Replit CEO Amjad Masad on 1 Billion Developers: A Better End State than AGI?

Amjad Masad set out more than a decade ago to pursue the dream of unleashing 1B software creators around the world. With millions of Replit users pre-ChatGPT, that vision was already becoming a reality. Turbocharged by LLMs, the vision of enabling anyone to code—from 12-year-olds in India to knowledge workers in the U. S.—seems less and less radical. In this episode, Amjad explains how an explosion in the developer population could change the economy, society and more. He also discusses his early days programming in Jordan, his unique management approach and what AI will mean for the global economy. Hosted by David Cahn and Sonya Huang, Sequoia Capital Mentioned in this episode: On the Naturalness of Software: 2012 paper on applying NLP to code Attention Is All You Need: Seminal 2017 paper on transformers I Am a Strange Loop: 2007 follow up to Douglas Hofstadter’s 1979 classic Gödel, Escher, Bach that explores how self-referential systems can describe minds On Lisp: Paul Graham’s 1993 book on the original programming language of AI

• 0:00Speaker 00

  You know, what is special about humans and what's replicable in in the machines and at least in the in the near term. My view is that AI is gonna get really good at sort of two things. Things that are highly represented in the data and things that you can construct a very good RL environment for. So what what can you construct a great RL environment for? Like obviously with AlphaZero games, right? Uh, games are famous for you can you can have these self-play sort of algorithms that develop over time.

• 0:40Speaker 00

  Uh, now with reasoning models, um, you know, math is an environment that especially with lean. Uh, it's like a code, uh, almost like an expression of of math that can be executed. Uh, that that's like a great, you know, RL environment. I think code execution as well, so running the code and and then, you know, um, doing reinforcement learning on that. And and things that are already represented on GitHub and and things like that. Uh, but but there's a lot of other domains where we actually still don't know what we're, how we're gonna make them better. Like, uh, new fundamentally new ideas, new knowledge. It's not entirely clear how we're gonna get there.

• 1:23Speaker 00

  You can't, can you use RL for these like more software things? Perhaps you you create a reward model, you can approximate these things. But I feel like the sort of the I ideas. And the creativity and the sense of like coming up with really novel things and understanding the world and very complicated, intractable way. And, you know, coming up with an idea that could fundamentally change how things work or change the world. I think will still have be the domain of the of the human.

• 2:11Speaker 01

  Hello and welcome to Training Data. I'm David Cahn and I'll be the guest host on today's episode interviewing Amjad Masad, the founder and CEO of Replit. Amjad's vision of the future is a world in which a billion people on the internet become developers. And in today's episode, we imagine how these billion developers will reshape the economy, society, culture, and more. A lot of people talk about AGI as a utopian vision of the future where people aren't working and there's universal basic income. But what if there's a different vision of the future? What if people are working, they're working as developers, and they radically change segments of the economy that we never thought we could revolutionize? Things like healthcare, education, industrials, and more. That's the topic of today's episode. Enjoy.

• 2:56Speaker 01

  Amjad, welcome. Thanks for coming on the podcast.

• 2:58Speaker 00

  Thank you. My pleasure.

• 3:00Speaker 01

  You and I have known each other, I think, five years now. And I had the chance to invest in you four years ago at Co2, so I've been able to see the journey. One thing that's been true for you since the first day I met you, and I think for maybe a decade even before I met you, is you've been talking about this idea of a billion developers coming on the internet, which is a bold idea. Maybe it's gotten more consensus as AI has come around. Maybe, tell us a little bit about that idea. When did that intuition first hit you? And how has that journey evolved?

• 3:27Speaker 00

  Before we get to that, I remember, I actually remember the first time we met, it was in our Bryant office, which is actually kind of a home, a loft. We're in the basement, sitting downstairs. And so, we worked slash lived in this like really small place in San Francisco. And in terms of the billion developers, it's just like, you know, ever since I was a kid, I started programming really early on. Like, my first experience with computers was when I was six years old. And like, by seven, I was trying to make things with it. The first program I made was for my younger brother to learn math. And I've done such a good job at it that he works at Replit today. So...

• 4:12Speaker 00

  It just always felt like making software is the natural thing to do at a computer, on a computer. And I was actually surprised that this is the domain of the expert as opposed to this thing that anyone can do.

• 4:25Speaker 00

  And you know, through thinking about why is that the case? It just felt like a lot of tools were complicated. Actually, they were getting more complicated over time. So if you think when, when I was a teenager kind of building a business, it was Visual Basic. And I can like, you know, make an app in Visual Basic with database and everything, kind of shrink wrap everything into an EXE and, and sell it.

• 4:48Speaker 00

  And then the web, web came along and it just felt like a lot more complicated. It was more powerful, you can deliver things over the internet. And then the complexity didn't stop, like if you think about a JavaScript application today. You know, there's, there's a lot of things you need to do, you, you need to spend perhaps hours. If you're new, you might spend days kind of setting up the development environment and learning all these esoteric things like, you know, what is webpack and what is transpilation, compilation and all that.

• 5:20Speaker 00

  And it just felt kind of worse than when I, when I started. And there was kind of no reason, I, I, I didn't feel like there was an intrinsic reason for that. There's all these perhaps social phenomena that made it so that programming is a lot more complex and one is this decentralized nature of open source. When open source took over and it was in Microsoft product managers kind of designing how programming should look like. But I felt like even then you can, you can have, you can have the open source ecosystem being this decentralized innovation machine. But you can, you can create

• 5:58Speaker 00

  experiences on top of that by mixing the best of the open source to create amazing experiences. In the old language of, of open source hackers, there's the cathedral and the, and the bazaar. And the idea of the bazaar is this complicated mess and this is where, you know, open source software lives. And the, and the cathedral is like something like Apple or, or Microsoft where, where you're designing it top down. But I was like, well, that's a kind of a false dichotomy. We can build cathedrals from bazaars. Right? And that's been kind of the driving motivation for Rappel. And I felt like, okay, if you make programming something that more people can do by removing this complexity, a lot more people would want to use it.

• 6:40Speaker 00

  And the nature of computing changes because no longer is there's this big divide between what it means to be a developer and what it means to be a user of applications, which was the original vision for computing.

• 6:56Speaker 00

  So this was the drive behind that. And it's just, you know, also the opportunities presented by being a programmer is kind of amazing. And I'm sure we'll get into my story, but like the fact that I was able to make all this money when I was a kid, was able to get an O one visa to get to the US. And I felt that opportunity could be a lot more accessible to people.

• 7:18Speaker 01

  Maybe take us, just to start, like, take us 10 years into the future, or I don't know. You tell me how many years from now it is. When there are a billion developers, what does the world look like? How does the economy look? I mean, I think you have this sort of imagination about all the ways that we're going to build software and the ways that businesses are going to be built are different. And to me, it is somewhat of a utopian vision of the future. And tell us a little bit about what that looks like.

• 7:40Speaker 00

  Yeah, I try not to be too utopian. But a few things on that. One is the nature of, if essentially anyone can program and most knowledge workers would want to develop or make make applications or solve problems using using software and AI. The nature of what it means to be a company kind of changes. Because if you think about companies today, we have these roles and we have these silos and kind of the way companies are structured are based on they sort of factory pipeline from the sort of industrial revolution. Actually, if you look at society today, a lot of it hasn't really been updated since since the industrial revolution.

• 8:23Speaker 00

  So the main innovation, the main collaboration slash sort of work innovation from that era is the pipeline. Is the idea, you know, you do this one thing and then you pass, you know, whatever you're making to the next person. Eventually there's a car or a toy, whatever, at the end of the at the end of factory chain. And you know, society is sort of designed, but that's like the main design principle that we have. And so you look at the school, for example, you start at like, you know, kindergarten, you go to first grade, second grade. And it's like, everything is like, kind of created like that. And even in companies, it's like, uh, oh, you have the, you know, the, you know, the product manager, you know, creating a PRD.

• 9:05Speaker 00

  And then goes to designer and then goes to uh, an engineer and then it goes to a release engineer. And then the product is out there, goes to marketers, goes to UX researchers. And then, uh, and so but always the silos and the, and this pipeline model. If you have journalists that can make, uh, that can make increasingly more complicated things and can use computing to its true core and to solve problems. I think, I think you're, uh, you're going to have people in the organization that can solve problem across the board. So if you're someone in sales and you want to, uh, you want to, you have an idea to drive more sales.

• 9:45Speaker 00

  You might spin up, uh, sort of like an SDR agent that does this one specific thing that's based on an idea that, that you have. There's no, there's no this separation, oh, well, I need to go to my boss to, uh, to go kind of, uh, hire someone to do the SDR. You can actually spin up an agent that does that exact thing that you want to do. Um, perhaps maybe you're on a customer call and a customer is asking, uh, how can I do this X, Y, and Z with your, with your product? And perhaps your API or SDK. Now traditionally you have to go back to engineering and you have to ask them how to do that and you have to kind of. But in this world, you can spin up something like Replit Agent and you can say, well, yeah, prototype this thing for me and show the customer in real time.

