We leaned into that tech and launched agent in September. Since then, you mentioned the chart. It's just been like a hockey stick, so absolutely insane.

I'm joined today by Matt from Replit. We dig into Replit's speed run to a 100,000,000 revenue, their new product agents, and how that kind of led to this explosive growth, and what you actually do in marketing when you're growing like crazy.

I think the people who have had all these really great ideas and have these things in their life that they want to do better or that they want to create, well, now they can just create those things. Having something that can write code, but then having something that can stitch all of the pieces together and make it super easy to deploy and share with people, that's really resonated. Video is a very powerful medium for showing people, hey. I tried this thing and I failed, but I'm gonna try again. Because you could read, you know, tutorials or documentation or whatever and not understand, like, my thought processes.

That's a good step in the direction that we're trying to head of, like, educating not just people, like, on tech Twitter, right, but also people in their work, people that wanna build cool stuff on a Sunday evening, people who are just trying to take things to the next level, really empower those users to create.

Okay. Wait. Matt, we just realized that you you just told me that Replit is not so much a dev tool anymore.

Yeah. So, you know, what we were talking about, right, scaling dev tools, mostly about tools for developers. And Replit used to be exactly that. You know? It's a company that's been around for a while.

I think 2018, I was in Y Combinator. And traditionally, the tool's kind kind of been marketed developer to developers as a place to learn how to write code, how to write code quickly and deploy it fast. And that's really what I joined Replit for. But since the launch of Agent, since a lot of the kind of like AI innovation that we've been seeing, we actually see a lot of users who are semi technical or even non technical using the platform. So that introduces a whole number of challenges, but also exciting opportunities for how to market to these folks, how to help them learn to use the tool in the right way, and how to make them as successful as possible.

And if I'm gonna draw kind of a parallel here, right, it's kind of like Neon, the database company, where, you know, who needed Postgres databases a year ago? Devs building web apps with Postgres. But who needs it now? AI agents that are building web apps for people and putting databases on their back end. And so there are a whole host of people that are being exposed to data and all these developed traditionally developer concepts who have never heard of a database before.

And it's the same thing with Replit. Like, there are a whole bunch of users who they don't know what a full stack app is. They don't know what a web app is. So it's my job to kind of educate them alongside, right, the more hardcore devs who are now getting into vibe coding and building with the tools in the space.

Yeah. And, actually, do they need to know, I guess, is the question before that as well.

Yeah. Yeah. No. I mean, that's super valid. And I think this kinda goes back goes towards, like, what is marketing and what is important and what is developer education and what does it mean?

Like, do I need to jump into the difference between, like, a OLAP and a OLTP database? Like, nobody's really going to care. Right? But they will care if I tell them, hey. You want to build an app, you know, where you can vote on restaurants in real time and, you know, store them somewhere.

We'll actually need somewhere to persistently store that data. That's what a database is for. And agents are pretty good at writing, like, basic data architecture, and implementing it. And so we can kinda stop there at the outcome level. This is the outcome you want to achieve.

This is what you need to achieve it. And then really, if you're interested, if you're super curious, you can dig dive deeper and learn about all the nuances of this tech.

Yeah. That that makes sense. And I hadn't really thought about that. And okay. So, also, just kind of like stepping back and before we we're gonna get into this, but, maybe we can pull up the chart of your someone's posted your revenue.

I'm sure you can't endorse that, but, like or maybe you can. I don't can you endorse that or

not? It was a we publicly announced that number.

You publicly announced that. Okay. Okay. So there's this crazy chart that's just like the, it wasn't zero, but it looks like it was almost like zero for a very long time just because of how big the massive the chart now goes up to in the last year. And I think it's like, how how much revenue did you use post that you'd

we we announced a 100,000,000.

100,000,000 in a in, like, the space of a year, basically. It's, like, growing

I think a little less.

Less. Wow. Okay. So unbelievable growth. And so we I wanted to, like, kind of talk about, like, how like, well, why why has that happened to that?

I wish I had all the answers for you, Jack. But my my response is going to be a lot of hard work and, you know, the techno the new technologies that have came out and all of the foundation that was built by the team. Right? Repla has been around for much longer than I've been a member of the company. And, you know, folks like Amjad kinda starting the company, building it up into what it is today, and really positioning ourselves to take advantage of these leaps in technologies that came out.

So if you look at it, right, again, nine, ten months ago at this point, maybe a little bit longer. Time time moves too fast. Replit was a platform where you could build and deploy things super fast. You still had to write code. It was still based around writing code, but all the infrastructure was there.

