Transcription
This is the first AI agent thing that has been a mind-blowing moment for me. I am not a programmer, I am not a coder, but I can now create software. Well, that's insane! There are apps built on Replit that otherwise would take probably $100,000 of developer time, and you can build it for like $25 paid to Replit. It's pretty wild how fast these companies are scaling. I don't think in the history of Silicon Valley we've seen anything like that, even in the Web 2.0 era.
So, what is a fast ramp for an AI company? What's impressive that kind of broke the frame of how long things would take? I would say reaching $10 million in 3 to 4 months ARR. Oh my God, that's wild! Can I ask a blunt, crude question? How can I use your software to become a billionaire?
I would say building. Okay, so how do we want to start this? So, OMJ, you're awesome! You are today in a position that I think a lot of people want to be in. You're doing the Silicon Valley dream. You had this idea, you went through YC, you've now raised hundreds of millions of dollars, you're valued at a billion-dollar valuation.
So, that's today. But then the cool thing about your story is that it didn't seem likely, you know, 10 years ago. It was a very unlikely success story. And yeah, you went through YC, but you were rejected a bunch of times. Like, yeah, you're in Silicon Valley now, but you started off coding in an internet cafe in Jordan. That's what's interesting to me.
We asked you beforehand, we're like, "Hey, what killer stories could you come on the podcast and tell?" And you wrote this, so I'm going to read it word for word, and then I want you to tell us the story. You go, "Rejected four times and Rick Rolling into YC, raising tons of money, and meeting amazing billionaires." So, let's do the first part: rejected four times and Rick Rolling into YC. Can you tell the story?
Yeah, so I left my job at Facebook in 2016. You know, Replit had been a side project for a while, and it had been growing. I had been working on it like nights and weekends. It grew to a point where the server cost was meaningful, and I was like, "Okay, I guess I have to start a company around it."
So, I went to my manager at Facebook and I was like, "Look, I have this side project. Can we make it somehow a project at Facebook?" We looked into that. I sent Zuck an email at the time, and he ignored me. I was like, "Okay, I guess I have to start a company."
So, yeah, I quit my job, applied to YC the first time. We did the whole thing: the form, the video, and all of that. We didn't even get a call or anything like that. It was just like we got the rejection letter. I was like, "Okay, I have to sell some Facebook stock." I sold the Facebook stock, put half of it in Bitcoin, and then half of it into the company, or just for us to kind of live.
How much money was that? It was like $70k or something like that. What was the original product for Replit? So, it was basically an editor and a console. You could type code there, and you could run it. You could switch a language, and that's it.
By the way, Sam, have you ever used it? I was using it today before this. It's magical! Also, your tweets describing what it is, like for example, your doctor saying he wants me to track my sleep, so I just uploaded the PDF that he wanted me to fill out into Replit, and it made an application so I can upload it much easier.
Yeah, it's like pretty magical. Sean, are you able to use it? It's definitely out of my league still. Both me and Sam have joked around because we both have maybe five or six times fall started, like, "I'm going to learn to code this summer." It's like a New Year's resolution thing where you just keep saying you're going to do it. You do 20% of it, 30% of it, and you give up.
You know, we buy the Udi course, "Learn Python the Hard Way," then you start doing it, and nothing really ever stuck. One of the biggest problems was that nobody really talks about this. You think learning to code is like learning Spanish. It's like learning a language. You're like, "Okay, so how do I need to say the thing?" But before you can even do that, it's like, "Oh, I'm supposed to download this program."
So, I need to download an editor, and then I need to download all these packages to be able to use it. And then you need to, and it's like just setting up the environment is so goddamn confusing to a beginner that you don't even get to do the part where you actually write the code.
And then it's like, "Oh, how do I run the code? I got to host it somewhere. Now I got to learn how to do hosting." And it's like, "What is that?" So, there's all these things around it that were confusing. Replit solved all of that, which was amazing.
I actually did your 100 days of learning to code. It actually made it really easy. You know, if I didn't have kids, I would just be doing a lot more because you solved that problem for me.
I know I'm asking you about the YC rejection. I want to come back to that, but to give Sam maybe a little more context, I think, correct me if I'm wrong, maybe I'm making this up, I think the reason you wanted to have this kind of online editor, online environment where it's all hosted there was because when you were younger, you were living in Jordan, and I guess you used to go try to learn to code out of an internet cafe.
But that means every time you go, you have to set everything up for the first time because it's not your home base, not your home computer where you set it up once and it's there. Is that true? Is that why you felt the problem like what a normal person would feel?
Yeah, basically. Every time I wanted to do a little homework, I had to spend an hour setting up the environment. At the time, the web was moving so fast until we had Google Docs, we had Gmail, we had this client-side JavaScript application sort of revolution. I was like, "Okay, why can't I type code into the browser and run it?"
I started looking around, and it turns out nobody solved this problem. There were some experiments, and it was kind of crazy to me because it was almost like finding a $100 bill in New York Grand Central Station. It's like, "Oh, I found an idea that nobody's paying attention to."
Is that true? Because it's kind of crazy. You know, the world is big. There's a lot of programmers. That seems like an obvious thing. I mean, I'm a total outsider, so my question is: was there some technical challenge to that? Because that seems like, I guess it's easy to say things that are successful are obvious ideas looking back.