• 10:30Speaker 00

  By the way, all the stories that I'm saying are real things, uh, from our customers. And so the, uh, you could think about it as just a, a company is a set of, uh, generalist problem solvers.

• 10:41Speaker 02

  Hmm.

• 10:42Speaker 00

  And you see that in startups, right? We see that in startups, but I think that's going to be the case at scale. The other thing is, um. The way we construct software, I think, will change. So if you think about how we construct software, again, when you look within one company, you see this like sort of factory pipeline, but also when you look at the economy in general, you also see the same sort of separation between different companies and how the products get made. For example, the supply chain, right? In software, we have some sort of supply chain where you have like a you know, database company, and then you have a

• 11:22Speaker 00

  sort of infrastructure hosting company, and then you have a, you know, whatever, front-end company that's delivering actually goods and services.

• 11:31Speaker 00

  And I, and I think the going from the industrial age to the network era, I think that the way we're constructing will become more like a, like a network. And so,

• 11:43Speaker 00

  you know, you can imagine the way software is constructed is if we have something like crypto, Bitcoin, stable coins, things that are able to lets people to transact without having to know each other, trust each other. You can imagine software being constructed where I am sitting in front of software agents and I'm going to say, okay, well, I need to, to create this product. And the agent is going to be like, oh, well, I'm going to go grab this database from this area, this, you know, thing that sends SMS or email from this area. And by the way, they're going to cost this much. And as an agent, I actually have a wallet, I'm going to be able to pay for them. And when I'm going to publish my software, my software is monetized.

• 12:27Speaker 00

  And whenever there's like a dollar that comes in, it kind of flows it through the entire network. And so there's this ambient services that I'm, that my, my software agent is able to compose without necessarily having any sort of centralized system. And all these services are operated by hackers and people that are making money on the internet. And so, so I think, I think the nature of software itself and the nature of companies and perhaps the nature of the economy would change when everyone can be a generalist software and AI agent creator.

• 13:01Speaker 01

  Yeah, maybe taking that last point on the economy to its logical extreme, it's macroeconomics itself changes in a, in a world where everyone's a developer. You have all these stories, a lot of your users are abroad, they're in other countries, you have users in, how many countries do you have guys that users in now?

• 13:16Speaker 00

  Every country, China is, is, is, you know, is harder, but we have some, some users there, but basically every other country.

• 13:23Speaker 01

  So how do you think like the global economy changes, the way the economy works today, the people you're empowering, how does this change how these economies work?

• 13:30Speaker 00

  So I, I think one concrete thing, I think Peter Thiel talked about this paradox of the internet, where the internet was supposed to be the great equalizer, yet it centralized all the wealth in Silicon Valley in this, this one, one area. And so why is that case, like this, this thing that is purely virtual could have been and perhaps should have been much more decentralized. And there are all these things you can say about network effects and, and things like that, but you know, ultimately I think that part of the problem is, yes, we think about this technology as accessible, but it's not as accessible as we think it is.

• 14:15Speaker 00

  For example, there's a university student in India that was living in some rural area, but he was going to school to study computer science and that didn't have a computer or laptop, he had an Android phone at home. And he, he would go on Rappler and start programming and learning how to code on his phone, because we have a mobile app. And then he started picking up tasks from our bounties platform. So bounties is the ability for people on our platform to provide services to others. Thinking behind it is like, we have a lot of people with time and then we have a lot of people with money and no time.

• 15:00Speaker 00

  So a lot of people with time and money and it's like obvious trade. So a lot of people learned to go on a platform and they want to earn. So we want to present opportunities for them. And so we had an entrepreneur here in the US.

• 15:13Speaker 00

  Build actually a technical recruiter, building a recruiting app, but he was running into problems. And he was able to go on our bounty's platform, hire this kid. And for this kid to kind of fix some of those problems and for this entrepreneur to be able to ship his applications. So and that kid made money, made more money than his entire family would make for an entire year.

• 15:35Speaker 00

  And so, yes, the opportunities will get distributed. Obviously, this is more providing labor services, but you can imagine it actually creating a company. And creating something that could, that could be scalable. And that way, perhaps that this massive wealth generation machine that we call the internet will be accessible to more people in the world.

• 15:57Speaker 01

  Yeah, that's amazing. When I think about your vision, I think it's fundamentally this empowering economic vision. I think it's a vision of wealth. It's a vision of prosperity. It's a vision of giving opportunity to people like, and we'll talk about your story, but people like you when you were a kid.

• 16:11Speaker 00

  I'm curious, yeah, to question to you, like, what do you think is the macro? Cause you know, you're more than a finance brain. Like, what do you think the macro implications of everything that we just talked about?

• 16:20Speaker 01

  I think that if we go down this route of having a billion developers, the question is, what are all the sectors of the economy today? They're not techified, if you will, they're going to get techified. And so you think about healthcare, you think about education, you think about these enormous percentages of GDP that fundamentally have not been touched by Silicon Valley and touched by Silicon Valley technology. And the thing that I think about is, well, the barrier to entry was high for these industries. And so in order to build something for these industries, you had to go sell to them. It was difficult, it was tricky. And so what were the lowest barriers to entry? Consumer, you and I can go download Facebook on our phone and e-commerce. We want to buy things and it's pretty easy to make money buying things. And so you think about social and e-commerce to sort of the initial engines that kind of got going as the economy was technified.

• 17:03Speaker 01

  And this is where the, you think about the wage rate for a software engineer as being a proxy for how valuable software engineers are. And so the wage rate has continued to go up. It's sort of one of these surprising things. You would think that as the more supply of software engineers came into the market, the wage rate would go down. And what that tells you is the value that software engineers drive is actually higher than the wage rate. And so I think companies like Replit, and as you see this next wave of developers come, you're going to have this incredible prosperity from the value that they're going to go create. And so the thing that excites me is what are all the sectors of the economy where we can go create this value such that the kid in India who's using Replit is going to extract some of that value and create prosperity for his family, but also such that the consumer at

• 17:48Speaker 01

  the end of that experience, call it a healthcare experience, call it a government experience, whatever it might be, is also accruing value. And so we've seen that value creation, but to the extent that there's 100 million people, I think, who've created GitHub accounts today, and there's going to be a billion. And so you have this incredible increase. I think there's a lot more to come.

• 18:08Speaker 00

  There's also like a cultural impact, right? Like where Silicon Valley, as much as we have this global view of the world and we can, because we have so many immigrants, we can actually relay to a lot of those people. But there's limits to that. You know, I remember when I was working at Facebook, we were redesigning a photos experience on desktop and we designed this amazing kind of vertical scrolling experience where everything was scrolling after the iPhone came out. And everyone was excited about it. And then we went and did an A B test and all the metrics tanked. It's like, what is this? Did we create some crappy product that didn't work?

• 18:47Speaker 00

  And then the UX researchers did some tests and then nothing came back as not working. Everyone was happy about the product. Then one product manager actually looked and dug into the metric and found that most, like most, like, you know, the majority of Facebook desktop users use it on the netbooks or like those laptops where it's like wire screen and they don't have a lot of vertical space. And that was mind blowing to me. Where everyone in Silicon Valley have this amazing sort of MacBooks. And so there are limits to how much we can relate.

• 19:29Speaker 00

  And I think the products that we built is often not as suitable to these culture and creates this flattening of the world. Whereas when you have sort of innovation decentralized, I think they're going to be able to create applications that are more local, that can like benefit their communities. And I think there's less of that today.

• 19:49Speaker 01

  Yeah. You've architected your product this way. Maybe you could, I think this is like one of the secrets of Replit that people don't fully understand is how do you make a product, how do you make a coding product that's actually good on mobile, right? Is sort of people think of this landscape and there's a lot of hard work that goes into that. How do you make it work on all these devices, right? Maybe talk to us a little about how you've done that.

• 20:07Speaker 00

  Yeah. Well, initially, I mean, the main breakthrough of the open source project that we've become Replit, the company, was that I was the first to compile a bunch of programming languages to JavaScript to run in the browser. So this technology becomes Wasm later on, but you know, I was on the grand floor. It was like a research project by Mozilla. It was called Enscripton. And me and my friend were able to compile C Python to JavaScript using this technology. And we contributed a lot to it. We had to create sort of a Unix emulation layer in the browser. It was like very complicated project. But it captures people's imagination.