And so then kind of when, you know, Sonnet started coming out, these models started coming out, Amjad had been thinking about this for a long time, kind of AI, writing code with AI before Sonnet. But these technologies enabled kind of a stepwise change in the ability to write code with AI. And so at that time, we were super well positioned to take advantage of all the new stuff coming out, basically what we call vibe coding today. And Amjad and some other leaders at the company made a really amazing decision to to lean into that. And so we leaned into those that tech and launched Agent in September.

And since then, you mentioned the chart, it's just been like a hockey stick. So absolutely insane.

Yeah. It's and it it it seems like a really interesting story because I I I saw Replit speak at AI engineers, like, two or three years ago, the first one. And you built this like amazing model that was like I think Amjad was announcing it at that conference. It was like this amazing model that was like pretty much as good as all the like foundation models at like cogeneration but was, like, really cheap so you could make it for free users. And it was, like, pretty incredible.

And, like, your I met some of your team members. There was, like, you know, like, guys that were, like, built the first versions of YouTube and were, like, incredible people. But but it sounds like that stuff still didn't like quite take off. It was the agents part that like really just drove this crazy growth.

Precisely. And so, yeah, today, you know, our our main product is an AI. It's a multi agent framework for writing code. Has access to all the same tools you would have access to in the workspace, which means that if you say, hey, I wanna build a full stack app that a CRUD app that has access to a database that can do all these things, that agent can provision a database for you, request secrets from the user, store them securely in the environment, build out the front end and the back end of that application, and then prompt you to deploy it in just one click. And so it's the combination of having the infrastructure, having something that can write code, but then having something that can stitch all of the pieces together and make it super easy to deploy and share with people that's really resonated.

Was it was it, like, immediately obvious that this was like was it just like wildfire? Like, how Well,

I mean, that's all relative. Right? Like, was it immediately obvious a year ago when we were building this thing and, you know, like, we were at the the base of that curve. Absolutely not. Right?

It was everyone coming into the office and just like putting in work to build this thing that was very much a hypothesis. Now after it launched, I think it became slightly more obvious in the sense that you get like an immediate kickback. Like you see the response to the product. So if you go back like, you know, in September and look at the launch, it was very well received, but we're still at, you know, like it is a slight blip on the graph, right, so to speak. And so at that point in time, we're all super excited.

We're super stoked. Again, all credit goes to the team for building this thing. I just get on podcasts and talk about it. But but, you know, everybody's really excited, and you just keep dumping energy and work and effort into the product. And it just feels crazy.

Right? It's like, okay, well now we're at 20,000,000 ARR. Like, that's insane. We didn't even this is great. We need to celebrate this.

We didn't expect to get there. And before we can even celebrate, the products just continues to take off. So very much, feels surreal, but also super fun and exciting.

Scaling DevTools is sponsored by WorkOS. If things start going well, some of your customers are gonna start asking for enterprise features. Things like audit trails, SSO, SCIM provisioning, role based access control. These things are hard to build, and you could get stuck spending all your time doing that instead of actually making a great dev tool. That's why WorkOS exists.

They help you with all of those enterprise features, and they're trusted by OpenAI, Vercel, and Perplexity. And for user management, the first million monthly active users are completely free. Let's hear from Utpal from digger.dev, a dev tool using WorkOS.

How it's designed is that you can start as early as day zero. But for us, it wasn't day zero. It was closer to when we first started monetizing because we didn't have a sign up at all. People could just anonymously use our tool. So it was a little later, coincided with when we wanted to start monetizing.

And, like, we needed a nice enterprise feature set. If you're open source and you're doing enterprise first, the minute you think about monetization is when you should think about WorkOS. To be honest, if we do that again, I think we'd think about that on day zero, to be honest, because, like, should have done it on day zero ideally. Anonymous usage should be permitted, but you should know who's using your tool. It should be optional, 100%.

It should be opt in, 100%. But it'd be great to have auth from day zero. You don't necessarily think about these enterprise features, but they still lead revenue, and it kinda is a no brainer in that sense. So, yeah, I highly recommend.

When you've got this kind of explosive kind of products that's, you know, taking off, what what things matter? Because I guess, like, the marketing job is very different under those circumstances Yeah. To, you know, regular, like, okay, we grew we grew a little bit less small. Like, we grew 10%, like, great or whatever. You know?

Yeah. So there are a whole bunch of considerations, specifically on like developer education and quote developer marketing. First, we made a huge pivot in what the product was and who it was who we're selling to. So our documentation doesn't necessarily speak to our ICP anymore, and our messaging has to change. And we have to reposition who we are as a company.

We're still working through that. Right? Maybe before we press record here, you're mentioning, oh, I've heard about Replit, but it was in the context of a developer tool. Well, it's actually really hard to change people's per per perceptive perception. Right?

And I still go to events and people are like, oh, Replit. Right? Like, I I used that in college. I learned how to code with this thing. Yep.