But yeah, probably two things, right? There's the technical challenge of being able to make this all work in a browser. That was not obvious. But then, it seems like the second thing was, I keep going back to the internet cafe thing because it's sort of like the hardship made the problem unavoidable to you.
Whereas anybody else who learns to code, if you're just doing it at home in America, you might do that setting up once, maybe have a little bit of the problem, but you're not running into it face-first every day as if you were working out of an internet cafe.
Yeah, absolutely right. I mean, you know, Paul Graham talks about it all the time. It's like, you know, the best startups are solving your own problem, and I felt that problem really deeply. I started working on it. I discovered why it's hard. Well, it's hard to run different languages in the browser. You can run JavaScript, but you can't run Python, for example.
So, we started writing interpreters, writing compilers to run on JavaScript. It took us a couple of years. We had a few languages running. It was a pretty rough prototype, but people started using it—my friends and people at school. I was like, "Okay, this idea has legs, so let me work on it more."
Then in 2011, we had a breakthrough. The breakthrough was we were the first to compile Python, Ruby, and a bunch of languages to JavaScript and run them straight in the browser. That went super viral. We open-sourced it, put it up on Hacker News, and that was my first experience of going viral on the internet.
I was like, "Oh my God, this is an amazing rush," and I still feel that rush. Can you put that in context for a non-engineer? Is the thing you guys did on a scale of like one to Satoshi Nakamoto solving the double-spend problem? How hard of an invention was that?
That was like the nerdiest analogy you ever could have come up with! That's what I'm here for. So, was it genius, or was it just that nobody had taken as much time as it would take to do that? Where was that breakthrough? How would you describe that breakthrough?
It's definitely not on the order of the double-spend problem, where it's like a fundamental invention. It was like, you know, pushing a huge rock up a mountain. It took so much grit and obsession to be able to hack the browser in order to run things that the browser wasn't designed to run.
I would say it is solving hundreds of problems as opposed to solving one invention. So, you were working on it as a side project for a number of years. That's a long time, by the way. Sean, can you imagine having a side project that's a hobby that takes three hours a night?
Doing that for two years is kind of a long time. No, dude, the only two things I've ever done that with are this podcast and my kids, and there's really no way out of the kids thing. The podcast was a hit right away.
The podcast gave me results right away, so it actually doesn't count. You were doing this without the kind of financial rewards or fame rewards or any other major rewards during that time. How many years did you do the side project thing, and what kept it going?
2009 was the original idea. 2011 was the breakthrough, and it went viral on Hacker News and the internet. I think that was the first time I felt a little bit of fame, a little bit of return on investment. I remember Brendan Eich, the inventor of JavaScript and CTO at Mozilla, tweeting about it. I was like, "Wow, this is amazing!"
You know, a kid in Jordan made this fundamental breakthrough in browser tech, and I'm getting this recognition. That's pretty cool. Also, some articles were written about it. People talked about it in conferences, and all that was evidence for my O-1 visa to come to the States.
Basically, my entire adult life, I'm working on this, which is crazy, right? I am 36. Wow, I think. And you've been working on this since you were 20, 21, I think. Yeah, 21. Yeah, that's a while. That's your whole life, your whole adult life.
You know, it continued to incrementally improve my life, so it wasn't this, you know, working in a room for 11 years and nothing happened. So, I get this visa to the United States, and I go work at Codecademy, and they used the open-source work that we did.
A bunch of companies in the U.S. There was this boom, and if you remember Udacity, Coursera, whatever, a lot of them used the open-source version of Replit to create interactive courses. Suddenly, the world opened up to me. I'm getting job offers all over the place, and I have choices where to go, and so we decided to go to New York.
A lot of people watch and listen to the show because they want to hear us just tell them exactly what to do when it comes to starting or growing a business. Really, a lot of people who are listening have a full-time job and want to start something on the side, a side hustle.
Now, a lot of people message Sean and me, and they say, "Alright, I want to start something on the side. Is this a good idea? Is that a good idea?" And again, what they're really just saying is, "Just give me the ideas." Well, my friends, you're in luck!
So, my old company, The Hustle, put together a hundred different side hustle ideas, and they have appropriately called it the Side Hustle Idea Database. It's a list of 100 pretty good ideas, frankly. I went to them; they're awesome. It gives you how to start them, how to grow them, things like that. It gives you a little bit of inspiration.
So, check it out. It's called the Side Hustle Idea Database. It's in the description below. You'll see the link. Click it, check it out, and let me know in the comments what you think. Naval has this great quote where he says people are always asking him about how to build a great network or networking.
What are your tips for networking? He's like, "My only tip for networking is do something great, and watch your network will appear overnight." People will immediately come to you because you've done something great.
Right? You didn't go try to get a coffee with Brendan Eich. You built something really cool that the creator of JavaScript and Mozilla browser was like, "Hey, that's awesome! I want to reach out and get to know you."
I think that's actually how you ended up getting into YC later, was Paul Graham thought what you were doing was cool. But let's go to the YC part. So, you quit the Facebook job, put half the money in Bitcoin, half the money in your startup, applied to YC, and got rejected. That was the first rejection. What were the other rejections?
