Greptile raises $50M Series B led by Menlo to automate code validation as agentic coding spend explodes

Speaker 2: Running two guests behind On public.

Speaker 1: We're we're

Speaker 2: Two guests behind.

Speaker 1: We're a little bit behind. But I like that we got some love in the chat for our next guest, Daksh Gupta from Greptile. Welcome to the show. Sorry for keeping you waiting. But your fans have been here. They were on time. We were we were running late, but welcome to the show. How are doing?

Speaker 5: I'm very good. Great to see you guys.

Speaker 2: Great to see Great to finally have you on. Yeah.

Speaker 1: You sent us a cool book of like the greatest bugs in software history. Is that right?

Speaker 5: Yes. We did.

Speaker 1: That was awesome.

Speaker 5: We are historians of the bug. We sent you that.

Speaker 1: Yeah. Very very cool drop. So many like, you know, you can only get so many t shirts and hats. So a book chronicling the history of your industry was a nice a nice touch. But clearly there's a lot going on. What's the latest and greatest? How is the company growing? How much money did you raise?

Speaker 2: I want to know it all.

Speaker 5: Yeah. Well, thanks for having me. We are trying to automate code validation

Speaker 1: Yeah.

Speaker 5: Which to us means that we want to make it so that you can review and test any code change and for sure know it is safe to merge. Sure. And particularly today, we're running a lot more code of course because the agents are getting better and and so we need this thing more than ever. Today, announced our series b. We raised $50,000,000 from a variety of investors. Menlo led the round.

Speaker 1: First garnish. Congratulations.

Speaker 2: Yeah.

Speaker 1: Mean, is the product just flying off the shelf because of this cyber security stuff going on?

Speaker 5: Yes. Cyber security is part of of it. It's also just... We're writing a lot more code and there there's kind of two choices. One is you like waste all of your time now like manually validating all the changes. And you lose all the productivity that you got from spending like, you know, like... I mean, like, we're spending more on coding agents than like the NBA. Like, this is like at scale sort of deployment of of spend and then and then the other end, your choices are either you like inject slot in your

Speaker 1: the NBA is kind of like one of many sports. Sports are just one of many things in the economy. Like, at the same time, like the AI boom is spending less than alcohol or so they compute build out

Speaker 2: CapEx.

Speaker 1: Less than less than all alcohol sales. But your point is that it's just growing so rapidly and it is a huge market and so Yeah. People will want the the the companion tools to enable that and make it safe and secure and efficient. Right?

Speaker 5: Yeah. Yeah. Yeah. You know, if you're spending as much, if you have engineers that are like, pay the engineer $20,000 a month, then spend $20,000 a month on tokens... Yeah.

Speaker 1: That makes sense. That... Yeah. Yeah. Yeah. That that that makes perfect sense. Got it. Jordy?

Speaker 2: And... Yeah. So talk about... Like, I'm sure this question came up a bunch and has probably perpetually come up in your fundraises like why why why an independent company or platform should be focused on this problem. Menlo, everyone knows, big and early investor in Anthropic. So like that is like a really powerful signal that like they've looked at the whole category and how the problem's gonna be solved and it's not just gonna be something that the coding agents themselves do, but like how how do you kind of wrestle or answer that question?

Speaker 5: Yeah. So I mean, know, last year when we started working on this problem, the most sort of common notion was that all of coding was gonna be the ModelLabs. And so there's just gonna be nothing else as we open the eye and drop a cloud And code is we took this somewhat contrarian bet that increasingly as companies started using multiple coding agents, that they'd want an independent and central code validator for all the code to go through. And we actually suspected that this is not something most companies would care about, but a select few that wanted their validator to be independent, that didn't trust a, let's say, coding agent company to also merge and approve and fairly review code that was written by the the the code generator. They wouldn't trust those companies to do both of those things, and we suspected that that would be some core of companies. I think we've really underestimated how much companies would care about this. I think virtually all enterprise... We heard this first from NVIDIA, which became our first large customer. They said, well, you know, we buy coding agents from, like, every good coding agent company that there is. You buy from the labs, you buy from every sort of application layer coding agent company, and they all have some version of a review product or testing product, but we wanted our validation to be separate. We wanted the separation of concerns between the thing that's writing the code and the thing that's merging and reviewing and approving the code.

Speaker 1: How far are we from, like the the the the fabled like lean, you know, lean implementation of all software?