And so that takes a lot of work. And from my perspective, one of the most important things was getting our docs squared away. So put a ton of work into what our documentation looks like and how what language we use, the types of frameworks that, you know, we use to structure the documentation. So I'm really proud of that. And then along the way, it's been adjusting kind of marketing language, brand guidelines, all of these things to make sure that we're catering towards the right audience, acknowledging that that takes a really long time to change perception like that.

But, you know, I think we're getting to the point where people understand kind of what the product is about and who it's for.

Yeah. It's, that's interesting. I like, a couple years ago, I interviewed, Logan Kilpatrick when he was still at OpenAI. And I was like, kind of I asked him the same question pretty much. Like, what do you what do you do that matters when, like, this is such an, like, kind of viral kind of product?

And I think he said the same thing pretty much. It was like docs. This is like his trying to catch up on docs and make docs really good and make sure, like it it feels like that's I don't know if that's if that's just what happens when you just it explodes.

Yeah. No. And I mean, I feel I feel incredibly grateful because where we are today, like, writing docs three years ago, I woulda had to sit and lock myself in a cave and, like, write docs, like, nonstop. Now I can build the infrastructure and build the tools, and we can rely on a like, AI to get us maybe 80 or 85% of the way there. Right?

And so there's a lot of really intelligent ways you can approach doing that at scale with relatively less resources or being on a really fast moving team. Like, we ship so much stuff. I'm I'm like trying to catch up with writing documentation for the products that are going out. But, you know, a lot of advancements in AI and coincidentally, of the products that Logan and others are helping build today are enabling us to write documentation faster.

Yeah. Very cool. Okay. So okay. So docs is a big thing.

What what other stuff do you are you working on?

Yeah. So I do a lot of, like, technical content developer education, and and that's mostly through video. I think video is super powerful, we say as we're recording a video podcast. Especially for for educating folks, and especially in the age of AI where you're working with tools that are, like, nondeterministic. I think this is a big piece, especially for nondevs, But for devs as well.

Right? Like, it's a very new phenomena that you might run something twice and get a different result. Like, there's it's just not deterministic. Or you might try something out and it might break and you might feel like you're wasting your time. But then you try the same thing with some different context and you get this amazing thing that works in like an hour.

Right? Like, this is a very new phenomenon. And so I think video is a very powerful medium for showing people, hey, I tried this thing and I failed, but I'm gonna try again. Or this is exactly how I think through the problems. Because you could read, you know, tutorials or documentation or whatever and not understand like my thought processes for how I'm approaching getting an error in the console and debugging that or getting my app to look like the way I want it to look.

And so primarily, documentation and technical content education. I did a course with deep learning a few months back. I think it was a good step in the direction that we're trying to head of, like, educating not just people, like, on tech Twitter. Right? But also, you know, people in at their at their work, people that wanna build cool stuff on a Sunday evening, people who are just trying to take things to the next level and, you know, really empower those users to create.

Yeah. Yeah. Yeah. That's, and video, I guess, makes more sense if you're kind of expanding beyond just developers because developers like reading docs, but a lot of people just wanna see, like you know, everyone's everyone's brains have been, like, just so used to just watching, like, TikTok and YouTube and I don't know.

Yeah. No. I think docs are important, but they serve a bit of a different purpose. So if you think about there's some, like, really great documentation frameworks out there. I prefer Dia Taxis.

That's, like, my favorite. And and it's very opinionated about how you group documentations. There's sort of reference material. Right? Like, I want to look at this thing to understand implementation details or how it works or like specifics of of this feature.

And then there are, like, how to guides and tutorials, which are much more like, oh, show me how to do this thing. Personally, I actually like reading more than I like watching video because I like to move fast. And a lot of times, video isn't fast enough for me, and I I, like, I I want to be able to, like, skip around. Yeah. But there are a ton of people for whom video is the much preferred media because there's like this this sense of certainty.

Right? Like, I sit down and I watch a video tutorial. I know by the time I get to, you know, like, thirty minutes or however long the video is, I should be able to do the same thing this person's doing, and I will see every single step along the way of what they're doing. Or like I just put this thing on the background while I'm cooking or driving to work, and I kind of like absorb some information. I think that's powerful as well.

And personally, when I'm really stuck or there's something really confusing or I just don't get it from the docs, I know that I can rely on video for some of those concepts. And so I think that's really powerful. You know, the flip side to it is that it scales. So I can do a video that a 100,000 people see. I can't get on a Zoom call with a 100,000 people and help them Yeah.

Build apps. I wish I could. But those are aspects to video that I really love. You know, the other thing is that I feel like it's one of the most personal kind of media out there because every single aspect can be controlled. The lighting is can be changed.

