I made $250M in one year. Here's what it actually cost me

We went from $2.5 to $250 million in one year.

To 250 in one year.

We're on our way to a billion this year. I'll just say that.

The idea of like losing to some kind of a competitor is like the worst feeling in the world. I'll call, I'll show up to their office. We'll do whatever we need, and we rarely lose deals.

I think for the first time, because making software is cheaper, you can actually create a multi-million dollar business and not have to raise venture, not have to grow the team a whole lot.

I think we are in the singularity.

I'm having mixed feelings, Sean.

Yep.

We're on our way to a billion this year. I'll just say that.

Yeah. And it stayed two or three million bucks for a few years as we were like navigating what do I actually do. We went from 2.5 to $250 million in one year. That was between '24 and '25.

We did. We actually passed a PwC audit a couple months ago.

And by the way, you know, we're gross margin positive, which is, you know, something that is rare in the AI industry, but by a good amount as well. And we're very precise in how we calculate run rate.

You know that the theory for one of the potential creators of Bitcoin is that he invented Bitcoin because he was making so much money he didn't know where to keep it. There's this story called The Mastermind about this guy who I think grew up in Rhodesia, and that was Zimbabwe, and he was a hacker. He was a computer hacker and he invented a lot of encrypted technology, very well-known guy. He ended up getting into criminal online activities, selling drugs online, one of the first to do that. And he generated so much cash he actually had ships with bullions of gold just in the international waters just doing circles.

And so there's a lot...

Paul Le Roux. And there's a lot of circumstantial evidence that he was potentially writing Bitcoin because it was like, I don't know what to do with this money.

He created the first like delivery online drug store.

He had like thousands of domain names. He was sending millions of emails a day. He was basically one of the first spammers, like internet marketers, and the monetization engine was selling drugs.

And now there are a lot of startups that are doing it, you know, quasi illegally, where they have a lot of doctors that are sort of approving those prescriptions and things like that. But yeah, he did it fully illegally and was wanted by the FBI and Drug Control and DEA and all of that for many, many years. And they caught him around the same year that Satoshi disappeared, around the same time actually that Satoshi disappeared.

Well, first of all, I have such high expectations of myself and my team that at any given point, whatever progress we made, the question is why haven't we made more if we're...

I am the Asian mom of myself more than anyone else, which is I don't really recommend. It's a kind of a miserable way to live. But you know, it's sort of like I was like, of course we're going to make this much money. Of course the company is going to grow this much. It was always the plan. It was always on the cards. That was always the vision. I'm also naturally paranoid and so I was fully expecting it to reverse at any given point. And so it was like, how can we make sure this revenue is sticky? How can we go up market, go to the enterprise, make sure our customers are getting a great experience, make sure that things that they create is not slop, it is secure, we're creating trust not just in us but in the whole space and that this vibe coding thing, it's possible, including doing podcasts like this, is not only possible, it's viable to build businesses. And so the first reaction is adrenaline and like, let's get to work, like this is an opportunity. Like if we had sold or failed before the breakout point, I would have been probably depressed about it, but not as bad as you now have the wind in your back and if you fail, it is fully on you. You've lost, it's almost like someone passing you the ball and there's no one else in front of the basketball ring and you could dunk, you could do whatever you want and then you somehow do a two-shot and fail. So that's the feeling. This incredible pressure from everyone, from the market, from even investors went from, hey, you got to do a layoff, you got to cut burn, to go invest, put money. Do you know, it's like a thrill of success only lasted like a week or two and then became sort of the reality of like how do you keep this motion going and how do you scale it. And the first thing that starts to break is people. When you have a supply and demand imbalance, companies start knocking on our door. It's like, hey, we want to buy this product, our employees are using it. How do we get an enterprise deal? And we're like, yeah, we have one person who does three other things, but also doing sales. They're like trying to close deals as fast as they can. I was like, okay, I've never done sales. How do you actually do sales? But there's something about the universe where these people land in your life at crucial points. Not always, but sometimes. And we had a VP of sales that came to us at the moment when the company was really in its darkest moment. We were about to do a layoff and I'm sitting in this meeting room and I'm asking him like, why do you want to join us? His name is Patrick Purvis. He had previously been a VP of sales at Zuma Info, exited, took a couple years sabbatical, independently wealthy guy. And I was like, look, the company kind of sucks. Why do you want to come here and...