Yeah, VCs kind of wouldn't talk to us. We'd get meetings with VCs. Some of them were yawning, and I think one of them slept. It was just not interesting to them. Dude, I had that happen one time as well. A guy literally fell asleep. It was like he was literally 80, and it was Friday at 4, and it was warm in the office, and he fell asleep mid-pitch.
It was warm, yeah. It was like a cold day, and it was warm inside, so I was like, "Yeah, I mean, you deserve this." But dude, what did they not see in you? Because it's so easy to look back in the past, but you seem like you got the it factor. This seems like such an obvious idea. Smart people are talking about it.
What were you missing? What was the case against it? Well, I think Silicon Valley is probably the most meritocratic place in the world, but it is also status-driven. At least then, it was very status-driven. If you look at the people who got into YC, they were Stanford dropouts and things like that.
I think since then, YC has improved and gets international people and all of that, but my background wasn't really interesting to them. I didn't have any fancy colleges or any of that. Also, being a married couple was somehow something that they thought was a disadvantage.
You didn't match the patterns. You didn't match the Stanford pattern. You didn't match the co-founder relationship pattern. You didn't match the trend of what categories have big exits. You weren't on trend at that time, right?
Yeah, and so we continued to apply to YC every season. We would send in the application, and our thesis developed more. We felt like we had started making some money. Some people started paying for our service. We had an API at the time that people paid for. A lot of educators and people learning to code started to pay for Replit.
What was the revenue when you got in? Maybe $10,000 a month. It was enough to sustain us at that point. It was like ramen profitability. But before YC, the person who actually was the first one to bet on us was Roy Bahat from Bloomberg Beta.
I knew him from my Codecademy days, and the meeting with him was so refreshing. He was a straight shooter. He would tell me, "Here's where I think the idea or the category is hard. Here is where I think the valuation should be." It was the first meeting where he just gave me everything he was thinking about. He didn't obscure anything, and I was feeling really good about it.
So, yeah, he gave us $500k on a $6 million valuation. That was the first check we got. Nice! And then how did you eventually get into YC?
So, basically, we were grinding, and the product was getting better every week. I started writing articles about what we were solving because we were solving pretty hard problems. These articles kept going on Hacker News, and Hacker News was really excited about what we were doing.
Paul Graham reads Hacker News a lot, probably still to this day. One day, in December 2017, I wake up, and there's a DM on my phone. It is Sam Altman, and he's like, "Hey, I run YC, and we're interested in what you're doing." I'm like, "Dude, I know who you are! You have to tell me you run YC!"
He's like, "Okay, let's meet. Come to this address." It wasn't the YC address. I was a little confused. So, I go there, and it was the OpenAI office in the Mission. I meet him there, and we're talking a little bit.
Then he turns his computer around and says, "This is Paul's email. He emailed Sam and told him this company is very important. You should reach out to them." He's like, "Okay, talk to PG. I'm going to give you his email. Talk to him, and maybe we can work on something to get you into YC."
So, I started this email relationship with Paul, which was really fascinating. He's a great writer, right? We talked about Replit, we talked about the problems of setting up an environment, the problems of hosting an application.
It turns out after he sold Viaweb, he started working on something like Replit. He started working on an editor where you write some Lisp, of course, because he likes this very obscure programming language.
By the way, Paul Graham is the founder of YC. At the time, he was starting to retire, and Sam was running YC. We had this email relationship where he wrote me essentially on the problem we were solving.
By the way, were you intimidated? You know, Paul Graham writing essays to you privately? Were you like, "Is that high-stakes replies there for you?" Yes! I would spend hours crafting the emails and trying to be as good of a writer as I can.
But you know, one thing about me is I was never nervous about meeting famous and established people, and I think that helped me over time because I can be myself, and I can talk to them at the same level instead of being a fanboy.
Why was that? Were you just oblivious to it, or did you have a different mindset about it? What was the reason? I felt like my life was taking on this trajectory that was not to be too superstitious, but it was this force.
I felt like everything's going to be great. It's going to be hard, but I'm meeting all these people, and things are opening up to us. When I go and meet people, my mindset is like, "I want to impress them, and I want to be able to get money from them," or like, "I have a goal."
I think having a goal when you're meeting someone actually puts you in a very different mindset than being a fanboy and just being very excited about the meeting.
Dude, have you guys seen that? Do you guys know the director Guy Ritchie? He's that British director. He's got this great story. He was on some podcast, I think Joe Rogan, and he was like, "You know, I just want to be the director of my own life, and I want to live my life like a movie."
What you're describing is sort of like that, where you're like, "I am destined for greatness, and we are taking on this amazing problem, and we are going to do wonderful things. It will be hard, but we will triumph."
I think that's actually a great story to tell yourself, and I think it's very motivating. It makes life more exciting. I think that's really cool.
Yeah, so I actually wrote a blog post. The title is "Do What Makes the Best Story." The idea is when you're faced with decisions where there's no obvious answer, like a fork in the road where the pros and cons are sort of the same, the heuristic I use in my life is, "What is a more interesting story?"
Obviously, Elon talks about this. The most entertaining outcome is most likely. I wasn't thinking about it in terms of entertaining, but in terms of what makes the story interesting. If my life was a movie, what would be exciting about that story?