The type of camera that you're using, the microphone, your background, all of these things are completely customizable. The way your screen shows up, the the phrasing you use. If you tell a joke or if you're like super excited or enthusiastic, maybe you can tell, I'm a pretty enthusiastic guy. I that's why I like doing videos. These are all things that that you can bring to the table if you're making video or creating content that like if you read my docs, like, I can try to be funny or tell a joke or whatever.

But like at the end of the day, it's just text. So

Yeah. Yeah. I feel like not that many jokes and docs these days.

Not enough.

Not enough. I

remember I I studied physics in in college, and there was this it's a couple, like, physics textbooks. Like, the authors would put this, like, crazy dry humor in them. And maybe it was because they're tenured and nobody could fire them or something. I don't know. But I love that stuff.

So that that's, you know, I take inspiration from that.

Yeah. Yeah. There's a lot of there's this like textbook in The UK that's like for the for like high school kids. It's like revision textbook. The most popular one is just like full of jokes actually.

Weird. It works though, I guess. So anyway, so yeah, that that makes that makes a lot of sense. One thing you said there that I actually missed was you said your favorite documentation framework, DiaTaxis.

DiaTaxis. So d I a d a t a x I s.

And what is that?

It's a way of thinking about doing documentation. So I think that's actually the tagline if you look up look it up. And so a lot of people a lot of people smarter than me have thought about this problem. And if you're if you're sitting there, you know, thinking about how do I write docs for my products? Or maybe you're writing docs yourself or you're a developer.

This is a great framework to think about how to divide up documentation, how to organize it in a way that's useful, but also discoverable and navigable for your users. And I think the separate separate thing about documentation, more of a personal point. Nobody I've met very few companies or people where where they're just like genuinely excited about writing docs. And I think it's like a really easy way to stand out or to like create something that people enjoy because everybody has to read docs. And so if you like say, it's kind of like taking out the trash or something.

Right? Like, I'm I'm gonna do this thing super well. I actually love docs. I think it's fun to create experiences like this for people. But I'm gonna do this thing super well, and I'm gonna have a super high attention to detail.

It's a really good way to be able to to kind of go out there and show people that that you care or to prove that, you know, like Mhmm. You can create high quality experiences for developers.

Yep. Yep. That's very cool. I had not heard of that and that's that's amazing. Okay.

So maybe you could tell us a bit about the agents products because I think a lot of people probably like, I I would imagine a lot of people listening is like, okay. So you press you type in a prompt and you get back, like, a result. But it seems like like the experience you've built is really, really nice. And one of the things I noticed is like, I hadn't seen anywhere else was that I type in the prompt. I say, I had a I had an idea yesterday that this is very embarrassing to me on the podcast, but, like, I snore, and it's it's quite bad.

Like, I want I want to, like, stop snoring. And so I I've used this app, and I was like, it's okay. I was like, I want I was thinking, like, maybe I could, like, vibe code, like, a better snoring app. Anyway, I I put it in and, like, into Replit, and you get back. Like, they draw out a, like, a actual, like I'm explaining this to this to people that say obviously about you.

But like, draws out like a mock up of it, which is not like coded up. And it seems like you have these kind of lots of like separation of like what like little agents doing different stuff and it's it's very cool. So maybe you could tell us a bit about how it works and

Yeah. For sure. I think, you know, from the start, we had this hypothesis that, you know, like, we should plan the app out for the user and kinda present that to them before they get started building. And then recently, we introduced kind of a visual preview, which is what you're describing. So you go to repl.

You type in a prompt. And the first thing we do is give you a plan of what the agent is going to build and then mock up a visual preview. And it's basically just streaming HTML into the viewport. So, like, in real time, you'll see the app being built. But that just serves as a basis for the design that's then going to be passed to the agent.

And you kind of can confirm all the individual features that are listed out of the app. So we'll give you a visual preview. We'll give you a list of features and then you can say, actually, you know, for my for my snoring app, I don't wanna, you know, like maybe track, you know, very specific snoring details. But I do wanna, you know, list like my the the the peak number of decibels or something. Right?

So you can adjust that and then start building.

That's a key part. Yeah.

Yeah. Yeah. Yeah. And so from there, you kind of get dropped into the Replit workspace and agent's gonna start scaffolding out that app and building it for you.

Yeah. And have you have you learned any things about like kind of the developer experience of agents? Because I know that's something that a lot of dev tools are starting to think about, like bringing little agent experiences.

Yeah. For sure. You know, again, the folks actually building this stuff much more talented than I at constructing multi agent architectures. Architectures. But But I, I, you you know, know, I kind of absorb this by talking to people, using the product, etcetera.