Speaker 5: I think far and I actually think it's pretty... So the way we look at validating code, one thing is like, it literally correct and then there's one thing which is like, does it do the thing that people want it to do and is the quality of the code high, which is like a thing that starts to matter more when you're changing the rate of code. Because humans don't experience bugs in terms of like bugginess per unit of code change. They experience it over a unit of time. Sure. The code is buggy and it's like, okay, I'm I'm okay with experiencing one bug in, my meal delivery app once every three months. I'm okay with, like, once a catastrophic bug, but I'm not okay with it all the time, I'm not okay with it being slow. Yeah. And if you're changing the rate of change for the code to like be 10 or 20 or 30 times higher, the human tolerance for faulty software is not going to go up.

Speaker 1: Got it. Yeah. That makes a ton of sense. Well, thank you for coming on the show. Congratulations on the...

Speaker 2: Last last thing Please. Question. What... Is there is there anything that you can do to solve and help companies with like the economics of all the code that you're generating? Oh. Routing like what the agents that are contributing all the stuff? It's not routing as much as almost grading like...

Speaker 1: Oh, like Yeah.

Speaker 2: Are you being intelligent with your spend? Is that something you guys do already or or you would do?

Speaker 5: We're working on that today. I think we're very excited about, what you can do and you can measure and and putting it in spend. I think there's there's a point in time when everyone was like, well, let's just burn all the money we have on coding agents. And at some point, you can start asking, okay.

Speaker 3: Well, you

Speaker 5: know, we've doubled our our engineering spend and not seen a subsequent doubling of revenue. So there's something that has to reconcile

Speaker 1: Yeah. Here. This this engineer spent $40,000 on tokens and introduced $12,000.

Speaker 5: Yeah. And then it's like, oh, this is the actual... You you can assess yourself whether the economic impact of the thing they built was, like, sufficiently high or not. But I mean, the... I think, like, just the fact that we have a little observability is crazy. Like, we will fluidly increase, you know, token spend by $20,000. And when you think about the relative scrutiny of hiring another person for a $250,000 a year, like, we've put a lot more thought into whether we should hire another person for a $250,000 then

Speaker 1: Yeah.

Speaker 5: We just, like, let the the coding agent math, just rip it. I think that's, like, a very ephemeral thing. People are gonna start caring a lot more about this in the near term.

Speaker 1: Yeah. Yeah. I mean, even on the consumer side, people feel like, oh, yeah. Like, there's always... There's this big coding agent boom, but, like, the apps I use seem like the same amount of bugs or I haven't really seen that many new features on x y or z app. And so, yeah.

Speaker 5: That's actually French about like the apps that you use. Like, you use, you know, use YouTube every day and like you're telling me that these companies are spending, know, like all these apps where their engineers are are coding 10 times faster and then the app seems essentially unchanged. It feels worse. Yeah.

Speaker 2: Another way to think about it is like in a world where coding agents get good. Yep. And so there's one world where a company is like, well, we're just gonna just keep making our core product better and better. We're gonna eliminate bugs from the experience. Yep. But in reality, companies just decide we're gonna launch a bunch of new products. And so, like, software quality can almost, like, stay the same just because you have more capability, but it's being applied across a greater service area. So the war on bugs continues.

Speaker 5: I think that's true. And and also just... I mean, I've been generally, just surprised that people can say that their individual productivity is 10 times higher, but not even the breadth of the software seems to be growing at the rate at which you'd expect. I think there's something weird happening that we're spending, like, these enormous amounts of money on coding and yet the downstream effects have been so modest. Like, there just hasn't been that much of an increase in the quality nor the quantity of software at the same companies. I mean, yeah, there's many more startups and they're coding faster than they ever are coding. Our team is a lot smaller than it would have been five years ago. But we dig a large company and they just seem like they are purely from a downstream point of view. They just don't seem to be very different from many years ago, which is like very weird and confusing. And I think that like I think validation has like a big role to play in this, but also it's just I think to be able to genuinely answer the question, I think we need like significantly better observability. I can't imagine that they'll keep spending these large quantities and then just be okay with the fact that there isn't measurable downstream increase in productivity or quality or something.

Speaker 1: Yeah. Makes a ton of sense.

Speaker 2: Well... Amazing progress. Great to finally meet.

Speaker 1: I'm sure I'll

Speaker 2: be back

Speaker 1: on soon. Thank you for the buck.

Speaker 5: Thank you.

Speaker 1: I appreciate it.

Speaker 2: Thank you. Congrats to the team.

Speaker 1: Have a great rest of your day.

Speaker 2: Cheers to you.

Speaker 1: Goodbye. Let me tell you about the New York Stock Exchange. Wanna change the world? Raise capital at the New York Stock Exchange. Tony Hawk's been to the New York Stock Exchange.