Amjad Masad: Resolve, Persistence, and the Rise of Replit

0:00One thing that gave me an edge is I actually sit down and try to predict where the future is headed. The thing that helped me in my career was to say, okay, in the early 2000s, that JavaScript is really important. The web is important. And if the web is important and it is the primary application delivery platform, then JavaScript is going to be very important. But I just became an expert in that. And so I think you want to try to do that. And it's this constant dance. You're always kind of having more insight. That's the voice of Amjad Massad, founder and CEO of Replit. His mix of grit, foresight, and technical obsession turned a scrappy browser coding tool into a movement focused on empowering the next billion software creators.

0:45This is Mike Maples Jr. of Floodgate, and it's go time with Amjad Massad. This is Mike Maples Jr., and welcome to the Pattern Breakers podcast, where we explore why some founders radically change the future and how they stand apart. Together, we'll learn about the counterintuitive mindsets and actions behind their remarkable success. Brace yourself for a world where chaos is welcome, naysayers are often a positive signal, and movements galvanize misfits who transform the impossible to the inevitable.

1:30Amjad Massad's journey is defined by relentless persistence. In Jordan, he borrowed computers to teach himself coding. And as a student, he earned money building software for local internet cafes, often rushing to fix problems for customers without hesitation. even if it meant he had to skip out of the back door of class. This unwavering resolve stayed with him. Determined to launch Replit, he applied to Y Combinator again and again, finally gaining acceptance on his fourth try. His constant drive fueled continuous improvement, culminating in the breakthrough success of Replit's AI agent in 2024.

2:10Amjad's story proves that true innovation comes from never giving up. Let's talk to him.

2:21Amjad, welcome to the podcast. Thank you. Great to be here. Maybe we ought to start before the beginning. So can you take us back to early days in Jordan coding and internet cafes? Yeah. So sort of internet just arrived in Jordan. I was probably 12 years old. I would like walk to the internet cafe and like you go there you pay you know one gd for like two hours$1.50 for two hours and then you go sit down and they write somewhere in their book that you know this computer someone has two hours or something like that they keep looking at this notebook and when the time hits they go and tap you under the shoulder and say hey do you want to leave or do you want to pay more I was like you know that's kind of dumb like you have computers all over the place.

3:12Like, why don't you use them to manage this? And you would have less workers and the system will be more efficient. You can also have an accounting system that's automatic and tell you how much revenue you made. And so I started building that. I got something working when I was 14, 15 and went around and tried to sell it. Where it started hitting off or have product market fit is with land gaming cafes because a lot of their users were younger. They're more mischievous and they find ways around the system and they install malware and they hack or do bad things. My system had security built in.

3:52It had user accounts. Every computer would boot to my application and then that has a username and password and then you can go in and just boot up a few games or connect to the intranets. It was going very well. I was making a lot of money. I remember taking my entire class to McDonald's. McDonald's had just opened and it was the most desired thing and it was very expensive. So I remember taking everyone. I was just bawling. And then I started running into all this trouble. So I had Microsoft Access running on the server and all the clients were writing directly to Microsoft Access. There's no software in between those things.

4:33And so they would have these race conditions and they would corrupt the database. I had a pager. And so these companies, these stores would like page me and I had to run away from school because I was in high school at the time. I literally had to like jump out of the window sometimes to go fix these things. Eventually, like my software got pirated and like pirated software like it kind of took over the market. And I kind of also lost interest a little bit. I was getting into college and just becoming a teenager and interested in all sorts of other things. But yeah, that's sort of like my first business.

5:08And so how did you find your way to the States then from Jordan? Have you ever seen this movie? It's called Pirates of Silicon Valley. Uh-huh. Sure. So it's like this low budget movie about the dramatized story of Steve Jobs and Bill Gates. After watching it, I'm like, man, this Silicon Valley place, it seems awesome. You get to yell at each other, build computers and compete on software and build billion dollar companies and all of that. And so I was like, yeah, I want to be there. I want to eventually be there. I want to be one of those entrepreneurs. I stayed in Jordan and went to university there.

5:47But I remember reading Paul Graham, getting into Hacker News. And Paul Graham was writing about the power of programming languages and how startups could win by just picking the right programming language. And he had this essays and a book on Lisp. And there was a sense in which programming languages were not merely tools, they're actually aesthetic objects, like you can actually get a lot of pleasure out of playing with programming languages and understanding them. And I didn't have a laptop at the time. So I would go to school, I would go to the computer lab, I would install gigabytes of software to run a certain programming language.