I think, you know, these multi agent frameworks for coding seem to be the future. And this is like what a lot of the kind of community is rallying around, like different SDKs and protocols for building for building agents. And then the actual developer experience, I kind of make it part of my job and personal hobbies to try out every tool that comes out. It's interesting to see where they're headed. And the experience of building with an agent, I think, is a lot like just being a software developer or maybe a manager.

Because all the time you're trying to think, like, what is the right context for me to give to this thing? Maybe you've seen on x context engineering, kind of prompt engineering, etcetera. What is the right context to provide? What context should I leave out in order to get the result that I want? These are always like things I'm thinking through.

And, you know, on the experience of building the agent, then you're you're trying to think, well, how do I get to what this person wants to build, what this person is interested in, and avoid all the pitfalls and speed bumps along the way?

Yeah. So basically, just if you're building a good experience for a developer, that's gonna be a good experience for an agent is like kind of, I guess,

the key. I think so. That's my that's my personal take. But, you know, if we're if we're thinking about the audience that I just spoke about, our users are maybe less technical. So it's like, how do you make that experience appeal to a broad range of of users?

And so a very tangible example is like our agent can read the console. Right? So it can see like back end errors. You know? It can read like logs.

Well, it can see front end and back end errors rather. So it can read like server logs and then the con the dev console. And so typically, that's something a developer would have to do. Oh, hey. Why is my app broken?

I need to go check check like what console dot log is printing out. Well, now our agent can just do that for you. Or as a developer, hey. I wanna go, you know, be able to search the Internet for the best framework and then implement it into my app. That's something we'll that will be have have launched by the time this podcast is out.

Our agent can perform web search on your behalf, finds interesting things, and then implement them into the app. And so it's really about shortening feedback loops, making for a better developer experience even if it's not framed that way. Right? Even if we're just, like, not showing this to the to the the less technical folks out there or trying to confuse them, that's kinda what we're doing.

Yeah. And actually, this brings me on to some of the I know we were talking about before was like, what kind of people are really, you know, doing well on Repla and, like, using it a ton and succeeding?

For sure. So I think there's a really broad spectrum from everyone from, like, hardcore devs still get value. Maybe it's like spinning something up really quickly or validating idea or building the scaffolding for something that they're then going to invest more time in. But I would say that, you know, some of our real power users, the people we see getting a ton of value, are these kind of semi technical builders with a lot of agency. And when I say agency, I just mean the willingness and ability to go and, like, debug things or figure out why things aren't working or, like, push the boundaries of what's possible.

And the the kind of parallel I like to draw is, like, someone who's really good in, like, Google Sheets or Excel. Right? Like, these folks, they're, like, really scrappy people who are willing to, like, sit in the trenches and figure out what formulas to use, which are, like, never documented anywhere, super hard to find. But they built I've built some really amazing things in, like, Google Sheets. You can build some really cool stuff.

And so those same types of people, if they're willing to learn a little bit about code or, are willing to fail or try again, you drop them in a tool like Agent. We're really democratizing the process of writing code, of creating. You know? The the code is kind of a means to an end in this instance, which is just to take a step back, is not to say we don't have any need for devs or software developers. Right?

I never wanna I I think that those careers are and will always be super important. But it is to say, like, there are a lot of things that devs shouldn't even really spend their time on. And so we're democratizing the ability to to create those things to really anybody. And so back to your question, who is getting the most value from Replit? I think the people who've who have had all these really great ideas and have these things in their life that bug them or that they want to do better or that they want to create, well, now they can just create those things.

Yeah. I I I think this is something that's really key, isn't it? It's like this agent agency you mentioned. And I I think also, like, some people just don't is it like like people who things occur to that you could try to do this stuff as well? I don't know if that's like, I think Yeah.

I think you're getting at it. I think you're getting at it. I've seen this pattern. I consider myself one of these people where I'm like, I go through the day. I'm always thinking, how do I do less of this boring stuff?

How do I do things better? I have this idea for something that I wanna create. Will it work? Can I sit down and try it out? I think in the past, like, that would be diagnosed in a doctor's office somewhere.

But as it turns out, it's also really great for, like, building stuff. But but I definitely would say that there there there are types of people, right, that that are trying to push the boundaries and that want to experiment and create things. And that's not to say that if you're not interested in that, something is wrong or that's bad. But it is to say that those types of people I see have better experiences in the product. Because they're really curious and they're really willing to explore and they're willing to try new things and potentially fail.

And these are all qualities, honestly, of developers. Right? Like, maybe you can think back to a time when you were learning a language. I know like when I learned Python, I learned it with like a textbook. It was like a terrible ex it's also a terrible experience.

And I'd spend, you know, back when I was a student, like, all this time trying to write, like, a hello world example or even just install Python, and it was the most frustrating thing ever.

Right? Because so Python is just the worst for getting started with. And, like, once you start writing, it's fine. But it's, like, just getting on your laptop is quite hard.

