲁尔（比如三年前的你），你有什么建议？在他们开始Cursor之前，现在应该做些什么？

What advice do you have for the audience if you have, like, a young Michael Truell? Maybe not just three years ago. If they wanna be like you three years ago before they start Cursor, what should they be doing right now?

我认为只需要专注于你感兴趣的事情，与那些你喜欢相处且非常尊敬的人一起工作，并认真对待这件事。对许多在校学生来说，有太多事情让你倾向于打勾完成任务，而不是专注于长期构建某些东西，真正投入你感兴趣的事物。

I think just working on things that you're interested in and, doing it with people both that you enjoy being around, but that you respect a ton, and taking that really seriously. Yeah. I think that for a lot of people that are in school, it there's so many things that pulls you toward, more checking boxes and less, you know, focusing on building something up over time, and really focusing on on something that you're that you're interested in.

好的。让我们为迈克尔鼓掌。非常感谢。当然。

Alright. Let's give it a round of applause to Michael. Thank you so much. Yeah. Of course.

谢谢邀请我。

Thank you for having me.