• 20:49Speaker 00

  We put it up on Hacker News and went super viral and people got really excited about it. I remember one highlight at the time was Brendan Icke, the inventor of JavaScript, like tweeting about it. It's like, wow, you know, we're like kids in Jordan, like building this thing and people got, people, you know, very important people got excited about it. And at the time, I thought that was like the best thing. The problem is when I tried to load them on phones, they would crash, especially low end like Android phones. And I even wanted to work on like these Nokia Symbian phones at the time. And it's because we were downloading tens of megabytes of JavaScript.

• 21:32Speaker 00

  So it worked for a lot of people. People are excited about it. But the problem with client side execution is that a lot of people's machines aren't just not very good. And so when we went to, when I went to work at Code Academy, and maybe I'm skipping a little bit in our story.

• 21:53Speaker 00

  We, again, we wanted more and more people in the world to learn how to code. And we started having users in Africa and other places like that. And their computers would not handle tens of megabytes of JavaScript to download. And then so I started building the sort of backend execution environment such that when you're, what you're using on your phone is really a thin client. And any code you're typing is just going to the server, executing and like coming back to the client. And then the, you know, make it so that you hyper optimize the JavaScript application such that it's very small. It does incremental loading. All that stuff with React right now is really easy, but back in 2011, it was actually quite hard to do. We had to invent a lot of kind of web technologies to kind of make it, make it work.

• 22:37Speaker 00

  And, you know, still to this day, like if you go to the App Store and download the Rapplet app, it's actually one of the smaller apps on the App Store. It's less than 100 megabytes, I think, whereas most apps is on the order of gigabytes.

• 22:52Speaker 02

  Why do you think that is? Like, the IDE market is one of the, I think, the last markets to move to the cloud. And you talked to a lot of Silicon Valley developers today. They still want, you know, the local thing on their laptop. You know, you cannot make me run this thing in the cloud. Whereas every other piece of software we've used, people have eventually given in. And, you know, Google Sheets and Figma and all these amazing cloud native companies. Why do you think developers have been kind of the last, last market to move to the cloud?

• 23:19Speaker 00

  What is the saying where like the shoemaker is walking around barefoot or something like that? There's a, at least in Arabic, there's the saying where, uh, you know, people who make things for other people are often not satisfying their own needs. I think there's some of that, and I think there's some cultural aspect like you, you know, you can't really, uh, you're, I can't really pinpoint a real sort of technical problem that can't be solved or can't be mitigated that prevents people from like coding in the cloud. I think a lot of it is cultural. There's this sense of control over the stuff on my machine. And you know, there's this old kind of joke in programming where progress in programming

• 24:01Speaker 00

  happens one generation at a time. Uh, so you need the whole generation to sort of retire so that for the new generation. By the way, all the folks that are learned to code on, on, on Rappel and Codecademy and things like that. I think, you know, they're coming into the market now. They have no problem with, with, you know, coding on, on cloud products.

• 24:19Speaker 02

  I love that. Another question. Back to the origin of the company. Did you first see that we were going to have L L Ms and, you know, Sonnet and all these amazing models to actually transform coding back when you started the company originally?

• 24:32Speaker 00

  So, um, when I was working on, uh, at Codecademy and then I went to work at Facebook, and before that on, on the open source Rapplet, what I was doing is wrangling code. So I was writing compilers, transmitters, parsers. And those things are very fiddly and very complicated, uh, programs. And it just felt like, you know, it just felt like this, this laborious thing that like machines should be better at. It was like, you know, parsing this character and like, you know, putting it in, in a node and trying to understand the structure of the program. And I read this paper in, in 2012, which I think sort of is an underrated, uh, paper

• 25:14Speaker 00

  that should be legendary, sort of on, on the order, maybe not on the order, but kind of similar to, to attention is all you need. It's called on the naturalness of software. And and this group, they they're making the argument that NLP can be applied to to code. And they make a statistical argument about how code kind of is statistically like natural language. There there isn't a lot of daylight between them. And so they build an Ingram model and they use the Ingram model to do completion. And by the way, you can scale Ingram Ingram models like Ingram models with the original language models.

• 25:54Speaker 00

  And famously Googled this like trillion parameter Ingram model that was like pretty good at translating. The very, very simple, simple things that's trying to, you know, predict the next character based on frequency.

• 26:08Speaker 02

  Yeah.

• 26:09Speaker 00

  So they constructed that thing and then it was performing well. It wasn't performing as well as IDE completions. But then they used static methods like ID completions to rank the completions from the from the language model. And the result was superior and program is rated as superior. And it actually did more completions than this static methods on their own or the classical kind of algorithmic methods. So at that point, it just felt like, okay, we're, you know, that's the future. Like, you know, you know, deep learning and things like that were coming on the scene and neural networks.

• 26:50Speaker 00

  And if this very simple thing is able to to to to to have such results, I'm sure we're gonna have better things. And so try to do something like that, but it didn't it didn't work really well at the time. But actually, when I pitched, you know, I I think the CDAG is out there on the internet. But one, one, one slide in it, I tried to do like the Elon Musk master plan, which everyone does these days. But basically, it's like, you know, the Elon Musk master plan was really amazing. It's like, oh, we're gonna build the roadster. It's gonna be expensive. We're gonna use that money to fund, you know, the next generation and that'll create economies of scale and then so on. And they did it.

• 27:33Speaker 00

  Basically, it was like, we're gonna create this website for hobbyist learners and teachers. And we're gonna grow it that way. And that way, we're gonna get a lot of data about how people use and learn programming. Then we're able to train machine learning models. And it's gonna be this like AI powered, you know. Uh tool and then that would be a tool that's like more powerful than traditional software, uh you know, tools and therefore would let people come to the side, learn how to code, make things and deploy them. Um and that was like in twenty fifteen.

• 28:09Speaker 00

  And so every year I would try to do something with machine learning and uh the first time that I got a glimpse of something wor- working was GPT two. Um and there's a lot of experiments at the time, uh you know, a lot of the early community around GPT kind of had the feeling that, okay, this is working because you saw GPT one, you saw GPT two, it scaled, even GPT two, there was like two sets of weights, one more parameters, you could tell that scaling is working. And so when GPT three came, it was amazing. It was like, you know, groundbreaking event. For a lot of us, it was the chat GPT moment. Uh because chat GPT was not, it was like a UI innovation, perhaps some fine tuning.

• 28:50Speaker 00

  Uh but it was, it was really clear that it was here and it was coming and I started orienting the company to take advantage of that.

• 28:58Speaker 01

  Do you want to talk to us a bit about Replit Agent? I mean, that feels like a transformative moment in this company and almost like the whole company, you got these millions of users, you had this community and almost like the whole company was built for this moment. Talk to us a bit about Replit Agent.

• 29:12Speaker 00

  Yeah, so um e- e- every year, so we implemented Copilot like features, we're actually the first startup to uh train code, code models outside of Microsoft and OpenAI. And we open source them for a small time. There were soda models, there were three billion parameters.

• 29:31Speaker 00

  And and that was really exciting, but the, the thing that I thought would be transformative is agents. Uh I actually had a, you know, thread that went viral at the time in twenty twenty two about how software agents will change programming and perhaps the economy. Um and so uh every, every year we tried it. I remember we tried it with GPT three. Um the context window wrangling that we had to do was insane, like we had to e- every time we generate like a class or a function or whatever, we'd have to like summarize it like into one sentence and put it back into context to be able to kind of iterate. Uh so that didn't work. Uh twenty twenty three, you had AutoGPT and things like that come out and you could tell like,

• 30:14Speaker 00

  oh, the models are getting better at like, you know. Like being coherent a little longer, but it wasn't working very well. October twenty twenty three, I I I gave a talk. Uh, at a TED Talk, uh, and I talked about how how in the future people make software. And in it, I talked about agents, I kind of predicted, uh, sort of test time scaling as well, where I was saying like, um, uh, you know, there's going to be a trade off between, you know, cost, speed and accuracy. And sometimes you want to like pour more compute or time into into a task. Um, and you're going to have these plans and the agent will iterate on these plans and and build a software.

• 30:57Speaker 00

  You're going to be a human in the loop where you're the sort of the creative engine. It was the plan, it was the roadmap for what we're going to build. And twenty four, I think around the time four oh came out, um, it felt like, uh, like we're almost there. And so I I essentially put the entire company on that. And in fact, we actually did a layoff, uh, in in May twenty twenty four, partly to to kind of cut burn and focus. Uh, but the other part is like, I knew in my bones that that was the way to go. And we were doing a lot of other random things that were unimportant. So between, uh, sort of March, the first time I got a demo from someone on the AI team.