Exactly. But these are the things nobody knows about, like, being a developer. Like, you're just failing all the time. Even, like, I try to write I just you just fail all the time. And AI has made that easier.

But I think for non engineering types, the failure or sort of the doing really hard things isn't as normalized as it should be. And so a really big challenge in educating for these AI native tools is like, hey, look, you're gonna fail. It's gonna be hard, and it's gonna maybe kinda suck a little bit. But once you learn how to fail and you learn how to learn, it's really fun. And I think that very closely mirrors the process of being an engineer or developer.

Mhmm. Yeah. And and what does the developer, experience, like, people use it? Like, are we is it like are the best people kind of just like going in, like, a loop, like, just changing their prompts? Like, how are they, like you kind of sitting down with like the people that are best with tools like Replit, seeing how they're using it?

For sure. The very interesting thing is that like everybody uses AI like a little differently because it it is so personal. It's like talking to somebody. Right? And so I'll every time I see someone using Replit, I'm like, wow.

I didn't even think of some of this stuff. The I mean, the parallels I I like to draw are are that it's very similar to being a developer. And you're always trying to think about context, which context you can give to the agent. So, hey. I have this idea from an app.

What does that actually look like? Maybe I have to sit down and draw it out. Maybe I have to spend some time thinking about the architecture or all the little puzzle pieces that I have put together. This doesn't really have to be technical. Right?

It could be, I need somewhere to store data. I need somewhere to persist this thing. Then you ask AI, what's the best way to implement this? And turns out it's a database. Right?

But spending some time conceptualizing what you want to build, creating a plan, and then going into the agent and giving it context like a manager of this is exactly what I want to build. And then along the way, something's gonna break. Well, how do we fix that? We have to think about debugging. A lot of debugging is just first principles thinking.

Something is wrong. What are the possible things that could be causing this error? And then one by one, I'm gonna go through those things and check them and try and figure it out. And so these are all just principles of, potentially, product management a little bit, but mostly software engineering and problem solving. And for me, it's like, how do I convey this to people without being like, you need to know debugging and console logs and all this stuff.

No. No. No. You just really need to know logical thinking, first principles, you know, thinking, and how to break down problems. And that's really it.

Mhmm. Classic physics guy. Classic physics guy answer to

Just gonna gonna get on here and talk about first principles thinking. Everybody's like, alright. Turn this guy off. I've had enough.

Oh, that's very cool. Yeah. I that that but joking aside, that does sound a lot like, my high school physics teacher talking about, like, first principles thinking, like, just break it down into the steps and break it down into the what what you know, what you don't know, stuff like that.

Yeah. Well, you know, I think it's been a kind of a recurring theme in my life. And a lot of people like AI because it gets you something fast, and that's really there's like a dopamine hit there associated with getting those those outcomes. But creating software, creating applications is no different from any sort of problem or thing that you're trying to build, whether it's like you're doing work on your car, you have a home improvement project, or you need to write an essay. It's like, okay.

I need to sit down, and I need to think about what my goal is, and I need to think about the steps that I'm gonna take to get to that goal. And I'm gonna think about how I can make incremental progress to get there. And along the way, there are gonna be bumps and things aren't gonna work out. And it might be hard, but I have to you have to trust that if you sit with the problem for long enough and you you think about it and you're you're diligent, that you're going to get to a solution. And that's how I see building with AI.

I think someone should just clip that and then we could play it to ourselves every morning before we start working. I'm inspired.

That would be if I'm if I'm that good at motivational speeches, that'll be that'll mean a lot to me. So

Yeah. If you're listening, tell us how motivational matters.

On a scale of zero to 10.

Zero to a 100,000,000. There you go. Okay. So do you wanna talk about the general, like, where Replit is sitting within kind of, like, the AI space in general? Absolutely.

I frame yeah.

I don't

know if I phrase it was that like an absolutely terrible question. I was just like, yeah.

No. No. No. I think it's a it's a useful question because I get people that'll come up to me and they'll say, like, tell me how Replit is different from Versus Code. And I'm like Yeah.

Where do I start? But Okay. The state of AI is that most of us live in the Twitter sphere, and we just assume everybody knows what's going on. And the reality of AI is that there are plenty of really talented engineers and really talented people that have no idea what they can accomplish. I see it partly as my kind of mission to help people understand what they can do and help them kind of believe in themselves.

And so the state of where Replit is, the state of where these tools are, I think about it in three kind of categories. So you have like chat apps, you know, Claude, Grock, OpenAI, etcetera. You have, like, local kind of coding agents and apps. So these are, like, Cursor, Claude Code, Gemini, etcetera. These these are CLIs or maybe Versus Code Versus Code forks or some other IDE that you run on your computer locally.