For example, when I was in college, I was coding all the time, and I wasn't really going to class. I was failing a lot, not because I was failing the exam, but because they would bar me from the exam because I wasn't showing up.
I decided to hack the university to change my grades. We're not talking metaphorically. I actually hacked into the servers and changed my grade. Is that what happened?
Yeah, I went into the basement. I spent like two weeks. I did the, what's his name, the famous inventor Michelangelo or something like that. I did his sleep polyphasic sleep where I worked four hours and then slept 15 minutes.
It was sort of like I was writing on the wall. It was like a full-on insanity. Were you angry? Why did you decide to act? I know so many smart people who work so much harder to cheat or get around the thing than just doing the thing.
There's like a 50% of the time they end up being losers, and then 50% of the time they are in fact like the greatest. They're on this podcast. Well, I think it is like some ADHD, right? You can't sit in class, but if you're interested in something, you're going to hack and work on it a ton.
I almost got away with it, but the servers at the university crashed, and it crashed on my record. One of the administrators there gave me a call, and he said, "Look, there's an anomaly in your record of your exam in school, and it's crashing our databases. Do you know anything about it?"
I was like, "What's the anomaly?" He's like, "There's a field in the database that says you're barred from the exam, and your grade should be 35. That's the default grade of failing the exam. Instead, my grades were like 75, 90, whatever."
That's what I entered into there, and I didn't understand that there was another field. By the way, that's not good design for a database. Since then, I had a fork in the road. I could lie, and I think I could get away with it and just say, "That's a bug on your side."
But I was like, "What's the most interesting story?" Is it they catch me, and it becomes a story that people talk about? I was like, "Okay, I'm just going to come clean and tell them what I did."
So, you're like, "Better than getting the grade would be getting the reputation." Yes, exactly! So, you tell them, and then what happened? Did they kick you out?
No, so I'm kind of a convincing person. I go the next day, and it's like all the Deans are there, and they're discussing my case. They're trying to find out what I did, and they're all computer science Deans.
I went in there, and I changed the subject to technical aspects of the hack. I drew on the whiteboard and showed them what I did, and they were very impressed. It's like a Good Will Hunting moment.
My interpretation back then is like, "I'm a loser. I'm failing everything, right? I don't show up to class." It is kind of like Good Will Hunting. Then, they say, "Okay, you have to go talk to the president because I think he's going to make the final call."
So, I go to the president, and he's a very intellectual person. We talk, and I tell him, "Look, I have this talent, and I feel like it was undiscovered. I feel like I was treated unfairly, and I used the university as my sandbox. I didn't mean to do anything bad."
He gave me the Spider-Man line: "With great power comes great responsibility." It actually affected me, and I was like, "Okay, I need to do something to pay back." I told them, "I'm going to work this summer for free to make sure I secure your databases."
So, they let me off the hook, and they're like, "Awesome!" What a great story! Dude, that is an amazing story! Sam, by the way, would you ever want to compete with OMJ at anything?
No, this mentality is scary. I would not want to. You're like, "Excuse me, Dean, have you heard of the word prodigy?" You're like, "My talents haven't been used well at this university. I accept your apology."
Yes, that's so good! Okay, so I love the principle of "do what makes the best story." I love the hack story. That's amazing. Where did we get here? We were talking about YC.
Okay, sorry, let's get to the story of YC because that's where we started. So, Sam's like, "Yeah, you should do YC. Actually, the batch starts tomorrow. Why don't you fill an application? It's just a process you have to do, and we can do a later interview tomorrow."
I'm like, "I'm going to fill the application again? You made me do it like four times! I don't want to do it again." So, I kind of do a bare-bones application about Replit, and then there's the video. I'm like, "Yeah, man, I don't want to do the video."
So, I pasted a YouTube link. We go the next day. By the way, for people who don't know, the YC application is like one page. It's like six or seven questions, but then they say upload a video, two or three minutes, you talking about your startup.
So, that's the video part, and then the interview is 10 minutes where they just rapid-fire. You have 10 minutes, and it's like this make-or-break thing. It's less than a lunch, you know? Less than a job interview. It's more intense, so you're waiting around for that.
Yeah, I mean, my view was they recruited us to YC. Why are you making us do this stuff, right? So, they call us to the interview, and I walk in. There were Jared and Adora and all these amazing YC partners, and there was Michael. He was the CEO at the time.
I shake their hand, and I shake Michael's hand, and I felt like his grip was a little too hard. I was like, "Okay, that's fine." Then I go sit down in the chair, and the moment I sit down, Michael looks at me and says, "Why did you call us?"
Oh my God! I'm like, "You know, we applied several times, and I thought it'd be fun to do, and I thought this interview was just a formality." He's like, "That's not how you get into YC." He was very, very angry.
Well, it turns out when we were sitting outside, they were getting red inside, right? Say, imagine their mindset looking at the application and getting the Rick Roll song. Then they give us a very tough interview.
In that moment, did you realize, "I messed up"? Did you realize how I was coming across? What was your mindset there? Like, they must be thinking I was nervous. I was very nervous and regretful immediately because you probably—it's like, "Oh, here's this entitled tech guy."