• 31:39Speaker 00

  Uh, Zen on our team, he he showed me something and I I could immediately feel like we're we're there. It was like almost like a baby, uh, a baby kind of agent. Uh, so we spent, uh, seven, eight months working on it. And then, um, it just felt like we were on on this treadmill fixing one bug after the other, you know, trying to make it coherent. Uh, the transformative thing that happened is like June twenty four, when Sonnet three point five came out. And Sonnet three point five had two important properties. Actually, the main property that that people don't really talk about all that much is Sonnet was able to generate, uh,

• 32:24Speaker 00

  A like thousands of tokens, like I think up to thirty two thousand tokens of code, coherence code, like so you can it can one shot a repo almost. Whereas with G P T four oh, when you're doing agents, you're writing one function at a time and then testing. If you remember, had they a cognition demo, it's like, oh, they write one thing, they go to the browser, they write one thing, they go browser. And that was like, that was intractable, that was very, very expensive and we thought that's not not gonna work. And the thing that's on it, bro, at the time is like writing high quality code and not being lazy about it. Because the problem with OpenAI's products were lazy, there's a real problem laziness where it's a

• 33:07Speaker 00

  it it like you ask it to do something and and and at one point it adds a comment and says the rest of the code here.

• 33:15Speaker 00

  Bro, I asked you to make that program. Why are you asking me? It's like.

• 33:20Speaker 02

  It is like a human.

• 33:21Speaker 00

  Yeah.

• 33:24Speaker 00

  And so so Sana, it was this huge unlock and we felt like we're gonna be able to ship soon. Uh, you know, there's always this tension, you know, the quality and when do you own ship and so we set up, we set a deadline for my birthday, uh, September fifth and then and it was so contentious, actually, at least one person in the team kind of quit because they thought we're not ready and he was partially right.

• 33:49Speaker 00

  And we launched a agent and there was a lot of excitement because really it's the first software that works on the market. Uh, the very first and and people got really excited because uh, oh, you can try this thing and you can get a glimpse of the future or this thing can bring it in a database and like provision it for you. Can run a migration, it can like deploy my app. But then as it escaped the kind of early adopters, uh, it was, you know, a lot of people were disappointed because it was actually kind of crappy. And so we sprinted between September and then December, we had something we were really proud of and we exited the beta.

• 34:30Speaker 00

  And although we're growing well, then we grew really fast from there and people got a lot of value out of it. Now we have version two coming and it's rolled out to some beta users. And we've done A B tests and it is is anywhere, depending on the metric we're looking at, between 50 and 500% better. Wow. Uh, so it created 50% better daily retention, 50% conversion, 50% better engagement. But the metric that was really great was um, new users are twice more likely to deploy an app that they made with agents.

• 35:07Speaker 00

  And um, power users are, you know, five, six times more likely to deploy an app that they made. Uh, and deployment of Replit is very, very important because that's one of what makes Replit great. It's that you can go from an idea to deploy thing and we see people who deploy are 10 times more likely to retain on Replit. So that metric we track very well. So now I look at V1, I'm like this crappy thing, like we gotta, we gotta ship V2 as soon as possible. And we're like on the team and so we're starting to roll it out now and it's really, really good.

• 35:44Speaker 02

  Can you give us some of the intuition for how you've gotten it so much better? Cause I imagine it's still Sonnet 3.5, that still seems like this other model that everyone likes right now under the hood. Like how have you, maybe what's the behind the scenes for how the agent actually works? And how you've gotten it to be so much better when the base models are very similar?

• 36:00Speaker 00

  Yeah, so um, at the 3.5 sort of era, the um, we had to build this like a fairly complicated multi-agent system. We're talking about Langraph. Earlier where, you know, there was a state machine component to it. And there was a lot of non-AI

• 36:22Speaker 00

  state transitions that we're having to make so that the more you put into the eye, the the worst decisions it often makes and less coherent. So so context window was still very important. Because there are actually studies showing that although you know, these models are kind of advertising a million tokens, actually after something like, you know, thirty two thousand tokens, reasoning and and a lot of benchmarks just like tank like crazy, right?

• 36:54Speaker 00

  And so there's a lot of context window wrangling. And so we had a manager and it had different contacts than the editor than the debugger. Because we had to kind of make sure they they're isolated and have their different memories and and things like that.

• 37:11Speaker 00

  And you know, the tooling between all these things had to had to be different. We built our own sort of protocol for the agent to call tools. Because actually, tool calling was not very reliable at the time. It would like hallucinate tools or hallucinate arguments. So so there was a lot of engineering behind it to to make it work. When computer use from Anthropic came out, so that was three point five V two. The moment we laid our eyes on it, we felt that there's something important that wasn't getting marketed. That was that was start of a of a of a transition, which is they actually, it looks like they actually fine tuned it for long for long horizon reasoning.

• 37:56Speaker 00

  Because if you want something to do computer use, you're going to have all these images and contacts. And it's going to have to kind of continuously like click on things, reason, think about what to do next, click on the other thing. And so you can actually roll out a long chain of tool calls without.

• 38:16Speaker 00

  That big of a degradation in reasoning. And so we started re-architecting the the model of the agent to be,

• 38:29Speaker 00

  let's call it, less multi-agent and more single-threaded. Because the because the models are getting good at it. And that's like a much bigger simplification. It added other challenges, but it's a much bigger simplification over the existing model. And so kind of the lesson to learn is that you have to constantly rewrite the systems. Because you want to make use of the next version of the and you want to be able to predict what's coming down the line. When 3.7 came out, you know, Cursor had a big problem with integrating it. Everyone hated it on Cursor. I think they just recently fixed it. And the reason is because, again, 3.7 is more agentic.

• 39:11Speaker 00

  And so when you try to use it and this composer style requests response, it is not very good at that. It's actually worse at that because it's trying to kind of make decisions that but you're not giving it the space to to kind of do this looping. And and so and so you there's a really kind of tough lesson for engineering and product teams. If you over-optimize for the present capabilities, you're going to enter a local maxima that is quite hard to move out of. There's like an innovator's dilemma problem on the order of months. Yeah, innovator's dilemma used to be like on the order of decades, right?

• 39:51Speaker 00

  Where, you know, a company has an innovation, reaches a certain height of success based on a product that they made that's people really love. But then there's a disruptive technology and they don't make use of it because it might be destructive for their current business where it cannibalizes their business. And so they tend to not pay attention to it and fight it. The classic example is like Kodak had a digital camera product, but they didn't actually launch it because I was going to cannibalize their film business, which was the thing that's making you the money, actually. And this kind of thing is happening on the order of months, which is really hard to wrap your head around. And so the the move fast break thing is actually very, very important today, right?

• 40:36Speaker 00

  More than ever and you have to have, you have to be okay with your, with sometimes like delivering crappy experiences.

• 40:43Speaker 02

  You mentioned Cursor. What do you make of the entire land? It feels like the landscape of coding tools has just exploded in the last six months. And you know, developers are reaching for, for new IDEs, you know, many more people who didn't consider themselves developers before entering the market. What's your view of the market landscape and how, how do you think of as you know, the ideal person that chooses Replits versus one of these other tools?

• 41:07Speaker 00

  Yeah, so I think there are incremental innovations and there are sort of more disruptive clean slate innovations. And I would put Replit in the latter category. So you, you take VS Code and you build an amazing AI experience on top of it. Really Cursor is like a fantastic experience. A lot of people on our team use it.

• 41:33Speaker 00

  The, the thing is, it is by definition incremental. You took a piece of software that you know, Microsoft has been building for over a decade. And you added this, your much better experience on top of it, but you're still, you're still kind of, it is still literally this additional layer. With, with Replit Agent, we actually took a clean slate approach to that. And I was like, okay, what do we need to make it that people aren't coding at all?

• 42:03Speaker 00

  Like, you know, we, we want to have drive towards this vision of, of no coding, not only no coding, no DevOps, no IT. Like, we don't want you to set up a database. We don't want you to write migrations. Like, writing migration is the worst thing ever. Like, you know, if, people don't talk about this, but one of the worst things about building software is being able to keep your version control, your database schema and your, you know, sort of deployment settings and configuration in constant, in lockstep. And most, actually most outages, when you read the postmortems from Google Cloud or AWS, it's like a configuration error. And usually it's outdated configuration. The software changes, but they forget an environment variable configuration.

• 42:46Speaker 00

  And so when you have an agent to actually do these things, it's actually much better at doing that. And so the system that we built for the agent is this sort of transactional lockstep system that does these transformation one at a time. When you hit revert and replit, it actually reverts the code, but also reverts the database changes and other environmental things, environment variables, things like that. So there's all these things that we thought about where you, as a developer coming to replit, the thing you have to worry about the most is your ideas.