And then you have these like cloud editors. Right? And so that's like Replit, some of our other competition, right, that run-in the browser. There's no installation. There's no configuration required.

And it has all the capabilities of an IDE of something that's on your machine, but it's a little bit abstracted. Right? Like, if you go to replica.com, separate from agent, which we talked about, you can also just create a new app, and it's like an it's like a container that runs in the cloud, and you can edit code, install languages, install all these things. That's actually super powerful on its own because it means you don't have to worry about, like I said, installing Python. You can, like, hot swap versions of Python, install Node, do all these things really fast.

And then on the top, we have a coding agent. And so these cloud editors, I think about the positioning of Repla is a cloud editor. So if you're super pedantic about, like, your dot file setup and, like, how everything works on your local machine, like, Repla is not the place to start a project. Right? But it is the place to start a project if you need something fast.

And of sort of the cloud development editors, of those editors, Replit's really the only solution that has databases built in. So you chat with agent. It just spins up the Postgres database, ties it into your application. You don't have to go to a different website. You don't have to sign up for a different account.

Authentication. Right? We can just with ReplitAuth, you say, add authentication, people can log into your application using Replit, etcetera. Safety features, you can scan for safety. Deployments, these are all kind of built in.

There are a whole host of features in this graphic. And so I think about Replit, the positioning among the market. It's the only all in one cloud development editor that you can build and create things in your browser without installing anything. And then because it's in like a little container, you just take that, deploy it in one click. And so really idea to app fast, that's the mission.

I think we get there better than the other competition. And then the only other thing I'll call it is that we also have like a mobile app, like a native mobile app. You install it from the iOS or Android store. You can do every single thing. This is actually crazy.

Every single thing that you could do on reblo.com on your phone, including write code. So big shout out there.

Yeah. Actually, I I heard that a lot of people actually, before the Asian stuff, I'd heard that a lot of people actually were like programming with Repla on their phone, especially yeah. I think in some countries, was like really big. There's like a lot of people doing this.

I mean, this is something back in the day that I played around with. I remember, like, because I I have an iPhone. And I was like, oh, how do I write Python on my iPhone? I don't know why I thought that would be cool at the time. Like, I'm a nerd.

You know? But you can actually if you just wanna, like, write Python scripts and then you can deploy them from your phone. You could be writing cron jobs and deploying them from your iPhone if you wanted to.

That's wild. That's so cool.

Use Agent. Right? And build full stack apps. You could do that too.

Yeah. And do do you think it's like it's kind of it kinda feels like sometimes people are like, everyone's ultimately building the same thing in the sense that I first saw that, like, meme on Twitter. Was like, we're all building the same thing, aren't we? How does it converge and how does it not converge with like Devin, you know, I mean, saying Devin, but like, you know, now like, Curse are doing much more like Devin like experiences and everyone's seeming to be like somewhat getting slightly closer. How do you see it like panning out with like the differences and like over time, you know?

Yeah. I mean, I wish I knew the answer to that question. You know, I think what you're saying has a lot of merit. And if you look at the industry, it is kind of like, oh, well, people are working on similar features. And even if Cursor is much more for hardcore devs, you know, some of those features parallel things that we've been doing and kind of vice versa.

And so, I mean, when I look at where agents and models have come since September in almost like ten months, it's insane. This time last year, I was writing code by hand. Now I write very little code by hand. Yeah. And to think, well, where will that be in a year, two years, five years?

I have no idea, man. I have no idea.

Well, okay. Well, it's the first truly honest answer we've had on it. So

That makes me happy. But, you know, if if we look at you get if you're really in the industry, you can see these things starting to converge. Yeah. But I I don't think anybody knows what'll happen when the next model comes out if there's a stepwise change in productivity. Like, what if we go from zero like, this step from zero to Sonnet three five?

We go from three five to like the next thing. Who knows? Right? And that's why most people that build build these tools that are thinking about building products in AI, you really just have to skate to where you think the puck's headed. There's a lot of risk there.

Right? Because if you go in the wrong direction, you wasted a lot of resources, but there's also a lot of reward if you go in the right direction.

Yeah. One one very, very extremely off paced question just to embarrass you, Matt. What's your personal record on the deadlift?

I knew you're gonna ask something like this, dude. I just so for those for those that are, you know, maybe don't follow me on Twitter or aren't plugged aren't terminally online. I post videos of myself lifting from time to time. So I think working sets, I'm usually working up to around 200 kilos on a deadlift. That's almost nine

pounds for those

I feel I believe it's like it was like 2.2. Right? So it's like four forty somewhere somewhere north of of four forty. And then, yeah, personal best maybe

of like five set five reps or something.