Just another tech entitled guy when they don't know you're like an immigrant from Jordan who's scraped his way here, right? The reality and how you were coming across weren't connected in that moment. No, they weren't at all.
So, I tell Ha, "Okay, this is done. Let's call an Uber and get back to work. We don't need to get into YC." So, I call an Uber, and just before I arrive, I receive a call. I take the call, and it's like, "Hey, it's Adora. You got in! Come back. The kickoff is about to start."
I was like, "What? Are you sure?" She's like, "Yeah, come back, sign the paperwork, and get started." I was stunned the whole day. It was, you know, we start, we go to the dinner, and I'm like, you know, phased out.
But it was really exciting. People who have never been to the YC office in Mountain View, it's all orange—bright orange—and the lights and everything. It feels like a cult-like environment.
Is it like—I think I've seen the inside. Isn't it like they have a steeple or is it in one of the rooms? It’s like a triangle, like a church almost. Yeah, exactly! Sam gets up and tells us what the experience is going to be. It's like, "This is the hardest time you're going to work. You better tell your friends and family that you're going to go away for three months. You can't help them move or all of that. You just got to be focused on work."
Ha and I took it very seriously. I was like, "Okay, these three months are very important for the success of the company." We transformed the product in these three months. It went from a simple sort of editor output to a place where you can host applications and build real things—all in three months.
We were working 12-hour, 13-hour days, and it was only three of us at the time. Our first employee actually was sort of a runaway kid. He grew up in California, a little down south, and he didn't want to go to school.
So, he leaves his home, goes to Hack Reactor, and becomes a programmer. He was 18, looking for a job. I knew the guys at Hack Reactor. They used Replit, and I was like, "Send me your best programmer."
He's like, "Look, this kid is a little awkward, but he's the best." So, he comes in, and he nails our interview. Music to my ears! Yeah, and basically, I like him. It mirrors kind of my life story a little bit.
Where's this guy now? We got him some liquidity after six years of working. I felt that's the right thing to do because he was kind of burnt out and didn't want to continue. So, I called my brother in Jordan. I'm like, "Look, you got to come out here when YC. You need to."
He's a programmer. I taught him programming when he was a kid, and I was like, "You got to come help us." He's still with us today, and I called my friend from Codecademy, Moody. He's still with us today as well. I'm like, "You got to help us. You can do it remotely."
So, we assembled a team of five people essentially, and we went really hard. We were one of the hottest companies in YC at the time.
Can you give some sense of the scale of it now? You know, I invested in it probably a year ago or two years ago, something like that. I don't know when, but the numbers were off. You had user growth. Your graph looked like a hockey stick because you zoom out, and it ignores all of the little years where nothing was really going on.
But you have this crazy growth. The crazy thing about it is that your growth was developers. So, it's like, you know, one developer user is worth, I don't know, 10, 20 times just like a normal internet user. But you have this crazy hockey stick growth of developers.
Can you talk about—can you just say a couple of like permission to brag? Can you say a couple of brag-worthy stats that would impress us?
Yeah, so Replit was very easy to get started with, and so people would start using it in college or high school and continue using it for many years. It was sticky, especially for junior developers when they're starting out.
It was spreading on its own, like word of mouth. You know, there was a viral component to it. People can share URLs, and then suddenly you're in the same environment as them. We have this multiplier coding experience, and so people were collaborating.
Also, COVID was really great for us because we were, I think, the only collaborative editor experience on the web at the time. A lot of people were remote and needed something to work with each other, and so Replit was adopted at that time.
The growth was off the charts, and the servers were going down. The marginal user in any web app is sort of like zero cost, but for us, it was, you know, we tried to optimize it a lot, but it was still on the order of like $1 to $5 a month.
The growth was off the charts, but I have to admit it was hard to monetize at the time because developers were actually not used to paying for things. Now they kind of are paying for things because of AI, but at the time, they weren't.
As we added limits and things like that, they felt like they could move on and set up their own developer environment. It took a lot of creative thinking to figure out how to charge for people. Ultimately, AI was the thing that people are paying for.
The reason is the productivity benefit of AI is obvious. People were like, "Okay, this saves me time and makes me a better developer," and so people are paying for it right now.
Can you give any indication on how many users or how much revenue the business does? You know, signups? We have like more than 30 million, I think 35 million users right now. In terms of active users, it kind of fluctuates, but you know, two to three million a month, probably 100,000 apps hosted on Replit because you can build an app and deploy it all in one environment.
In terms of revenue, I can't share right now, but especially this year, it's been exponential growth. Sam, check this out. This agent thing—I got to show you this. So, you haven't used this, right, Sam?
No. Alright, so watch this. So, yesterday, I was like, "I'm going to mess around. I was doing research for this," but I was like, "I just got sucked into Replit." I started doing that, stopped doing research.
So, I go, and I go to Replit, and it's changed because now when you open it, before, it would be like, "Here's a coding screen with a blinking cursor, and it's like write some code." I'd be like, "Oh cool, I don't really write code, so I don't know how to use this product exactly. Maybe I could learn to code. Maybe I could pay somebody to build something on here, but whatever."