• 43:29Speaker 00

  You know, you don't have to worry about where do I get object storage and how do I configure my buckets? Right? Whereas when you're sitting in Cursor, there's either somebody in your company, a backend engineer, worrying about these things and you're basically hooking into these APIs. Or you're having to build them from scratch and like it would help you hit those APIs, but you still have to architect the larger system. So I think there's like a fundamental category difference between these things. These products are hooking into existing systems and require a lot of existing support around these systems. Whereas replit is trying to be the final tool you have to adopt in order to build a piece of software.

• 44:14Speaker 01

  One thing I've heard from users when I talk to them about replit agent and deploying on replit. One of the things that people love is that you can go straight to deployment and that replit does the whole thing. You start from nothing and you end up with a fully deployed system. Can you talk a little about that? How do you think the deployment space plays into this? You guys in some ways are not comparable to a lot of these other companies in the sense that you can go all the way on replit. How do you think about that last mile? Who cares about that and how does that play into the long-term vision?

• 44:41Speaker 00

  Yeah, so in the same way that I can say that Cursor helps you with the coding, doesn't help you with the other stuff. Versal helps you with hosting, doesn't help you with the other stuff, right? And again, It's like this back to the cathedral of the bizarre, sort of like a bizarre. You need to kind of find all these tools and configure them and figure out. Whereas sort of Replit, I mean, another sushi analogy, there's a la carte and there's omakase. Replit is omakase. We're gonna make, we think we have great taste and we're gonna make the design choices for you. And it's not for a lot of people. If you want to make all these choices, don't come to Replit. I mean, there's a lot of other products out there.

• 45:19Speaker 00

  So in terms of like the deployment system, the, if you want to reach a billion people eventually, like making software, a priori, most of them will not know how to set up a deployment environment. And therefore, the Replit product needs to have a deployment environment. Basically, that's how we approach product in general, you know?

• 45:44Speaker 01

  In the early days of Replit, there was always this narrative of it's students, it's young people, it's people learning how to code. It feels like in recent years, that's really changed and maybe AI has enabled that. I was talking to a Netflix engineer who's telling me about how he's using Replit. Paul Graham has obviously always been a big advocate of Replit and one of the early believers in the company. How do you think the user persona has changed and how will it continue to evolve over time?

• 46:07Speaker 00

  How amazing of a visionary is Paul? Like, how can he look at a crappy toy like Replit and just see that, oh, that this one day could be really big? Or the idea that people are living, you know, are sleeping on mattresses and other people's homes? That's gonna be like a hundred million dollar company.

• 46:25Speaker 01

  It's incredible.

• 46:26Speaker 00

  Yeah, it is fascinating.

• 46:29Speaker 00

  In turn, he's also incredibly good human and he's been very supportive during very difficult times in the company when others weren't. It was frustrating for a long time that like, oh, Replit is a toy, hobby toy, students, you know, kids, teenagers. On the one hand, I was excited by it because like we had a group of users that are willing to experiment, that are willing to kind of try to find a future of software, the future, what we just talked about of, you know, being able to have an economy built into the software infrastructure. And we tried a lot of that and I think it's gonna work. We still have, like, the main missing thing was the agent, but all these things are gonna work and.

• 47:11Speaker 00

  There's this co-evolution that happened with these users that was was very important. The other thing that that perception was important because sort of competitors weren't paying attention.

• 47:24Speaker 00

  They looked at Replit like, oh, that's a toy thing. Why would I pay attention to it? A very close friend of mine and someone who's looked up to in the in the industry visited us the other day and visited, you know, a lot of our friends.

• 47:40Speaker 00

  And he was telling him he was using Replit to kind of prototype ideas for a new company. And everyone was like, well, isn't that isn't that like the toy thing that I'm just working on? It was like, well, no, it's very useful for me. And I that's how I'm able to iterate on these ideas very quickly. So now, yeah, the perception is is changing. And partly, look, I mean, we we took explicit decisions to make the product more premium. Actually, what's fascinating now is the product, Replit is the most expensive product on the market. With Cursor Composer, I think one request is four cents. With Replit, one request is twenty five cents.

• 48:18Speaker 00

  And the reason we did that is like, we just really want to build an agent. We don't want to build code gen tool. And if you want to build an agent, it's going to be expensive. It's gonna iterate and and it's going to call a bunch of tools with every request. And so it did alienate some some users. And that partly also shifted the perception and and kind of more professionals got on board. But if we're gonna reach a billion people, we need to we need to go down market again. But I think you know, I think we started in one end and now we're starting another end. I will meet in the middle, I guess, at some point. But I guess that Replit is the sort of the more roadster now.

• 49:02Speaker 00

  And as we drive costs down by you know, perhaps using more open source models and and things like that, it's going to become more and more acceptable to people.

• 49:12Speaker 02

  What do you think of Vibes coding?

• 49:14Speaker 00

  I don't really like this term. I think you you shouldn't fight it because

• 49:19Speaker 00

  I actually didn't like Gen AI either. It's just, I don't know.

• 49:23Speaker 02

  Same.

• 49:24Speaker 00

  Yeah.

• 49:27Speaker 00

  I don't know, it just like cheapens the possibilities. Yeah. It's like, oh, this thing that generates things.

• 49:35Speaker 00

  Well, actually, you can build agents and not just generate things. It can actually reason.

• 49:41Speaker 00

  And vibe coding is, um, makes sense if you're sort of starting from a position of coder and you're Andre Karpathy and you don't want to kind of worry too much about the code and you keep hitting enter or whatever. But if you're, if you're starting with Replit, you're actually not starting from a position of code. You're starting from, um, an idea. You're starting from an idea that you're iterating on and, uh, and then you go in and the agent is unfolding this code in front of you. Actually, when you're using Replit agent, you don't have, uh, the luxury to, to, to look at the code. So, so, uh, you know, we have another product called Assistant and Assistant is for more advanced people.

• 50:24Speaker 00

  And that you can do more vibe coding there because it's like request, response, and you can like kind of review the code and, and do all that. But, um, yeah, if I were to explain Replit, it's just vibe, like don't code vibe. Not vibe coding, just vibe.

• 50:41Speaker 01

  You've always been a bit of a contrarian. One idea that's very popular now in Silicon Valley is AGI. There's this idea, there's going to be no software engineers and you sort of have the opposite view. There's going to be a billion software engineers. Talk to, how do you think about AGI? What is AGI in your view? And, uh, how do you, what do you, what would you say to all the people who say, I talked to a lot of software engineers and they're like, well, I'm not going to have a job. I'm worried about not having a job. I'm worried about, no one's going to have a job. We're all just going to be on universal basic income and all this stuff like. Replit is in some ways the opposite vision, right? Of like, we're actually all going to have jobs. They're just going to be very powerful and we're going to be able to do all these amazing things.

• 51:15Speaker 00

  I, I think it's like a fundamental, it's philosophical difference. Like what, you know, um, what is special about humans and what's replicable in, in the machines? And, uh, at least in the, in the near term. My view is that AI is going to get really good at, um, uh, sort of two things. Things that are highly represented in the data and things that you can construct a very good RL environment for. So what, what can you construct a great RL environment for? Like obviously with AlphaZero games, right? Games are famous for, you can.

• 51:56Speaker 00

  You can have these self-play sort of algorithms that develop over time. Now with reasoning models, you know, math is an environment that especially with lean. It's like a code, almost like an expression of math that can be executed. That that's like a great, you know, RL environment. I think code execution as well. So running the code and and then, you know, doing reinforcement learning on that. And and things that are already represented on GitHub and and things like that. But but there's a lot of other domains where we actually still don't know what we're. How are we going to make them better, like new fundamentally new ideas, new knowledge?

• 52:41Speaker 00

  It's not entirely clear how we're going to get there. You can't can you use RL for these like more software things? Perhaps you you create a reward model, you can approximate these things. But I feel like the sort of the ideas. And the creativity in the sense of like coming up with really novel things and understanding the world and very complicated, intractable way. And, you know, coming up with an idea that could fundamentally change how things work or change the world. I think will still have be the domain of the of the human. And AGI will have AGI, but it would be a functional AGI, meaning it would do the jobs that a lot of humans are doing today.

• 53:24Speaker 00

  By virtue of the training data being available and by virtue of some of these jobs, having like ground truth that you can you can train on. And the reason I call it functional AGI is because it's fundamentally not general in that you can throw it in a super novel environment and for it to efficiently learn things, especially when they're, you know, when there's not explicit feedback. And be able to to be successful in that environment, which which which you have the definition of the universal AI, which it doesn't feel like we're trending that way. But I do think that you can reach something.

• 54:08Speaker 00

  So the definition of AGI at a lot of these companies is doing economically useful activities in front of a computer. Right? I mean, they it's like a remote worker is what AGI? I feel like we're going to get there, but. If I have a remote worker, I'm gonna like create a hundred more workers. I'm gonna implement all my ideas. And and I still, it's a tool, it's useful for me. Is it gonna replace me? Well, if I am like a code monkey, it's gonna replace me. But if I see my place in the world as as someone who can generate ideas and create products and services because I understand what people want and how how the economy works and all of that. I think I think that's still irreplaceable.