Couple reps usually. Think, you know, like, I'll work up to a heavy single and then I'll usually drop it down. So, yeah, 200 kilos usually like maybe three to five. And so then I haven't taken a PR attempt. PR is on a deadlift.

Once you get started getting really heavy, it kinda take a lot out of you. So I'm not entirely sure. I think squat, I hit two zero seven kilos on my last cycle. So that's four fifty five.

What? Okay. And then one have you got a tip for someone just getting into lifting? Tips.

Yeah. You know, I think people approach fitness. Gonna hit you with a physics take and you're gonna everybody's gonna hate me for it. People that are getting into fitness I for some reason, health and fitness has a lot of non, like, logical thinkers or just, like, random dogma stuff where people don't understand how things work. We keep humans understand exercise science extremely well.

Like, you can go read research papers. If you find stuff that's evidence based, it's it's actually very much a science. And people don't look at nutrition or fitness that way. And so if you approach these things, I approach them the exact same way that I approach building. It is very much a long term investment.

And people wonder, oh, Matt, how do you lift weight like that? Well, as it turns out, you know, if you do something for fifteen years and you do all the dumb stuff and you struggle through it and you find out what works and what works for your body and how to perform, you can get to some pretty amazing things. And so the same way people for some reason, like, if I look at a software developer or a CEO that has built something over the course of twenty or thirty years, people don't draw the connection between, like, that success and, like, working out or lifting or nutrition or whatever. But it's like, if you sit down and you understand the basics, it's actually a lot easier because, like, these are known known problems. You don't have to meander through life and and struggle.

You sit down and you figure out the basics. You figure out what's important and the evidence based approaches to doing these things, and then you just do them in a sustainable way consistently for decades. That's that's that's my advice.

But during the parallel with Repla, it would have been like doing all the right things, but then not really not really getting bigger until like one day you come into the gym and this Repla is Well guy in the gym.

Well, so that's interesting. I've I've thought about this. Like, the the cool thing about life and and business, right, is that there are asymmetrical rewards and and and nonlinear growth. And so, like, you know, there are very much limits to how much muscle or how much stronger you can get as a human being. Like, there are limits in our biology.

Those limits don't exist in business. And so, like, the classic I think this was, like, maybe Jeff Bezos or something. You know, like, in baseball, you can only hit a grand slam. You can only get four runs in the in one at bat at the most. In business, you can hit a you can hit a a $100,000,000 grand slam.

Right? Like, could hit a $100,000,000,000 grand slam. There there are no caps to the upside in one turn. So with fitness, it's very much like there are known linear progressions. With career or other goals that you're trying to achieve in life, it's much more like you could spend you could spend twenty years just goofing off and messing up and then do something amazing if you have the potential, but that's maybe not the right way to think about it.

Yeah. But that's that's brilliant, though. That's also, I'll add it to my matte mixtape of for the for the listeners. Okay. So, if you have one takeaway for DevTools listening, especially over the last since you launched the agents, what would it be?

I think everyone should reevaluate what is possible. And that is very broad intentionally. But if I was to give a tangible example, there are things that I couldn't automate a year ago, five months ago, three months ago that I can automate now. Primarily as a function of one, I didn't have the time to sit down and immerse myself in the topics well enough to understand them and then get to a solution, purely a function of time. And then tool, there weren't dev tools.

There weren't infrastructure tools that were good enough to accommodate those things. And so if you reevaluate what's possible and you reevaluate what you can learn and you give yourself like an hour or two to experiment stuff, sit down, say, I'm gonna work with Claude to see if I can come to a solution. I'm gonna work with Replit to see if I can come to a solution and deploy it. I think you're gonna start to find, like especially if you think about, like, you know, levels of talent and the ease of doing things, like, the difficulty is falling. So at some point, all of these things become possible.

And so I would say reevaluate what is possible and what you can automate and what you can build and then give it a shot. Because there's stuff that I'm building that I see other people building, I'm like, holy cow. How how is that how is that even how is that even feasible? And it's only gonna continue to increase. Right?

That's the exciting part.

Amazing. And do you have any shout outs for Replit where you want to direct all the DevTools people?

DevTools people. Yeah. You know, you can follow along on all of the Replit socials, YouTube, x, LinkedIn, I believe. And then I'm posting a ton on x, LinkedIn, and then videos very frequently on YouTube. So I'm sure Jack will list those out in the show notes.

You can find me there. I'm online a little too much. And then, you know, I'd I'd be remiss if I didn't say you gotta go to to replot.com, build an app. You gotta go to docs.replot.com and learn how to build an app. So

Amazing. And, and message us if you found Matt's inspirational words good or you're gonna go to the gym and lift more scientifically. Thanks, everyone, for listening. Thanks, Matt. It was really fun.

Amazing, Jack. Thank you.