I was stuck. So now you open up Replit, and it just goes, "So, what would you like me to create?" So, I go on there. Watch this. So, I give you the exact prompt I said: "Build me an app that will text me every morning asking how I ate yesterday. Let me answer via text messages and then track the results on a monthly calendar grid. If texting doesn't work, you could also use WhatsApp or something else."
Okay, so basically on the right here is just like the chat, and it just goes, "Absolutely! Let me propose what we'll build." Then it just kind of explains to me like a project manager. It goes, "I'm going to help you create a food tracking app through SMS messaging with calendar visualizations. We'll start with the SMS. Later, we can add WhatsApp as an alternative."
It's like, "Okay, okay, prioritizing things. That's interesting." Then it goes, "The app will send daily messages," blah blah blah, and then it goes, "How would you like me to proceed?"
There was like, you know, "Add more features, change the instructions, or like go ahead and build the prototype." So, I clicked "Build the initial prototype," and then literally, I don't know if you can see this, but it starts like auto-scrolling as it's writing code.
Like, this is all just code it's generating. So, you know, I'm not doing anything. I'm literally sitting back with popcorn while this is happening. So, it's like, "Here's your calendar grid," and it's like, "Hey, I need to use Twilio for the SMS."
It decides I'll use Twilio for the SMS. "Can you go to Twilio and give me your account and your phone number so that we can use Twilio for sending SMS?" So, I go to Twilio, I give it my SMS, and then it's like, "It made it! It literally made the thing exactly how I want."
So, this works now? Yeah, I actually got stuck on the Twilio step because Twilio has to verify my phone number, so it hasn't verified it yet. But I can go into—I see it tries to send me the message, and it's just awaiting Twilio verification to be able to use this.
So, I'm like a little bit stuck there, which is like a common thing with agents, I feel. It's like almost absolutely incredible and then kind of frustrating at some point where you have to fight through some walls.
Well, I think I just tweeted. I think he said, "I want people to be able to build an app faster than they can just Google the answer to a question," and that's exactly what happened here.
Well, that's insane! So, this screenshot is the agent looking at the result, trying to verify. So, this is not a running app. If you click run, you can see it running. It says, "Top took a screenshot," and then it shows it to me.
It's like, "Hey, is this how you want it?" I was like, "Oh," because before it had it where it was like not the right month on top. I go, "Oh, put the month on top. Don't say monthly food tracking, right? December."
And then it also said, "Hey, would you like any other style improvements? I can make it broader. I can change the color scheme." I'm like, "Dude, this is literally better than an employee!"
Right? First, it's instantaneous. Second, I don't have to pay somebody to sit at a desk to sit around waiting for me to do something. I had an idea on a whim, went to Replit, and did the thing with the agent.
This has been a few mind-blowing moments for me in my tech career. You know, I graduated in 2010, so I'll start at that point where it's like the first time I took an Uber. I was like, "Holy [expletive], that was amazing! I pushed a button, a car showed up, the guy got in, I didn't even have to pay for it. It just paid through my phone."
That was magic! I could see him on an app getting closer and closer to the restaurant. That was like one of them. You know, ChatGPT for sure was another where I could just tell it to make something, and it would write something for me.
This is another one of them. This is the first AI agent thing that has been a mind-blowing moment for me where I am not a programmer, I am not a coder, but I can now create software. This is amazing!
Can I ask a blunt, crude question? How can I use your software to become a billionaire? Because I see this, and I'm like, you know, the ridiculous analogy that I use is I feel like an artist sometimes where I feel like I have the ability to conceptualize certain things, but I can't paint.
It's like I can't [expletive] paint what I want to paint that's in my head because I literally don't have that skill set sometimes. So, I'll be working on stuff, and like, "Dude, I want this to do this," but I got to go talk to this developer, and I don't want to have this conversation, and that's just like a pain in the ass.
So, like, you basically are making it so I can finally express myself easily. I like how you're on the first date. You're like, "How can I get you to take your clothes off?" What you're like, "How do I use your thing to get really rich?"
Yeah, I mean, that's basically like—and you had on the document, you're like, "Here's just the opportunities. Just use Replit to do X, Y, and Z." I want to go through that because this is amazing.
This is actually, you know, there's like a viral clip on YouTube or Twitter, a bunch of places where it's like the headline, which we probably have used, which is like billion-dollar one-person companies or something like this. You're the closest person to this probably to that question to answer that question.
Yeah, so there are apps built on Replit that otherwise would take probably $100,000 of developer time, and you can build it for like $25 paid to Replit. I will say that there are limitations, right? It is not perfect.
This is like the worst it's going to be. It sometimes gets stuck with problems. You need to have some skill in prompting to coax it to figure it out, and it sort of teaches you over time because it tells you what it's doing as it's editing the code.
Over time, you're learning how to use it. You're actually learning how code works. You're learning how—maybe you're not learning how to exactly type code, but you're learning the different components and where things could go wrong. You're learning about databases.
We have a database. You can go in and look at the tables and look at what's happening. The vision for this is that that's all you need to build an entire startup. Every day, we're inching towards that.
I talked about pushing the boulder up the hill, and I think that's one of my talents. It's like, "Okay, what are the problems that you can make progress on every day and every week such that in a year's time, you have this exponential progress, and the product is so much better?"
The other thing is we're riding this wave of the foundation model getting better. Every time they get better, we plug in a new foundation model, and the product is suddenly better.