And he's like, look, I spent the past few years reflecting. I want to do something meaningful. And your mission of empowering people and democratizing software just seems one of the most important things to... I was like, okay, welcome to the team. But it's not, we don't want anyone for you to sell.

Yeah. As you call, being the Asian parent, I was also kind of financially responsible and so we never had like, we never was like, oh, we're weeks away from not making payroll. So when we did the layoff, it was more like about extending the runway.

Yes.

Yeah.

For a long time. Like I just have high pain tolerance, I think.

Yeah. I think the worst part about it is the belief that your team have in you, your vision, your leadership. And when that goes away, you can see it in their eyes. And that is the most hurtful and depressing feeling. And so after we did the layoff, the layoff is a bubble bursting event.

So we actually wasn't that big. It was like a third. So we were like 120 went to now to 90 or something like that. And by the end we were 60. By the end, like 3 months we were 60. And so we lost 50% of our team, but most of it was voluntarily. And the reason is because suddenly, you know, I've been for years getting in front of everyone and talking about this vision and mission. And suddenly it just sounded like lies and sounded like I don't know what I'm doing. And we had also at the same time moved our office from San Francisco to Foster City. We got a huge office. I really knew that we were going to hand it off and scale, but I was off by a few months. So we had already gotten this office and so the office was empty and cold and kind of dark. And so every day I go to work and just like the mood is super dark. And I know that at the moment I'm going to get to my desk, someone's going to walk over, they're going to tell me they're quitting. And so every night, the whatever sleep I got was like, okay, who's going to quit the next day? And it just kept happening. And that was really draining. And for me to put on a face and tell everyone that things are going to be okay was incredibly hard. However, there were a group of people that were working on Replit Agent. Replit Agent is the breakthrough product that we created. It actually created a breakthrough in the entire industry because this was, remember, this was before Claude Code or anything came out. We showed the world that it was possible to build end-to-end coding agents. And the people inside that team had almost a secret. We're on the precipice of a breakthrough. We were all kind of dogfooding it and playing with it and having a lot of fun with it. So whenever I go to the office, everything is depressing except this one room we call the war room. And I walk into the war room and the mood is intensely different. Everyone is super pumped. They know we have created something that's going to be super valuable. So I had this night and day, almost schizophrenic feeling of, you know, large parts of the company and the business totally lost faith in me and the company, and then a small part that were totally engrossed in the possibilities of AI and what we were inventing.

Why would they stay?

But the other one is you mentioned investors, it's just a broader industry. And you know, I remember my calendar emptying and I'm no longer invited to the hot parties in many ways. When I go to one, it's like, oh yeah, Replit, oh I remember that thing, you know, and...

And our partners were not including our logo anymore and our vendors removed our logo.

And it felt like, you know, a big part of my identity has been about Silicon Valley. And it's about I've read everything I could get my hands on, every biography, and read about all the different figures from the biggest to the most obscure. And I just felt like the weight of Silicon Valley on my shoulders because I was backed by the best as well. Like Marc Andreessen, Paul Graham, David Sacks, you know, all these people are on our cap table. And I was like letting all of them down.

We launched Replit internally at Replit for employees outside of the core engineering group working on it. And the first time everyone tried it, I was especially paying attention to non-engineers. Are they going to be able because that was the mission? Are we going to enable them to be actually successful using this? And you know, Jeff, Jeff is our head of partnerships, Jeff Burke. And I always take Jeff as like the prototypical potential customer because like he's a consulting background. He's super sharp but he can't configure Python to save his life, right? So he tried it and I remember him posting the feedback. It's like, oh, this is miserable. I failed. I couldn't make the simple app the first day, the next day, third day. I think one day when he posted was like, oh, I was able to make this thing. We're like, okay, we got it. I think the product is ready right now. And it was around August in 2024. And the team felt they weren't ready. I was like, we got to launch this. I don't care if it is semi-broken. If you get 50% of the time get amazing results, it's going to wow the world because for the first time an agent can not only write the code, debug it, create a database for you and deploy to the cloud. This has never been done even if it is in just like demo shape. And one of the habits that I picked up during this darkest hour is playing video games. I hadn't played when I came to the US. I was like a pro gamer back in Jordan. When I came to the US, I was like, okay, I'm going to focus. I'm going to be the Silicon Valley entrepreneur. I'm going to work really hard. I want to make a lot of money and I want to be successful. So I just dropped all the hobbies that I had. But I bought a Steam Deck because I just wanted something to take my mind off all the depressing stuff. They do this thing in games. It was called early preview. Sort of like a beta, but the intention is that you're getting a semi-broken thing, but you're expected to kind of give us feedback. I was like, okay, we're just going to release an early preview of the product and we're just going to call it that. We're going to say, hey, don't subscribe to this thing because it might be buggy, but if you're already subscribed or you're okay with getting buggy software, join it. So in September 2024, I posted a video of me in the office shot on an iPhone saying, hey, AI coding is amazing, generates a lot of code, but so far you had to do all this other stuff. Today for the first time we're showing you that an AI agent can do the end-to-end thing. And there were a few moments when this tweet started going viral. One of them was Andrej Karpathy, who is head of AI at Tesla, early OpenAI AI researcher, co-tweeted and said this is the aha, the AGI moment. And so that demo showed the industry luminaries what AI could do. People inside research firms like OpenAI, Anthropic reached out to us and told us we did not know our models are capable of doing that. And so that moment was like, okay, we have something here. And then the revenue first day made like a million dollars of ARR. Second day, $2 million.