• 54:50Speaker 01

  I want to talk to you a little bit about your life story. You touched on coming to the US, the O one visa, growing up in Jordan. Maybe start at the beginning for people who don't know your story because it's a really inspiring story. And in some ways, your whole life has built up to Rappel. It's not just a company. I mean, your wife works at the company, your cousin works at the like, this is like your whole life and your whole mission.

• 55:10Speaker 00

  All in.

• 55:11Speaker 01

  Bring us to the beginning, like, how did how did this all come together? How did you get to the US? How did you get into Silicon Valley?

• 55:17Speaker 00

  Yeah, yeah. So my, one of my first memories, um, just as a as a child, I don't know if you remember your first memory, but I remember very vividly that my father, uh, kind of getting this this machine and like opening this this box or like putting it together. And I was fascinated by it even before I knew what it does. And I I walk walk over him and I kind of look over his shoulder and sitting at the keyboard and sort of, uh, finger, uh, sort of typing. He had like a big manual, like, you know, you throw it at someone, it could really hurt them.

• 55:56Speaker 00

  And he would, he was reading into it, but one by one, and he was like finger typing, um, CD, MKDIR, you know, these DOS commands. And and I remember feeling that it's just, wow, this is like, uh, this is a machine that you can talk to. Like, you know, what is a DOS? It is like a Rappel. You know, this is where, uh, the Rappel name comes from, read eval, print loop. Like, you can, uh, have a essentially a conversation with this machine. So my father would like go and like attend, uh, classes to learn, uh, computer, computers. And he had like, you know, every night he would come in with all these notebooks. Uh, he's like such a such a big nerd.

• 56:43Speaker 01

  The apple doesn't fall far from the tree.

• 56:46Speaker 00

  I I'm different than my father. So my father is a Palestinian refugee. Grew up with like this intense focus on education. Like if we're gonna make it, we're gonna, we're gonna have to be like the best educated. We're gonna have to be better than anyone else. We're gonna, so he's like an A student and working really hard. My mom is sort of the, the opposite. My mom is like a sort of a free spirit. She, she was a into poetry. She taught me a lot of poetry when I was a kid and I was actually able to. It was like there's an art to telling Arabic poetry.

• 57:29Speaker 00

  And, and so I was this mix of two things because I can be very intuitive. I can be, I can go on purely in intuition and, and take a lot of risk based on some kind of idea or vision or something that I, that I feel, feel good about. And I also have this like very analytical mind. I can sit and be, be such a, you know, a pain in the butt about, about being so rigorous about certain things. And that's, that's really my father's inspiration. And, and so, you know, you know, one day my father comes home and I, I'm sitting in front of the computer and he was, he was mad.

• 58:07Speaker 00

  He put all his savings into this machine and, and I was like, you know, opened the computer apart, took the parts out, put them back in. And I was like, you know, don't get angry. I know exactly how to use it. I know exactly how it works. And I, I showed him how to use it. I showed him essentially how to create things, how to open applications that he's been studying all this time. And he was like, okay, this is fascinating. You're right, it's yours.

• 58:38Speaker 00

  Yeah, and the first program that I wrote, I think I mentioned it earlier, was to teach my younger brother math. And I thought, okay, I can use this this thing, this conversational machine. I still think about it that way. Obviously with AI, that really happened. Yeah, and this idea that you can sit in front of it, learn something, play games, do something fun.

• 59:01Speaker 00

  And so that was the first program that I built. And then, you know, when I was a teenager, I was really obsessed in Counter-Strike. So I would go to these internet and LAN gaming cafes. And I would like, you know, whatever money, scraps of money I have, I would like put it into that and just play, play a lot of Counter-Strike and strategy games and things like that. I got very good at them to the point that it was like a source of income. I was actually winning tournaments and things like that.

• 59:33Speaker 00

  Yeah, so I got into esports early on. So that's like another bench of my life. But one of those, one of the things I noticed about about those businesses is they were running on pen and paper. I'm like, you have all these computers, you can just write software for it. So I did write this like client service software that did accounting, that get people, you know, accounts and use username and password and manage their their time on the system and their.

• 1:00:03Speaker 00

  And I also did security so that people can't format the computers or do, you know, install malware and start selling that. And I made a lot of money on that. I sold it to a lot of businesses in Jordan.

• 1:00:17Speaker 00

  By the time I got I got to college, I had this idea that, you know, AI is gonna get so good, you know, we're not gonna have to write write software. It just felt like software is this like pedestrian thing that that that you do. It's like not that interesting. It's just like you just do it to like make things. And at the time, like these wizards of code generations were coming out from Microsoft. So when I went to school, I actually studied more on the electrical engineering side, especially because my father thought that computer science was not a real field because the Engineering Association of Jordan would not admit you into the, as an engineer, if you don't have like, you know, electrical engineering.

• 1:01:02Speaker 00

  And by the way, my father is the vice president of this organization.

• 1:01:04Speaker 01

  I didn't know that. I learned something new.

• 1:01:07Speaker 00

  He really loves that engineering organization. So. But but throughout my college experience, I got I got really into programming languages. I started reading Paul Graham, started reading Hacker News. Paul Graham wrote a lot on on Lisp. He actually has a book called On Lisp. And Paul Graham's view of programming languages is more of an art rather than than a science. Like, these artifacts are aesthetic artifacts and not just functional artifacts. And that also played into the reason to create Replin, because I wanted to try all these programming languages. And there wasn't a place on the internet to be able to try all these programming languages. And, you know, after I created Replin, had that breakthrough that I talked about earlier,

• 1:01:52Speaker 00

  it went viral in the US. A bunch of ad tech companies started adopting it. If you remember, 2010, 11, there was the MOOC kind of hype. You know, we have AI hype now. There's a like blip period where it's like the massive online courses, right? Udacity came out at that period. Coursera came out, Code Academy. And I got a bunch of, they all started using Replin, by the way, and I got a bunch of offers. I eventually decided to come to the US, got an O1 visa, because a lot of my work was published in the news and everywhere else. And so it was it was it was possible to go on O1 visa. Landed in New York early 2011.

• 1:02:33Speaker 00

  The only money that I had was taken from me at the airport. And the reason is, in the airport in Amman, I had like perhaps $700. Those are the money that I'm going to the US with. And. The folks at the airport did not know what a no one visa is. Apparently, like, no one in Jordan had ever gotten no one visa to get to the US. And they didn't think it was like a resident visa. They thought it was like a visitor visa and that I needed a ticket back. And I was like, no, it's like, I can go there, I can work, it's a work visa. And they didn't believe me. I was like, go look it up. They're not, they didn't believe me. And they made me buy a ticket back. And that was something like five hundred, six hundred dollars.

• 1:03:18Speaker 00

  So I arrived there with like about a hundred bucks.

• 1:03:22Speaker 00

  And my salary was eighty thousand dollars, seventy or eighty thousand dollars in New York City.

• 1:03:26Speaker 01

  That's pretty big.

• 1:03:27Speaker 00

  Yeah. Well, not in New York City. When your rent is like two thousand dollars.

• 1:03:32Speaker 01

  I think at some point you had someone offered to buy the company for a lot of money. How did you decide to turn that down? That seems like that was a pretty big decision. Coming from this background, you grew up, you made it to the US, you got this job. And then you have an offer to buy your company for life-changing amount of money.

• 1:03:46Speaker 00

  Insane amount of money. We were like six people at the time. And, you know, the numbers that were thrown around is between five hundred million to a billion dollars. We're six people. And so it would have made me insanely rich, right? So.

• 1:04:04Speaker 00

  And it was a tough time. I.

• 1:04:08Speaker 00

  I wasn't really happy about the culture then, like, we grew to six, seven people or something like that, from the three that are family, essentially. And we hired people that I didn't like very much and the culture was changing. And I kind of wanted to do a reset.

• 1:04:24Speaker 00

  And at the same time, my mom was diagnosed with cancer back in Jordan. And I had to go back, the entire family had to go back, which is half the team.

• 1:04:35Speaker 01

  Wow.

• 1:04:36Speaker 00

  To spend time with my mother. And it was a very stressful time.

• 1:04:42Speaker 00

  And the thing that made me, which is like the absolute rational thing to do is to stake the money, go home. And, and, you know, live, I guess, happily ever after or something. But I felt two things. One is.