6:26And then I would go to work, I would do the same thing. And then And I would go home and do the same thing. I'm like, why isn't there a website where I can go play around with Scheme and then go play around with Python? And I was like, okay, how hard could it be? Let me go build it. And then I wanted to do more languages. And then that took another two years. And sort of similar to my teenage years, I had persistence to kind of continue hammering at it. Eventually, I found a research project at Mozilla. It's called Emscripten. Emscripten would become Wasm. And the idea behind it is you can compile sort of C and other like native programs to JavaScript.

7:10Of course, there's a lot more complicated things and we ended up contributing a lot to this technology. And so we open sourced like an engine that could run Python, Ruby, Loa, a few other languages in the browser. And it went super viral. And then it really gained a lot of traction. A lot of companies in the US started using it. A lot of people were talking about it. It was presented at conferences as the future of what a browser could be because it could run all these different languages. I started getting job offers. And one of those is from Code Academy. YC company, 2011. I joined right after YC.

7:51And then they got me in 01 and came to the US. So you were working on Code Academy. And how did you make the transition to some of the key ideas that framed the opportunity for Replit? So the theory I had that even motivated early Replit was more people would want to code. More people would want to use software. The tools were getting better. We could build a browser-based environment. If you did that, more people would want to learn it. More people would want to do it. But it's hard. Programming is hard. and if you have a job, kids, who has time for this? And then I left Code Academy and I went to work at Facebook.

8:32I was really excited about React, what was happening in the web. I ended up building, being at the ground level of all the modern web frameworks that everyone uses today. I was specifically a founding engineer on the React Native project. The goal was to create this cross-platform, or ostensibly this cross-platform, development, you know, mobile development framework. The thing that I was like more interested in is how do you make mobile development accessible to more people? Because like if you want to write a mobile app, you have to go learn Java or Objective-C. And then you have to like pay Apple like a hundred bucks to like make anything on the phone.

9:13And with React Native took the same approach that we're taking to Replit, which is like get started in five minutes. And so built like a command line interface that walks you through building a Hello World app. And you literally in five minutes have an emulator app running Hello World. And then you can iterate live. You can write web code and see it refresh quickly. And I think we increased, we created a new market for developers. Those who only knew web technologies are able to now build mobile applications. So again, it's sort of this continuation of this idea of like, you can actually expand the potential for developers.

9:54And developers are such an important thing for the economy. I mean, the reason why, you know, startups have been able to create so much wealth is because technology is moving fast, but like being able to apply it to parts of the economy is hard. And that's the act of startup. It's like finding problems in the economy that could be solved by technology. And was there any sort of flash of insight that led to Replit? Or did you just kind of find your way there gradually? Yeah, it's this dance between the potential of technology and where it's headed. I feel like one thing that gave me an edge is I actually sit down and try to predict where the future is headed.

10:39The thing that helped me in my career was to say, okay, in the early 2000s, that JavaScript is really important. Like the web is important. And if the web is important and it is the primary application delivered platform, then JavaScript is going to be very important. It was a crappy thing. It was like a buggy thing, still is, but I just became an expert in that. And so having a little bit of foresight and trying to predict where the future of technology and same with AI, we got into it so early, 2019, 20 with LLMs. And so I think you want to try to do that. And it's this constant dance, you're always kind of having more insights.

11:18Yeah. And is there a method or heuristic that you use to try to predict the future or sort of think about where it's going? I think you need to have this duality of getting maximum information input, especially, you know, the kind of high velocity information like Twitter, and sort of hacker news and technology outlets and just read widely. But also, you're digesting that actually sitting down and thinking about what does that mean playing with these technologies, prototyping, you know, just the other day, I spent some time playing with Gemini Flash, which is really cool LLM that I think is underrated from Google right now, it's able to do all these things.

12:03It's the first truly multi model LLM. And I feel like there's a lot of potential there. And so you kind of like want to to see sparks of potential by having this high bandwidth kind of information processing and then like figuring out what to sit down and like spend a lot of time on and trying to think about what the future of it is. Yeah. And I wonder if you've experienced this, some of the really great founders I've worked with, they spent a fair amount of time on things that they were interested in just for their own sake. One of the things that you see with a lot of the breakthroughs is that by definition, they haven't been discovered yet.

12:39So you have to be treading on territory that's undiscovered. And therefore, most people think it's a waste of time. They don't understand why you're doing it. They don't see the immediate ROI impact or the immediate professional tangible impact. But it's like you're never going to discover something new without exploring the unexplored. And so I find that a lot of times the great ideas come from just going down in a rabbit hole. You know, you just get a notion and you're just like, this is interesting for its own sake and I got to get to the bottom of this. And then in the course of encountering the barriers of getting to the bottom of it, you start to figure out, okay, this is the thing that's missing in the future that needs to get built.