Yeah.

But at the same time, we were just hustling to make the product better.

Yes.

You know, I've always heard from other entrepreneurs is that the feeling of product market fit is like stepping on a, I don't know who said this, it's like stepping on a landmine. And I knew that we didn't have it for a long time. That's why pivot, pivot, pivot. And then that felt like stepping on a landmine. I was like, okay, we got it now, we go from there.

But I sort of, it's almost bad to talk about that. Not talk about it, but I don't think you should expect that because I think that most great companies have never had that. If you look at the subset of companies, most don't work. Some work okay. A handful work a little bit better and then a very small percentage are amazing. And even amongst the amazing ones, what they're describing, what you guys are describing, they don't feel that. Like for example, we were talking about like Mars Candy the other day. Like that's been around for 150 years. And more likely than not, it was like a grind, a grind. This year was not a grind. It kind of did pretty good, but this year it was kind of a grind. You know what I mean? Like there was never like this moment where it's like we just grew to a billion in revenue from 2 million in 24 months. I think you're right that it doesn't happen right away even for you, even though it happened super fast, it's like 10 years in. But there is a difference between push and pull. And I think the companies I've seen, which is a very small handful, that actually have product market fit, it starts to feel like pull. In fact, Emmett, the CEO of Twitch, I remember asking him, you did Justin.tv, then you guys pivoted to Twitch, was it obvious right away? And he basically said product market fit's the main thing that matters. He goes, basically you feel like you're pushing a boulder up a hill every day and then at some point you kind of wake up and you realize it's like you reached your hands out but the boulder had already moved and now you have to start running to catch up to it. And suddenly the whole game feels like a sprint catching up to the boulder going down the hill. It takes time but there is a difference between like you are manually pushing the crank versus it starts to get pulled by the market.

Just to offer like a distinction between the different types of businesses. There are, you know, Clayton Christensen, RIP, passed away recently, Harvard Business School professor, came up with this concept of disruption, right? Disruption became sort of a meme, but it's actually an economics theory almost about technology. And he talks about sustaining technology versus disruptive technology. You can be innovative on a sustained curve. You can make things incrementally better, but there are things that are disruptive and those are like moments of time businesses or technologies that create this explosion of demand because for the first time it became possible, right? So if you Mars Candy, there was a lot of other candy. Maybe you're really good in your recipe, you're really good in your execution. Most businesses are trying to compete in a more or less zero-sum market. So they're trying to capture market from existing demand on existing competitors. Right now there are market creation moments and I think for Replit this was a market creation moment and most internet, right, whether it's PayPal, Facebook, Google, the reason you have this explosive growth is like you just invented something new and then everyone's rushing because there's capability that didn't exist before and there's no other alternative by the way.

I mean, it depends what you're trying to build. Are you innovating something new? Are you creating something that wasn't really possible? Or are you entering a business where you're like, okay, I know you've done the business plan, right? Here's what the market size is. Here's how much we're going to capture from it. Here's how we're going to be better than competition. If that's the case, you're going to slog. And maybe every now and then you have some kind of innovation or some kind of breakthrough that gets you a bit of a jump over competitors. But prepare yourself to be very good at execution. Build an amazing team. Build amazing processes. Just be very good at the day-to-day execution to outcompete your competitors. If you're a young guy trying to make something that didn't exist before or you know like Sean and some of the guys in social media, they're trying to create a new, you know, Clubhouse or some kind of thing that I didn't know I would enjoy talking to people, you know, taking a walk on my earbuds. This is just like didn't occur to me. So if you're inventing something like that, you will, you know, the point is pivot, pivot, pivot until it hits because if it's not hitting, you're not hitting on some human nature or element that you're finding or you're like finding a secret in the universe almost.