So, I think in a year's time, it's going to be really mind-blowing. In a couple of years' time, I think we're going to see stories like someone getting super rich making an app on Replit that goes viral.
We're adding Stripe integration right now. You can already use kind of Stripe on Replit, but we're adding integration that makes it super easy to start monetizing your app.
So, Sam said, "How do I get rich?" and you're like, "Disclaimer, it's not fully there yet." But now you still have to answer the question. I mean, the question is like, what kind of applications? What are the ideas? What kind of applications can you build?
I would say AI applications are growing really fast. The revenue in some of those AI applications is kind of crazy. Can you tell the story of Magic School? I thought this was really interesting.
Yeah, so Magic School is like an AI application for educators. It's basically helping them use foundation models and LLMs to do their work, to do assignments for kids, to have an interactive AI experience.
It's like a full suite of AI for educators. The guy who created it was a teacher, right? The guy who created it was a teacher. He took some time during COVID to learn how to code, and he started using Replit.
Him and I think another person built the initial thing totally on Replit because you can go from an idea all the way to deployment, and it immediately started growing. You know, these AI apps, when the adoption starts happening, it goes super viral.
You don't need a ton of market kidding, and the revenue ramp was one of the craziest ones I've seen, especially for education. Yeah, it was like a known thing. It was like the hardest thing you could do: sell into schools, into teachers.
They're overworked, they're underpaid, they don't have the time to figure out your new tool, but this thing is great. So, if you go to it, it's basically like because teachers spend a lot of their time not in the classroom.
It's after school is done. They have to grade papers, they have to create the lesson plan for the next day, they have to create the quizzes or the multiple-choice tests, and they have to constantly do these.
There were these platforms like Teacher Pay Teachers where if I don't want to make it myself because I'm tired after the school day, I might be able to go buy one for nine bucks from another teacher who teaches fifth-grade science in some other state, and I would take that and I would buy it that way.
What Magic School did was it was like, "Cool, generate a—you just say like, 'I teach fifth-grade biology. I want to do a pop quiz about how mitosis works,' and then it'll basically create either a lesson plan or a quiz or a student interactive workbook that they need to create or whatever."
So, it lets a teacher not have to spend four hours a night creating the materials that they need just to teach class because AI can do it for them. This thing looks—I don't know these guys. I don't know anything about them, but it says over four million educators are using this.
Four million educators and their students, which I don't know if they're counting. Well, if you go on SimilarWeb, they have millions of monthly uniques, so that's like a pretty huge signal. They raised like $20 million too, so I mean that's like a pretty huge signal.
So, they launched in like, I want to say July 2023, so they are like a little over a year. Do you know these SaaS metrics are like how long to get to whatever, like $100 million or whatever?
The AI apps, and I would say Magic School is on that trajectory. The curve is like, you know, all the way straight up. This is kind of weird, but maybe this is like a feature of yours that you helped this company become potentially one of the faster-growing companies of all time, and you only earn $20 a month from that.
Yeah, so Replit had always a problem of value capture. Partly that's why VCs struggled with it for a long time. There's some logic for why it is hard to monetize these things and capture some of the value.
I will say, you know, I invested in Magic School, so there's some of that. With AI, I think we're going to be able to capture at least a little bit more of that value if people are monetizing these apps on Replit via the agent.
There's a way I think where we can potentially take a cut out of that, especially if we make it super simple to start monetizing an app. Also, once we reach scale, you know, it is like ChatGPT. You don't need a lot of skill to do that, and it's going to get easier and easier once we reach scale.
You have millions of people paying for this, and it's not just like $20. You're going to pay incremental after you finish your credits. We give you monthly credits, and then afterwards, if you want to continue, you can buy more credits.
Are there other companies like Magic School, like cool companies like that you've seen that maybe we haven't heard of that are using AI? Yes! You know, I'm very excited about agents right now.
I think I predicted earlier this year on a podcast that this is going to be the year where agents are born, and next year is where agents are going to scale. There's this company called 11x, and 11x creates AI SDRs.
Basically, you don't need to hire SDRs. There are some companies that feel like they can boost their sales without SDRs. You can have one AE, and then an account executive is running these tens of AI SDRs.
The revenue ramp on 11x was also crazy. It's pretty wild how fast these companies are scaling. I don't think in the history of Silicon Valley we've seen anything like that, even in the Web 2.0 era.
So, what is a fast ramp for AI? For maybe not 11x specifically, but just for an AI company, what's impressive that kind of broke the frame of how long things would take?
Yeah, so I would say reaching $10 million in 3 to 4 months ARR. Oh my God, that's wild! I invested in Jasper, which was like one of the early chat GPT wrapper type of companies where it was like, "Hey, like marketing, you need to write a blog post, you need to write a description for a product or whatever."
You could use it for writing any kind of marketing copy, and their graph was—I'd never seen it. It was like in 10 months or 11 months, they scaled like $50 million in annual recurring revenue.
It was like I've never seen anything even remotely close to that. It brought up a question: is this sustainable? What is happening here? It just doesn't compute, but it definitely broke my frame of what is possible because I've been working in Silicon Valley since 2011, 2012, and that just wasn't a thing.