13:19And then lo and behold, there you are. You're in the future before the rest of the people. A hundred percent agree. That's where you want to go fishing for ideas. The GPT community was very small. People who actually cared about GPT-2, even early GBT3. By the time chat GBT landed, all of us who's been playing with these technologies for four years at that point, we're like, yeah, sure. I mean, of course you can build chat GBT. It's not a big deal, but the world kind of explodes. It's always these moments similar with Bitcoin and crypto, where a lot of people were just tinkering with it and found it really cool.

13:59And then you're constantly surprised that the world has not woken up to this until they'll finally like, you know, there's some tipping point where the world kind of wakes up to this. So if you want to fish for talented people, the future of technology, you want to find these communities that are tinkering. Yeah. And I can think of so many startups where the team are, you know, there's some technology trend and they're just tinkering with it and trying to figure out how to build the best thing around it. It's interesting because there's a lot of opportunity to experiment at the edges with things.

14:37But I think you're right. I think you have to develop your own independent judgment about what you think is interesting. You almost have to, by definition, if it's hot and everybody's chasing it, whatever idea you have, somebody's going to have had that idea. If you're encountering barriers to getting the most out of something that not many people are doing yet, I find a lot of the unlocks come from there. Yeah, it's almost like some form of efficient market hypothesis. If you're looking for change on the floor of the Grand Central Station, it's impossible. Someone else will find it. But if you go outside of New York and just walk down an alley and potentially look for change left there, there's a much higher likelihood that you're going to find it.

15:24I mean, there are a few heuristics. One of them is like doing hard things, you know, opening AI, spending, you know, blood, sweat and tears for five years before they found something interesting. And I would say Repl.it is kind of the same, you know, really took off in terms of like a Bible business only recently. But we had to build all this stuff. We had to build the entire platform to do that. So I think that's one heuristic is like, what is the path less traveled? and there are a few reasons for that. One is hard. Another thing is boring. Paul Graham talks a lot about that. What are things that Silicon Valley hackers are just not interested in?

16:01I love this efficient market notion that you just brought up. I'm not sure I've internalized it completely until now, but let me try this on you for size. One way to look at this might be that pretty much there's an efficient market for startup ideas. there's probably not a startup idea you could have or I could have for the most part that somebody hasn't had so like one form of arbitrage is to leverage an inflection because for the first time you can try startup ideas that were impossible before one valid reason for startups to exist at all is that inflections happen and create the conditions a window of time where where it's an inefficient market for startup ideas yeah but but you brought up another one too I think is interesting, which is, I would internalize it as patience is a form of arbitrage.

16:50Like when you're patient and you pursue something longer and with more authenticity than most people are willing to, than most people are patient enough to do, you're going to be in an efficient market as well, because you're not going to have any competitors that show up that are willing to stick with it for that long. That's a good way to formalize it. I think worth writing something about it. But something like Roblox, right? I mean, they started in the 90s and they were exploring the potential of 3D and maybe they caught that inflection. And then they caught the next inflection, which was like, what do they call it?

17:28Massive multiplier games on the internet. And then another inflection is the creator economy and having the foresight to kind of catch every one of those inflections. I think that's how you kind of build a great business over time. So in the early days of Replit then, did it get product market fit right away or did it take some time? To some definition of product market fit, like if we're talking about a commercial viability, if we're talking about users, yeah, I mean, we've never had trouble onboarding a lot of users. And the people was providing the most value to was the people with no disposable income.

18:10You know, students, sort of hobbyists, struggling entrepreneurs, low income communities and countries where they have to like code on their phones or the laptops that don't work very well to install. And for the developers, they really loved it. But like they wouldn't pay for it necessarily because the alternative is free. you needed to actually expand the market with AI such that there's a lot more people trying their hand in programming and making software that enough of them find value from it and actually either make some money or makes it more productive at their work or makes their small business better or some capacity like that or they save money on it such that paying for it is a no-brainer.

18:56And every day I open Twitter now and I see someone saying, I saved$100 ,000 on this piece of software that I used to buy by generating it for$5 on Replit. And throughout all of this, we kept building the components necessary to get someone all the way from an idea to a deployed app, obviously the editor, the underlying runtime, the language support, the package management, the databases, the deployments. And so imagine it as like GitHub, Visual Studio, AWS, or Heroku in one platform. And it's the first and only of its kind. And it took the patience to build all these tools. And another way in which having this massive scale helped us is to optimize cost and fight abuse and become a secure platform.

19:50A lot of platforms that are coming online now, a lot of coding tools are facing massive abuse, are running at an insane negative gross margins. You have to go through the time and pain in order to solve these problems. When AI came on the scene, our advantage is like, hey, we've been through all of this. We have a user base. We have a reputation. We plugged in AI. It was obvious that this thing is so valuable that you would have to pay for it. And that's when we took off. Yeah, it's like you go from pushing to being pulled. And that turning point is not always obvious. You kind of know it after the fact that it's happened, but it's not all that obvious.

