How Cursor Built One of AI’s Fastest-Growing Companies

We don't need to compete with Anthropic and OpenAI on models right now. The interface between the human and the model is the key thing.

If you looked at the competitive landscape, it was almost silly.

I asked Michael, I was like, what do you think about Claude Code? And he said something to the extent, look, we are going after the biggest market in the world. You're always going to have formidable competitors. That does not scare us.

For us, the decision to invest in Cursor was actually pretty obvious. You remember Andrej Karpathy was using it. It was clearly a phenomenon. It was clearly like a known brand.

So to kick off, I actually wanted to take us back to early 2024. And you know, obviously you guys led the deal in, I think it was May of 2024 that year. But can you walk us through what was going on in the Q1, Q2 of that year and maybe how you guys started working with Cursor well before we ever partnered officially with them?

So, you have to sort of transport yourself back to late 2023, early 2024. I went back and looked it up. The leading models at the time were GPT-4o, Claude 3, and Llama 3.

Feels like forever ago.

Llama 3.

That was actually pretty good.

That was great. Yeah, it was a great model. You know, change in strategy obviously now.

The leading coding harness was Copilot by like a mile, right? This was like the consensus thing. Microsoft was doing their Microsoft thing, right? They had Copilot, they had Office, you know, they had VS Code. All of these things combined, they were going to somehow win AI. And so it was clear that AI coding was sort of working, but it was a different universe compared to today, right? Like you had the sort of early signs that you could do meaningful work with AI, but agents didn't really exist. You know, loops didn't really exist. Reasoning didn't really exist at the time. So, it's sort of a crazy time.

And not only models, there were a million companies doing plugins too. And both of these were very reasonable things to do, done by very smart people, right? Like that's what's so interesting about these new spaces. It's like smart entrepreneurs show up and do really interesting things. And in this case, model training was a very reasonable thing to do because the models weren't that good at coding yet. And there was sort of a theory that if we focus on this and do a coding-specific model, that's the breakthrough. People doing plugins on the other side were sort of saying, hey, let's not toss out the whole IDE, like that would be a crazy thing to do, but the interface between human and model is really important, so a plugin is a great place to go. You know, it just kind of turned out that the thing that Cursor was doing was the right thing to do, right? They, like us, they were very bitter lesson pilled. You know, at the time they would have said, oh, there's no need for us to go train our own model. Of course that changed later for different reasons which we're going to get into later in this podcast, but it's like we don't need to compete with Anthropic and OpenAI on models right now. The interface between the human and the model is the key thing. And this is something that I remember Michael and Armin saying like a lot, and saying just saying very clearly from the earliest days of Cursor, saying code in the future will look like pseudocode. Like I remember Michael making this point a lot, like when you look at old, you know, like computer science papers, there's a section, it's like here's the algorithm, not in code, but in like pseudocode, which, you know, I'm sure many people, you know, are used to. And you know, we don't exactly write in pseudocode now, but probably you could, like probably you could just take the pseudocode, you know, block from an old paper and like the models now would implement that. This is like really what the Cursor guys thought from the very beginning. And they've proven to be exactly right about this, right? It's like this sort of natural language-ish specification, like pairing the programmer's intent down to the minimum possible spec. That's what really sort of matters now. So anyway, so we were in this world, as Martin said, we, you know, we're talking to a ton of very smart, very interesting people working on coding in various ways. You know, the Cursor team kind of stood out because not only did they sort of align with the things that we thought about the world, more importantly, smarter people than us, you know, engineers working at the time at the leading AI companies like OpenAI, Midjourney, Replicate, a bunch of others. You know, all kind of use this and subscribe to this model. And the team frankly just seemed kind of special, right? Like they had an unusual degree of focus. I remember we got Michael to come pitch our GP group. And 90% of the meeting was him saying no to things, which is just like, for founders out there working on products that you love, like VCs will always ask you kind of dumb questions about other things you might do or things you're related to or whatever. And Michael literally just sat there and like very politely listened to all the questions and said, hm, interesting. No, like we're not going to do that.