Yeah.

Yeah.

So growth by building tools.

So growing the product by building tools for education, essentially people learning how to code, and then build a simple network and AI-assisted interface that blurs the distinction between learning and building. That's essentially what Replit is today. Like you're learning how to make software but you don't need to sit down and learn how to code. You're just like vibe coding and it's just happening as you go and you're kind of picking up these skills. And then evolve the platform into a place where people come to learn, build, explore and host applications. So not only is it the development environment but the hosting of the applications. And the hosting of the application I think should have another point because that unlocks the next step which we're getting into now, which is monetization of those applications, scaling of those applications. Right now if you go to Replit, you build something, you can just say monetize it and it will integrate Stripe for you and all of that. And pretty soon you should be able to say market it and like...

This one is a little controversial, but Medv, the billion-dollar one-person business covered by the New York Times.

GLP-1. And the reason it's controversial is because people don't like their marketing practices.

The entrepreneur I think already fixed a lot of these issues that people are reacting to, but he runs a big part of his stack on Replit and he says it's been a crucial part of their growth because it's not only just the core application but the automations. He's been updated today. So for example, he's working with a vendor and every day he does something manual like a marketing vendor. It's like, okay, I'm going to build this website for you. You go and you speak to an agent. The agent will give you all the information you need or the agent will give you the assets. So he'll build interfaces for all the different parts of the business, all the back office stuff of the business. And he's the fastest person to go from idea to prompt. Even when he was on Twitter going viral, he was like building a game in Replit. And he built an experiment where he sent me that experiment where you can scroll a site with your head so you can read something without looking at your screen. So I've never seen someone after spending some time chatting with him be so quick to create things. So boom, create that, create that, automate this, automate that. That's one example. There's been a lot that I've spoken about over the years. In the past they used to have to migrate off of Replit because the platform was not complete, but now the platform is actually complete and you can run a full business on it. We have a lot of venture-backed companies such as Spellbook, I think is a multi-hundred-million company started on Replit. Magic School, I think a $500 million business started in Replit and many others. More recently, since Replit Agent, there hasn't been enough time for them to really hit massive scale, but there's a lot of multi-million dollar businesses. For example, recently I think they just did a round, but there's a company that's building influencer marketing for local restaurants, local shops. So you're a local restaurant, you want someone to cover your restaurant, you go to the platform, you hire an influencer, they actually physically walk to your restaurant, eat there, take a TikTok, whatever. Because now most people, especially Gen Z, they actually don't go to Google Maps anymore. If you ask a Gen Z, oh, let's find a restaurant, they're like, go to TikTok and search. And so that's already over 100,000 ARR in a few weeks.

Try Nearby.

It's surprising to me that the Google search queries keep going up. They just broke records. The revenue in search keeps going up. But yes, people around me, they're not using Google as much.

Yes. Silicon Valley have always approached problems with hyperscale in mind. And I think for the first time, because making software is cheaper, you can actually create a multi-million dollar business and not have to raise venture, not have to grow the team a whole lot. So local-style businesses is very interesting. There's a guy in England that I think he's well on his way, he's 100,000 already in run rate. He's well on his way to a million. He's building software for ice rink management. So skating rinks and all of that. I think that you go around in your life and there's a lot of things that are just not computerized, there's no software running it.

So that would be my first approach actually. That's what I did when my first business was building software for internet cafes and LAN gaming. And so that's always been a great place to start.

You know another one I would say is being lazy in programming. There's this old saying that laziness is a virtue. And the idea is to automate a lot of things, but now everyone is a programmer essentially. So part of the reason why I made Replit is because I don't want to keep setting up the freaking IDE every time. And so go about your life with this laziness attitude. What are the kind of things that are so annoying that you have to keep doing over and over again? There's probably a million, maybe a billion people that feel the same way. And perhaps you can make a piece of software to automate that. And more important than ever, this lazy mindset of viewing things through the lens of automation is very important because AI is truly like, you know, we have a magic automation machine. So not only you'll optimize your life, you'll become a lot more productive, but you might stumble on something that a lot of other people would want.

