Vibe Coding All Day? Stop Delivering PrototypesㅣGraphite, Merrill Lutsky

Intro

00:03that's really been exciting about building this company, especially over the past year as code generation has gotten so good. As agents get better and better at generating code, the the way that you're interacting with them probably isn't going to look like how we we write code today.

00:16it's v 0ero bolt new replet agent like something like that where you know where you're interacting with these agents and prompting them and and iterating on their experiences and it's almost more analogous to code review than it is to how we would write code previously and that's why we think like owning code review as as the starting point is what matters you know far more so than the fist fight that's happening over like in loop code generation right Hi, I'm Mel Lutzky, co-founder and CEO of Graphite.

00:50Graphite is the code review platform for the age of AI. We take best-in-class workflows and tooling from companies like Google and Meta and combine them with our agentic AI code review companion to help engineers uh create, review, and merge code changes faster than ever.

01:05We work closely with the team at Enthropic um to power our AI code review agent um called Diamond. The series B that we just announced was $52 million. Um, and Graphite has grown 20x in the past [Music]

What If Everyone Could Use Big Tech’s Internal Tools?

01:24year. Graphite started on a completely different idea. We actually we actually had sort of two one small pivot in the beginning in the first three months and then then the iOS tool. We'd hired in the process then a few engineers uh a few of Tomas's ex-coorkers from Meta and this we'd had this really peculiar thing that happened when when they onboarded.

01:46So they'd start on their first day we'd get them set up on GitHub. We'd show them how we do code review at the company and all of them had this kind of allergic reaction to it immediately and they all said you know this feels like so much slower than the tooling that I'd had at Meta.

02:00I I feel like a caveman using this. Google and Meta have both built amazing internal tools to help their their teams of thousands or tens of thousands of engineers um all working on on one monor repo at once even across across time zones and across you know spanning these massive code bases.

02:18And um if you go outside of those walls though u it's very hard to to get that same level of of productivity and uh have that same that same tool chain. And that was a lot of why, you know, why we started Graphite was we saw kind of the the gap between what we'd had at at the large companies that that all us and the founding team had worked on and what was available to every company when they're just getting started.

02:43And um that was really what inspired us to to build graphite. I think it's so important to

The Most Important Factor for Success

02:49like deeply understand and care about the user that you're building for. You're going to spend if your company is successful, you know, 10 plus years talking to these people every single day. And if you don't genuinely care about making their lives better, about solving their problems and about building uh product experience they really enjoy using and get a ton of value from, it's going to be really hard to just sustain your your level of enthusiasm about that.

03:12And one of the the things that's so fun about graphite and one of the best parts about getting to work on this is that it is something that we all on the team use every day. A lot of like great companies use every day. our friends at at these companies are are using and are constantly texting me with feedback.

03:28It's just fun to be able to to build something that that's for um a user that we're so you know that we're so deeply familiar with and um that we genuinely care about and and want to like make their lives better. So

How Is Graphite Different?

03:41in the beginning we started with tool for creating stacked pull requests. The idea of stack pull requests um is that instead you you break and having that all be one pull request um instead you break that up into a series of small dependent changes um that all sort of chain together and build upon each other.

04:00And the advantage of this is that you know first of all nobody likes to review like a thousand line pull request. that's like, you know, there's that meme that's like, you know, thousand lines like looks good to me. And I think that that really is is what you see play play out in the data.

04:11Like nobody wants to to like give a detailed review to a change that big. Um that's really why, you know, why stacking is is so powerful is it just it gives you so much time back um that you would have spent kind of sitting on your hands or trying to find something else to build.

04:28And we've seen like really meaningful improvements in in development speed at many of our top customers. ASA was the one where where we save 7 hours per developer per week. At RAMP, we increased their development speed by 3x. One of our adviserss kind of famously said like once you stack, you don't go back.

04:45And I think that that's really true in terms of the both like the user excitement and then also just the the retention numbers and and you know the engagement that we see with graphite pretty much every

Why Code Generation Isn’t Enough in the Gen AI Era

05:02day as every engineer has tools like you know cursor and windsurf and copilot and even some of these more agentic code generation tools now uh they're able to like write code faster than ever uh they can they can do or they're creating more code changes than than they ever have before.

05:18That's only really the first part of the story. I think even what we're seeing with with vibe coding right now is a lot of those products are kind of MVP like there. You still need to take that change and get it through code review. You still need to test it.

05:31You still need to deploy it and measure the impact. and that whole kind of outer loop process um where you're collaborating with your teammates that's like a huge part of of your life you know almost as much in many cases as the inner loop of of how I'm like writing code testing locally and getting it ready to uh to like push up to to a pull request.

05:52So I think that as you know as you have like this this idea of like vibe coding and AI code generation it's

What Should Engineers Do in the Gen AI Era?

06:00almost not enough to be just you know engineer anymore like the best engineers are also great product thinkers and great business thinkers great designers in some cases it's really it used to be the case I think that the limiting factor on creating anything was was you know asking like how do I actually build this and how do I implement it now you have these great tools that can help you answer that question and kind of get to an early version like a working prototype of of something like quite quickly.

06:28So then the question shifts from being like how do I build this to a little bit further upstream of like what should I build in the first place that's actually going to be of value? you know how do I actually interact with this and what is the optimal experience here and then how do I differentiate that and how do I in a world where like anybody can you know go on replet agent and prompt you with a few prompts make you make make like a working prototype of something how is that different from anybody else's like vibecoded version of this and I think that that software

07:02engineering looks a lot more like you know reviewing or even designing or or like you pming in the future versus like writing code as we know it today.

Building the Right Thing at the Right Time

07:17the transformation of software engineering uh with AI code generation is has just been the sort of seismic shift that's happened in you know as we've been building the company and you know we were fortunate in that we were working on solving a problem that was now made 10 times worse by AI code generation code review has always been and this outer loop of review and testing and deployments has always been a bottleneck at large companies it's always been an area that that needs better tools And that was true before AI, but now with with all these great AI code generation tools, that need is is

07:52all the more acute. And I think that's that's one of the things that's been I think we were lucky that we were building kind of the the right thing at the right time. U because we built uh we built tooling for accelerating the outer loop and we got that into a lot of great enterprise customers.

08:06We also built the trust with those organizations to be able to now deploy AI to help them review code faster. Uh I think it's really hard if you're just starting an AI code review company to convince an enterprise or convince a company like Shopify to give you access to their entire GitHub repo from day one.

08:23It's taken us a long time to build up that level of trust. And I I think there are a few factors there. One is just like building just adding value progressively over time, shipping features constantly, like showing the customers that we work with early on that we are consistently able to like deliver on our promises and you know continue to just add more and more great features to the graphite platform.

08:45Um we spend a lot of time making sure that that everything we do is like you know best practices for security. We're pentested very regularly. Like we really want to invest in in making sure that that that piece is is central to what we're doing you know.

08:59And then and then part of it is just climbing the ladder. I think you have to you have to start with like smaller companies first like you're always going to have kind of more you know smaller companies that are willing to adopt tools tools in the beginning and then like building up trust over time and I think now that snowball has really like started to roll and um we have you know some of some of the largest technology companies out there that are are trusting graphite every day and u you know we're we're really privileged to be able to serve