Wait, wait, actually on that point, can I ask you guys a question? Because I remember that meeting so vividly when you brought Michael in for the A. And one of the things that he said no to was actually, and you know, to your point on like there were a lot of plugins, right? And so there was this maybe more consensus thought that a VS Code fork versus a plugin into VS Code, one would be better for enterprise and it wasn't the former, it would be the latter. And I recall, I think it was you Martin, or maybe it was both of you, asking him, well, are you going to get enterprise customers with this medium? Right. And he looked at all of us in the eye with high conviction and you know, well, I'm going to turn it back to you and like what did you think of his answer and what gave you guys the conviction to take a pretty, I would say, a pretty contrarian bet at the time.

They actually had a very reasoned articulation for why where they were in the design space makes the most sense. So for example, why wouldn't you build a coding-specific foundation model? So what they said at the time was it turns out that you don't speak to these models in code. You speak to these as humans in natural language. Therefore the model has to understand the breadth of human experience in language, right? And so that means in order to build a coding model you're actually building a frontier language model and that's just a very hard thing to do. And then of course there are other companies doing it and there are much more focused problems like tab-complete to do. So I think every one of their decisions was very reasoned in our opinion. They were also very product-focused, which you actually don't see a lot in AI. So they really believe that you can build a product, the product, you know, if it's really good will be adopted, which by the way is very similar to how Elon thinks. And that dictated a lot of their decisions. Like if you really believe in the product, you would never do a plugin because then you're part of somebody else's product. If you really believed in the product, you wouldn't do enterprise early, you'd do it later because the product itself is the kind of go-to-market motion. And so there was, I would say, and Matt let me know if you would agree, tremendous clarity on exactly what type of company they were and the decisions that they said all fell from that consistently, right? Where many of the companies we saw were kind of these pastiches of we'll try all of these different things, we're not quite sure what's going to work, you know, so their plugins and their services and their whatever, whatever, whatever. This was a product company going after what they considered to be a product problem.

I also remember by the way when we had the lunch after we did the investment and you know I was talking about something about like internet days that were similar. He was like, I was five. And we're like...

But like I mean what's amazing about not just Michael but all the founders is like even though they were pretty young when a lot of these things happened like they actually were kind of students of all this history which is a trait that we see, Alex Rampell talks about this a lot. We see this across all of our top founders that they're like able to study and really learn from what's kind of happened before.

I'm going to turn the question back on you. I think Matt would agree for us the decision to invest in Cursor at the time was actually pretty obvious. The team was phenomenal. It was incredibly focused. All of us were users. Like half the team actually used Cursor. You remember Andrej Karpathy was using it. It was clearly a phenomenon. It was clearly like a known brand. The growth was asymptotic. So from an early stage investor it's a pretty straightforward decision. It was a much less obvious decision for a growth investor. And so like I think probably what you heard in that first meeting maybe was a little bit dissonant with how you'd growth invest but then you very very quickly led the next round. So maybe talk through how you kind of like got past the typical growth evaluation and got conviction.

So you know you got the bug you know cuz you were calling us and we were so how did you convince David George? Like don't you have to show like you know...

But specifically, I agree with that, but specifically it was just kind of crazy to think you're going to fork something from Microsoft who owns the thing and then you're going to compete with them with smaller models. Even they own the OpenAI weights and they have enterprise sales force and they have... greatest enterprise distribution ever. They have 100 million developers. They own everything, they literally own every piece of that like...