I think we are in the singularity, like the original idea of a singularity. I think Vernor Vinge, one of the early computer pioneers and sci-fi authors, talked about the singularity, borrowed this concept from physics. In physics, it's like the black hole and anything that happens after the black hole is undefined. So we actually don't know what is like not knowable. So anything that is beyond that point you can't even have formulas for because it's literally undefined. You know, a lot of the early AI thinkers talked about this moment of singularity where it is incredibly hard to predict what happens after because the pace of innovation is not only fast, it is accelerating as fast as it already is. Meaning the second derivative is exponential. And so I think this is the moment we're in where in '22 we got GPT-4 at the time and GPT-2 had come out in 2019, GPT-3 in 2020. So every two years we were getting a model and now it feels like every few weeks we're getting a model, sometimes every day, and there are these...

Fundamental capability shifts that happen either in autonomy or cybersecurity or computer use or what have you, and every one of those capability shifts has a lot of downstream applications that yet need to be discovered. We haven't actually productized it yet. So it's almost like you get a bundle of energy. That's how I think of models. They're potential energy that entrepreneurs need to go figure out how to make them into products, right? And this is why it's an explosive time for entrepreneurs and entrepreneurship, is because we have this potential energy that needs to be harnessed. Now, the reason why there's like a delay is because we have this capability overhang. The world will change, but it's really hard to know whether — I can't tell you in 12 months like all entry-level jobs will be gone.

My bet would be that I don't think unemployment numbers will move a whole lot. Perhaps they'll marginally increase. I think Dario, the CEO of Anthropic, said we're going to get to 20% unemployment or something like that. I don't know if I'm misquoting him, but he said all...

Yeah. Like I don't think we're — because like what I'm seeing from our customers, we have some customers that automate a lot of things and end up laying off some people. We have other customers that automate a lot of things, make more revenue, and want to hire more people to automate more things and create more products. And so net-net, I think there's going to be a lot of new AI companies, a lot of companies that make use of AI and want to hire more people. So net-net, I think the unemployment question will probably be neutral.

I like business theory. It's not always useful, but it often is useful. Hamilton Helmer wrote this book called The Seven Powers, and everyone in Silicon Valley talks about moats. Seven Powers is the theory of moats. The moat of these LLMs is the technology fundamentally commoditizable, right? Like it seems to be very easy to replace these models. Like, you know, a lot of developers every day, professional developers, are switching between these models in some cases. And if you're in Cursor, it's like literally one click away to switch to the next model. And I think a lot of the fighting has to do with this fact that it is hard to find a way to zoom heads too far ahead of the competition. You're always sort of neck and neck competing with others. So you might want to try to block them in different ways, whether it's in the court or with the government or regulations or things like that. Whereas I think in the past, you could — if you're Microsoft and you achieve this dominance on the PC, it's really hard to unseat you, right? And so there's this natural monopolies that used to emerge, and they're protected by certain moats such as network effects, economies of scale. We haven't seen any of those moats sort of emerge from these foundation model companies. And so my hypothesis is that maybe the one natural moat is capital. You need not just one-time capital to enter the market. You need continuous capital to train the current model, the next generation model, and the next generation model. And then you need enough installed base and revenue to cover your cost. But that is not enough moat to keep the other big players away. So you can imagine all the big companies have a fighting chance, and all the big governments can also participate in this. And maybe China, as a place where you can't really distinguish between country and company, and maybe the country is going to do to the foundation model market what they did to the EV market. Well, they'll subsidize the hell out of it in order to have a competing chance, and maybe even destroy the market globally by subsidizing it. All of this to say, I think it is good for entrepreneurs because if we ended up with a monopoly or just oligopoly and you're building on top of it, it is actually very hard to build a successful business. Because like you said, they'll come for you. But if it ends up being a technology that's kind of easy to replace, I mean, there is a place for a lot of entrepreneurs to participate.

No. To the first question, we're not burning nearly as much cash as foundation model companies. And like, you know, sometime last year, profitability was in sight. We had something like, you know, 30 years of runway or something like that. Since then, we decided to spend a lot more on especially sales and marketing. And my theory on this is, or like my thesis is, that there are a lot of businesses that will need help adopting this technology and adopting it broadly. And so we — and also there's this imbalance between supply and demand that I talked about. So I thought, okay, we just need to scale the sales team. And so the sales team was like four reps or something like that by the end of last year. I think it's going to be more than half the company by the end of this year. Which is very surprising for me to say because I was always like as sort of a tech...

I love it. I didn't know because I think I'm...