• 1:05:05Speaker 00

  My dreams, right? I I've had the dream of being in Silicon Valley for so long. Like, the first time I knew about Silicon Valley is through a low-budget movie called The Pirates of Silicon Valley, where it's the dramatized fight between Steve Jobs and Bill Gates. And I was like, wow, this Silicon Valley place is like, there must be like flying cars and like really, you know, like mass advanced technology. Of course, you come here, it's like, it's like the suburbs.

• 1:05:36Speaker 00

  But but I always wanted to be here. I thought the innovation, I was really, I was reading about all these entrepreneurs. And I I I felt like this is the most important thing, and those people are heroes. And I felt like the weight of Silicon Valley also on my shoulders because

• 1:05:54Speaker 00

  Paul Graham, you know, Mark Andreessen and other people that I really respect invested in the company. And like, they all have like very high hopes for it, and and they're all really excited about it. And and I felt like I don't want to let people down. And I felt like if I sold the company, it wouldn't be, it wouldn't be, I wouldn't have achieved the potential of it. And and and maybe I would regret it in the future. And and it's like, okay, being rich is is good. You know, I think I think money is actually great and and improves your lives in many ways. It allows you to focus on on the things you love.

• 1:06:37Speaker 00

  But, you know, what are you gonna be? You're gonna be another, you know, rich Silicon Valley dude. And there's a lot of them, you know? Go like, right, invest, do angel investing, it's like your life has gone a little boring. And and you you never kind of want to take the pain again unless you're ill musk to kind of like go start a start another company and put your life's force and energy into it. And so for all these reasons, we decided to to turn it down. Now you have forty million users? Yes, and, and you know, we achieved evaluation over the, what we're gonna sell for. And I think the company is actually underpriced now.

• 1:07:15Speaker 01

  It's a lot of grit to get here from six people. I want to talk a bit about your management style. You have a unique management style. I remember seeing one time on the internet, you, you said like, I'm now the VP of Engineering of Replit. I feel like over the years, there's always been, you're a hands-on leader. Your leadership style maybe has become more popular, but it wasn't, it wasn't popular six, seven years ago and you were doing this. Like, how did you develop your leadership style? What advice do you have for founders as they think about running their companies?

• 1:07:40Speaker 00

  In some way, it's a deficiency, you know, Jensen now talks, talks a lot about him not being sort of a great manager and him having all these reports and only, you know, talking to everyone in these big meetings and, and being, you know, giving feedback publicly, not, not really doing performance reviews in the traditional way and, and all of that.

• 1:08:01Speaker 00

  And in some way, I would say it is a perhaps lack of skill in terms of like how to do traditional management. The management style that I have is similar to like how I would like lead a open source project or how I would lead a sports team back home. I've always been a leader and I've always been sort of a hands-on leader where it is this duality of micromanaging and trusting people. Like, it is actually not at odds to, to do these things. And the most inspiring leaders both can go dig into the details and give very precise direction,

• 1:08:46Speaker 00

  but then like really trust people to deliver on those things and also trust people to have their own innovation, their own ideas.

• 1:08:54Speaker 00

  The way I managed the company from the start is I had a, a text file inside Replit. I used it as a sort of a notebook and the text file had everyone's names. And the one thing that I think they should be working on or the one thing that I expect them to deliver on. And every week when I meet everyone, we would go around the table and I would tell them, did you do this thing? And it's either yes or no, or something don't go right or something like that. And then the other question, what are you gonna do next week? Uh. And so it's like, okay, they did that last week or they didn't do it. Why did they fail? What happened? And here's what you do, uh, what they're gonna do the next week. By the way, my execs still write that everyone in the company every week on Friday, they write,

• 1:09:39Speaker 00

  here's what I got done this week. Here's what I'm planning to do next week. And so, uh, and so I can still keep in my head what most of the company can do uh, or is working on, like I can walk around and tell you this, this guy's working on this. This, this person is working on this. Um, so partly is I can keep a lot of complexity in my head and I can like really be able to kind of um, make these, um, these, you know, very deep decision about how like a sort of one button should work uh, inside the product or, or how uh, certain marketing ideas we should, we should run.

• 1:10:20Speaker 00

  And so I can go between all these different departments and, and, and be able to go all in into the details and then kind of zoom, zoom back out. And also, you know, uh, just having really high expectations. If someone, if week over week, that person who said they were gonna get that thing done and they, they couldn't, it's like an obvious reason to let them go. It's like not that complicated. And so Replit has and then high attrition rate, especially in the first few months of people joining. Um, twenty, thirty percent of people are gonna either leave or get let go because they're not able to keep pace of the environment or they get confused about how to work in that kind of environment.

• 1:11:06Speaker 01

  The other thing I think that's unique about your culture. I remember I attended some all team dinner at an all you can eat barbecue restaurant. I don't know if you remember this, and there's all these young people around the table and some of them didn't even go to college, graduate college. I mean, you seem to recruit for raw talent.

• 1:11:21Speaker 00

  Yes.

• 1:11:21Speaker 01

  And that's something a lot of people talk about doing, they wanna do, but it's hard to do. How do you do that? How do you filter people? How do you find these people? Um, how do you give them leeway? How do you train them? Talk to us a little bit about how that, how that happens.

• 1:11:34Speaker 00

  Yeah. First of all, I am, uh, uh, you need to be able to work with weirdos and misfits.

• 1:11:42Speaker 00

  And I was able to do that, whether it's instinctively or whether I relate to them being sort of myself, a weirdo and misfit.

• 1:11:52Speaker 00

  And and I can see talents, certain people, like even back in in my college years, like finding those people that like have hidden talents and being able to harness it somehow. I had this feeling of like, if if someone has a talent that is not being put to good use, I feel like a sense of waste. And like, you know, this this needs to be harnessed somehow. And I think that's sort of a good management skill. But in terms of, so first of all, you shouldn't have allergy towards these people. I remember my wife and co-founder, Haya, when we hired Mason, he was like eighteen year old kid.

• 1:12:33Speaker 00

  He he was, you know, a joke that he's a runaway kid from from Santa Cruz, which is partly true. He left Santa Cruz in high school. He wasn't happy, he wasn't happy with his family, with the school. He went up to the city. He went to one of those bootcamps.

• 1:12:51Speaker 00

  Those bootcamps were using were using Replit at all of them. And one of them was sending me bug bug reports. I'm like, look, I don't have a lot of people who work on this. Why don't you send me your best engineers? It's like, look, I'm gonna send you this kid. He's really awkward, but he's one of our best. So he came in and I remember Haya one day was like, oh, this kid like doesn't doesn't open the door, doesn't keep the door open behind him. He like literally slams the door on me. He's like, well, he's not really thinking about you. He's like thinking about code as he's walking around, right? So and still to this day, I see people doing that. But the first thing is, most people, I think, just reject these people outright when they can't communicate with them or they can't relate to them.

• 1:13:35Speaker 00

  And then the other thing, like, you know, they're gonna spike on certain things and then they're gonna be not great at at some other things, right? And so, but, you know.

• 1:13:48Speaker 00

  You can construct a team where it all fits together. Uh, where the the peaks uh of of someone is the valleys of someone else, right? Uh, someone is really good at, you know, shipping thousands of lines of code. And someone else is really good at testing it and very methodical about code reviews. And you know, if if you have one who's like a cowboy slinging code and others like a little more careful. Uh, and like meticulous and rigorous and perhaps a little annoying of of how how meticulous they are. Put these two things together. There's a lot of tension, but then they they get you get a good, good product. So you want to balance the team uh in a way.

• 1:14:30Speaker 00

  And uh, you, you want to go to to hire from places that uh other people aren't going to. And you know, that's, that gives you an advantage because everyone's competing over the existing talents uh on the market. So for us, a lot of it was go go into Rappel, a lot of it was running hackathons or running uh sort of prizes on Rappel. And uh, and you know, some kids win and we fly them out to to SF to work with us. And you know, we've had anywhere from sixteen, seventeen, eighteen, twenty-one years old uh, you know, join the team. And even on the older side, um, uh, people who haven't had uh traditional jobs before. Or were were ND game hackers that that joined the team.

• 1:15:17Speaker 00

  So yeah, I mean, it's it's a sort of a diversity of sources. If I am starting a company today, I would try to like find um niche communities. Uh, this is where you'll find like some really underrated uh talent, whether it's a niche group community around. A niche crypto project or a niche um program language.

• 1:15:38Speaker 00

  That's where I would look first.

• 1:15:39Speaker 01

  You mentioned your wife, Haya, a couple of times. Uh, what's it been like building a company with your wife?

• 1:15:46Speaker 00

  Depends on the day.