You know, it's so interesting, right? Because like that's definitely the conventional wisdom and like it's usually correct, the conventional wisdom in many of these cases. One of the data points at the time was it was actually Marco on our team set up a dinner where Armin sat right next to John Schulman and they talked like a lot about Cursor. And I think John's thing was like, oh yeah, I tried Cursor but like our code base, I think he was at Anthropic at the time, like our monorepo just like immediately killed it, right? It's like I couldn't even load the code base. And like I think like the next day Armin came back, he's like, oh, we made some changes, like here you can. And not only is like speed of iteration, but like you know we just look for these little data points where it's like actually like doing a plugin or just like writing on top of VS Code is clearly not going to work because like the smartest people in the world have tried it and it doesn't work, right? So it's like so these guys were sort of listening to... And historically we as a team have always bet on the independent in a large market, right? If there's a leading independent we'll always bet even if the incumbent seems very scary. I do think it's worth talking about competition because my experience with, all of our experience with Cursor is that there was always like a looming existential threat and we almost forget about it. But like you know, of course, Copilot was the big one and then remember the whole Windsurf thing. So like Windsurf actually did this very smart thing where they went to YC and there was a lot of YC users using it. So you kind of had this kind of whole movement on Windsurf and you know Cognition was a phenomenal company actually was making real waves with agents and they've done a great job with that. And then you had the first version of Claude Code that came out and like that was kind of like a competitive thing. And so like you kind of go through like there was always some competition. And one thing I would say, and I would love again to actually hear your perspective on, is like it is such a pleasure to work with founders that are entirely unfazed. You know like normally like you know founders have this, things happen, are calling and they're kind of freaking out and whatever. They were 100% unfazed by the competition.

We invest in every Cursor round.

Sarah co-led the B and the C. That's real conviction. That was a harder conversation to convince...

The team for the C. You know, at the B you could argue, God, just bet on vertical momentum. That was actually obvious. Who grows from 4 to 50 in like four months or whatever. But this, the, or actually no, it was the D. So Claude Code, not to take us on a time travel trip again, but Claude Code came out May 2025 with a splash. You know, the pickup wasn't right away, but obviously formidable. And I remember there was a round, I think a couple months after, three months after that, and I asked Michael, I was like, what do you think about Claude Code? And he said something to the extent, actually, you know exactly what you're saying here, which is, look, we are going after the biggest market in the world. There's going to be competitors. In fact, there's been a rotating cast of characters from the very beginning. Copilot, Microsoft Copilot was the first one. Claude Code is one of a rotating set of characters. And you know, we think in big markets you're always going to have formidable competitors. That does not scare us. And you know, I just remember being somewhat blown away by the clarity of that comment, the lack of fear, right? There's humility as well, like it's humility mixed with bravado. And I just think that is kind of the winning combo when you're going after these juggernauts like that.

And I know it sounds vague but like things like this, it's almost a Bezos-ish thing which like if you're smart and right, it actually gives you I think the confidence and the focus to have that kind of attitude when you have this like mega competitor sort of coming after you.

Yeah.

That's true.

Yeah. Tab was the big thing when we, you know, Karpathy had that famous post. I guess he has a lot of famous posts where he's like tab tab tab. Yeah. Exactly. Like that kind of doesn't matter that much anymore.

Yeah. I actually think that's a good segue to maybe some of the other strategic decisions they made along the way around business model and to your point on oh [__] moments, I don't know if they would consider this that because I don't think the vagaries of Twitter impact them as much as the rest of us just like scrolling all day. But it was probably summer of '25 where there was a lot of [__] written about their gross margins, etc. And then of course, you know, post the Opus 4.5 maybe January period. I mean, we should all talk about that because those are both different but interesting times.

Yeah. I mean one thing else is I just feel like X is driven by like tail chasing, adrenaline, schadenfreude and there tends to be a massive disconnect between what people say on X and how the business is actually doing. And so even though there was a lot of this, the business tended to do very well. And what's so interesting is so much of this often comes from investors especially things talking about like business quality. But tech transformations always precede figuring out the business. Like if you guys remember the internet, like we didn't even figure out how to monetize it for a long time. Forget margins, making any money at all, right? And here you've got like unlimited demand, unlimited growth and margins. So, you know, I think this tends to be, you know, a, it's its own kind of echo chamber as people say it, but also b, every single wave you have investors trying to say something that sounds smart, which is like picking apart the business model of a new tech transformation and I just think historically they've always been proven wrong.