20:31You're still dealing with cutting through the jungle with a machete day by day. Was there ever a day, though, when you reflected and kind of said, okay, this is clearly working? Yeah, it was September 2024. We launched Rough Lead Agent. and although we had like sprinkles of AI all over the product, this one was, okay, this is the mainstream of the product. You start from a prompt and you develop the software in a conversational manner with the agent. You can still go code and use some of the other AI tools, but we force you through the prompt-based thing as the initial kind of creation experience.

21:11And the thing that created that sort of conviction or reaction was the scaling of ARPU. Like most prosumer and consumer applications call the net dollar retention is always less than 100%. You're churning users and it's hard to capture more and more value from users. With Replit, we see individual users spending up to$1 ,000 a month. This thing is so useful for them and their business and their lives that they're spending sometimes hundreds of dollars a day. Of course, we're adding users, so horizontal expansion. Vertical expansion, meaning people are buying more services. So we built this usage-based infrastructure.

21:59Again, man, we've just been through these slogs for a long time. We built this infrastructure to charge for more and more services. So we have a database product. We have object storage product. We have a compute deployment product. We have all these things. And so the more you use the agent, the more it's going to use of these products and the more you're going to be able to build more complex things. And once you put these things in production, they're going to drive more value and more consumption. It's a cloud product. But it's a consumer quality cloud product. And so everyday people are becoming cloud customers.

22:38This is when I felt like we had something really, really special. Even if some people don't like it and are turning or don't know how to adopt it and make the best use of it, just the fact that we have these users that are continuing to expand their usage means that we can eventually get to more and more people and a billion people maybe at some point that get a lot of value out of this thing. How far do you think you can take this then? Like at the limit, do you want everybody in the world to be a developer? Or like, you know, is there a boundary condition where you don't think you apply to certain people?

23:19Or are you just trying to push the boundaries as far as you can and see how far you can go? There are two components here. One is your bet on AI and how far AI could get, at least in the near term. another one is what are the skills required to be able to be good at like working with an AI agent and just to start with this one no matter what your background is you need to be a good systems thinker even if you're like an artist or nurse or or what have you you still need this skill of breaking down problems and thinking about it in a way where you can describe what you want the agent to do next and what do you want it to try and being able to have this almost search process in the space of solutions.

24:09The question is how many people have that skill? Is it intrinsic? Is it learned? And that will determine some of the addressable market for this. My guess, it's learned. I think that some of it is intrinsic. I think you're really intelligent. that comes really naturally to you. But I've seen people kind of learn it. This is where education is going. If you believe that some form of AGI or AI is on this tremendous trajectory that's gonna eat more and more of our routine jobs, then the work of people becomes being able to direct these systems in a way to create value in the world. And to do that, you need to be a good systems thinker.

24:53Yeah. It's funny you say that. We talk this way internally as well. I have this database of startup companies where you would have made a hundred times your money if you'd invested. And I want to understand those companies in the way that Buffett and Munger understand Fortune 500 companies. And so I try to get a complete time capsule of exactly what it looked like at the seed round. And what I find is if you do that and you're curious, you get rewarded in unexpected ways. So then I started to create some AI scanners, like one will scan a pitch for Clay Christensen's jobs to be done, and another will scan it for the business model canvas, and another will scan it for SWOT analysis, and another will scan it for Peter Thiel's zero to one, eight factors, and another will scan it for Sequoia Capital's investment rubric.

25:46And then another one will scan it for the pattern breakers insight stress test. You have to have been curious enough for a very long time about what makes a great startup. But once you do, and once you start to create this lattice of mental models that you can apply against these companies, you start to find patterns of how they cluster. You start to learn things about the founders, about the opportunity. How long did it take to get to 10 million revenue, how long did it take to get to 100? You're able to do that analysis across 200 companies. And you couldn't ever imagine doing that even a couple of years ago.

26:24But now it's almost like you can be a tiny little organization and you have the force multiplier capabilities as if you had 1 ,000 people. I'm curious, has it proven predictive power yet? I would say rounded off to the nearest yes or no, no. The thing that it does let me do is it lets me sanity check things much better. So I can kind of, I can look at an opportunity and I can say, I don't want to forget that jobs to be done is important. I don't want to forget that knowing what type of business model it might be is important. I don't want to forget that I'm looking for certain things about the founders.

26:58I'm looking for certain things about the problem it solves. And so it's almost like before a plane takes off, you check all the gauges and the switches and the knobs and you have your pre-flight check. There's elements of that where now you can do that much faster. But in the end, you're sort of looking at this founder and you're saying, do these people have this stuff? And I don't think AI is there yet. This is so stupid, but I just want to run it by you. One way I kind of try to suss out founders, I mean, it's not the primary way, but I try to imagine them being super rich and super successful.