• 1:15:49Speaker 00

  We um we've worked together, uh you know, we met at work uh back in Jordan. We were working at this uh company, we were recruited. Uh actually, a foreigner came to Jordan as part of a job from Belgium. And he decided that he loved Jordan a lot, he loved the desert. And he wanted to build a company there. So I was the first employee there and Hai was the second. She was a designer, I was an engineer. And he pitched us on this idea, we're gonna do consulting. And then we're gonna build a product on the on the side and that and then we're gonna become a product company. By the way, every company that has this idea never becomes a product company. I'm sure you've seen some of them.

• 1:16:28Speaker 00

  Um and so uh the company was very dysfunctional, he was non-present and so it was like the inmates were running asylum. Uh so I and I ended up like working a lot of projects together, just just for fun and um. And um and then we we started dating, which dating in Jordan is not real dating, you like go out for a cup of coffee. She had like a seven, seven P M. curfew and so and so um. And then by the time I started working on on the open source uh raplet. She actually contributed the um the logo and a few other designs, she helped with a few other things.

• 1:17:07Speaker 00

  And then when uh I got the um uh visa to come to the U S, um we we got married and were very young. I was like twenty four at the time. Um and then she she joined me in the U S. We continued working on projects uh the whole time. We've worked on art projects, worked on games, all these other things. Um and then when the time came to to start the company, I was kind of like looking for co-founders. Uh because I'm like, oh yeah, Y C says that you you need to have co-founders and I guess it's like best to have co-founders. And I was like going on these co-founder dates and trying to meet people and all of that. And then she was like, well, I can, you know, I can I can start a company with you.

• 1:17:50Speaker 00

  I was like, it's gonna be really painful, like it's really, really painful, like are you sure you wanna go through that? And she's like, yeah, yeah, of course, like I saw you, you know, with Code Academy and like, you know, I can do that. Uh three or four months later, she's like, I underestimated how hard this thing is. But uh she she obviously lived up to the to the challenge and I think overall it um. It strengthened our relationship because when I was at Cordera, I mean, I was, you know, working, you know, twelve hour, fourteen hour days. She she don't really relate, like, what kind of job requires that kind of commitment? And then with Replit, we were both doing that.

• 1:18:33Speaker 00

  So it's not like, you know, your your partner is going off in the morning, not returning until midnight. On the weekend, they're they're wasted and they they can't really do anything fun. And and so you don't really have this great relationship with them. If you have a startup founder and you're kind of, you have a regular job. But if you're both founders, you're actually going through this together. And I think I think that ends up being being good. And, you know, at the time, it was actually working against us because venture capitalists do not like, you know, husband, wife founders. There wasn't a lot of examples at the time. Obviously, the most popular famous example is Paul Graham and Jessica.

• 1:19:16Speaker 00

  But then now there's like they get Canva founders, a bunch of other companies that that made it work. But there are challenges too, like, you know, right now, we're fully sort of fully committed to the company, but also we have kids. And so how do you, how do you manage, how do you manage this? That that adds quite a bit of stress and and pressure on us. And then the other thing is, when something is going wrong in the company, like, a lot of people can go home and can disconnect because they, you know, find comfort in in talking to their spouse about other things. But when Hai and I go home, we're talking about the problems and the problems kind of percolate. And and and they end up seeming bigger and and worse.

• 1:19:59Speaker 00

  And so you you start to have, you start to need rules around when do you actually talk about work? And you're constantly breaking those rules, obviously, because you're thinking about something. So I I think it is a challenge, and I think it is, at the same time, there's there's something very.

• 1:20:16Speaker 00

  Very unique and good about it.

• 1:20:19Speaker 01

  Yeah, well, it's impressive how well you made it work and you guys have a really great dynamic.

• 1:20:23Speaker 00

  Thank you.

• 1:20:24Speaker 01

  All right, lightning round. We have some lightning round questions. Are scaling laws gonna hold?

• 1:20:29Speaker 00

  Yes, scaling different things. We're just gonna keep finding things to scale.

• 1:20:35Speaker 01

  What is the best piece of advice you've gotten?

• 1:20:39Speaker 00

  Paul Graham asked me this question. I was stressed at the time and he told me, is this your life's work? Like, are you gonna be working on this 10, 20 years from now? And I said, yes. And he said, well, why are you worrying too much about the daily tribulations? Like, if to just settle into the fact that even if things are not great today, you're gonna be able to fix them and you're gonna be able to turn things around. As long, obviously, as long as you're not dying. And that's why, you know, Paul always talks about not dying and surviving as a company.

• 1:21:14Speaker 01

  What is your favorite new AI app?

• 1:21:17Speaker 00

  I kind of love that because I make a lot of things myself and I use Replit to make them. I use the basics like, you know, ChatGPT, perplexity. But then I spin up, I spin up pieces of software every day to use. I guess Manus is an interesting demo.

• 1:21:37Speaker 01

  Yeah.

• 1:21:39Speaker 00

  It's actually pushing on this idea that we talked about of like, how long can models work while staying coherent? And I think they showed that they can go for an hour with some coherence.

• 1:21:51Speaker 01

  Since you brought up Replit apps, what's a cool Replit app that you've seen recently?

• 1:21:56Speaker 00

  There are a lot of business things that are fascinating.

• 1:22:02Speaker 00

  One cool thing that I, I went to New York, I was meeting a few investors and customers and I met with Sears Home Services. This is a century old company. I didn't know it still existed, but apparently the home services department still existed. I met this very cool, trendy team. They kind of easily work at a startup. They're working at Sears and they told me that six months ago, they finished the cobalt migration. And it took them six months. And then I started asking him like, what kind of other tools to use to use any SaaS software? And they don't have any ERP software, any of like these modern SaaS, SaaS software.

• 1:22:43Speaker 00

  And they leapfrogged an entire generation of, um, of software to, uh, start using Rappler to create agents to manage their business.

• 1:22:56Speaker 01

  That's really cool.

• 1:22:57Speaker 00

  Yeah, so they, they have these, uh, field workers that every morning they wake up and they're like, oh, how do I, you know, optimize my routes to kind of earn the most and service the most customers? So they built like an AI tool that actually gives them the most optimal route that they use every day. For example, yeah. By the way, that, that team is non-technical. They're all operations.

• 1:23:17Speaker 01

  Wow. It's insane that you go from Cobalt to Rappler and you skip everything in between. And it does, does, we were talking about this at the beginning of the conversation, but it gives you a sense of the sectors of the economy that are going to get transformed by this technology. Favorite book?

• 1:23:31Speaker 00

  I Am a Strange Loop.

• 1:23:32Speaker 01

  Hmm.

• 1:23:33Speaker 00

  Uh, Douglas Hofstadter.

• 1:23:34Speaker 01

  Yeah.

• 1:23:35Speaker 00

  Uh, I actually disagree with the conclusion slash, uh, premise perhaps of the book, but it's still like one of the best books exploring concepts like AGI, concepts like consciousness and the soul and what it means to be a human versus what it means to be a machine intelligence.

• 1:23:51Speaker 01

  Um, what is the foundation model that you're, you like most or most impressed by?

• 1:23:56Speaker 00

  Uh, that, you know, in, uh, Claude is the, uh, is the best. I mean, three point seven is, is really the best at doing agent stuff.

• 1:24:04Speaker 01

  Person who's influenced your life?

• 1:24:06Speaker 00

  Uh, I think, um, my mom had this, uh, supernatural belief in me, like she, uh, she would talk about all the great things that I was gonna do, even when I was a very small kid. And she, uh, she gave me this, again, uh, um, unwarranted at times confidence, uh, in myself.

• 1:24:29Speaker 01

  Recommended reading on AI?

• 1:24:33Speaker 00

  You know, Twitter, like, it's really awesome. You know, I think it got degraded recently, but you, but like, there's so many papers that you find on Twitter, so many snippets of information, just like, like, go read those papers, skim them, and I think you'll learn about what's coming down the line by just reading it a little, sure.

• 1:24:53Speaker 01

  Last question, who is the most underrated person in AI?

• 1:24:57Speaker 00

  I think, um, Michaela Gatasta, the head of AI now, president at Replit.

• 1:25:03Speaker 00

  Like, I don't think he's very public, but he was one of the early pioneers of LLMs for code, and he's a great leader as well, and visionary around where AI for code is headed.

• 1:25:21Speaker 01

  Super impressive guy. I love every conversation I have with him. All right, well, thank you for coming on the podcast. I really appreciate our friendship. Thank you for doing this.

• 1:25:29Speaker 00

  Thank you. I mean, I appreciate your belief in us, and I think you saw something special a real long time ago, and like, all these things that you notice about what's special about Replit, and I really appreciate that.

• 1:25:40Speaker 01

  Thank you. We're still 4% there, 40 million users. We got 960 million to go.

• 1:25:44Speaker 00

  Yeah.

• 1:25:45Speaker 01

  Exciting times ahead. Thanks, I'm done.

• 1:25:47Speaker 00

  Thank you.