You would never see a graph like that. What are some companies that got into that like $10 million or similar trajectory in 3-month type of businesses?
Yeah, so I wanted to kind of give a disclaimer about this, which is the big question in the investor community right now is the Moat question. That started around the time that ChatGPT kind of came out, and there were these GPT wrappers.
It was sort of this condescending way of looking at a lot of these companies. It's like, "Ah, if you can create a GPT wrapper in a month, then a lot of other people will create GPT wrappers in a month, and you're going to be competing on price, and the margins go down."
Yes, the ARR is great, but you're kind of a middleman, and you're going to have a hard time having margins. I think it's totally a valid question. Now, I think Moats develop over time through strategy and technical excellence.
Some of these companies can go down pretty fast, and there are examples of that right now. But I think if you can start building technical, like with Replit, again, this idea of pushing a boulder up a hill, we have this runtime environment, we have this infrastructure, we have the deployment, we have databases, we have all these integrations.
I mean, it's the only one in the world that is like an end-to-end environment to make software, and to catch up with that, it's going to take years. Technical advantage is also not a long-term moat, and so again, it's a big question.
I don't think it's answered yet. There are strategic things you could do if you reach scale. If the switching costs are high, that may be a way to have sustainable moats. But it is definitely a big question.
You know what's crazy, Sean? For the longest time, building a website or a web app, you just literally couldn't. Now you are making the technology that everyone can do it.
What I think is like guys like Shawn and me or people like us who have an audience, it's like why don't we just—why aren't we constantly launching companies using this technology? Because our ability to get users, because we just get on the microphone and talk about it, that's actually a competitive advantage.
Whereas being technical is no longer—it's still an advantage, but it's not as much as before. It's like getting customers now is actually the only hard part, which is still hard, but it's way easier if you're popular.
Yeah, so the playbook I would use is like I would go into some inefficient market industry. So, a deal from Magic School went into this hugely inefficient industry, which is schools and education.
By the way, another product is Synthesis Tutor, which is also going viral right now, and they have also this revenue ramp that's kind of crazy. Both Sean and I invested in that company.
Think all did, yeah. For a while, they had this thing where they had educators on the payroll and whatever. They replaced all of that with AI. Now, the kids sit in front of the iPad, and they're talking to the AI and learning really fast, and it's much better than the previous product.
So, basically, find an industry where you're familiar with and just build the GPT wrapper to automate some of the work there. You could do it like a hundred times, and one of them will take off.
Yeah, it's the era of the idea guy now. It's our turn! It's our turn to shine, right? Because now the limits and the kind of the value creation is: do you understand a problem well enough to know how to take this really powerful magic wand and point it at that problem and be able to make that more efficient?
Then, of course, do all of the other hard things: go get customers, make it sustainable, build a good team, you know, do all the normal entrepreneurship stuff. But it seems like more than ever, having a great idea is the key unlock to doing these things because building has become easier.
I'll give you kind of my personal epiphany that I had while I was doing this. I invested in Replit mostly when I just thought you seemed really smart, and I saw a growth curve of developers using it. I thought, "Oh cool, like a one-stop place where I can come in, write the code, host it all, all the stuff you talked about."
I don't have to download Java. I don't have to do any of that [expletive]. That appealed to me at the time. I think actually in the same way that Synthesis took AI and really like 10x the value prop of the business, I think you guys are doing the same.
So, here's my quick pitch, which is now that I think of Replit as basically what Shopify was for creating online stores, I think Replit is that for creating software. To me, you guys are—dude, eyes just brightened when you said Shopify for software!
So, I'll give you my example. I recently celebrated a milestone that was both I was proud of it and really embarrassed also. A few years ago, I started an e-commerce brand, and we just crossed $50 million in revenue—kind of like cumulative lifetime revenue.
Half of it was like, you know, this year, but $50 million total. I was like, "Wow, like $50 million, that's great! I had never created a business. I had done $50 million in revenue." So, that was like a personal pride point.
At the same time, I was telling it to a friend of mine who's not an entrepreneur. He's like, "Yeah, man, I would love to learn how to make websites and make products, manufacturing." I was like, "Oh, I don't know how to do any of that."
I was like, "This brand that is on $50 million in revenue for me, I don't—I just stacked Alibaba times Shopify. I've never manufactured a product in my life. I still don't know how to, and I've never made a website that's actually used by customers. I still don't know how to."
But I was able to skip all the work and get to the brand part. I did the thing where we created a product that people liked, and you know, it's a successful company now. I thought, "Wow, Replit's going to do that for the software space."
I was like, "It used to be that the job was software engineer, and now it's going to be software creator." It's like I can be a creator of software without being a programmer myself. That little shift is a big shift because the way I think about it, I don't know how many developers there are.
I think GitHub has like 100 million or 200 million accounts, so I'll just use that. Like, there's 200 million, let's say, developers, you know, software engineers in the world. Well, now there's going to be 2 billion people that can create software because if you got the internet, you got your phone, you can create software now.
You can just tell the agent, "Make me an app that does this. Make a tool that does this." You 10x the number of people that can create software in the same way that Shopify and Alibaba 10x or more the number of people who could create products and go sell them, like hard goods.
That's how I see what you're doing. Yeah, so even at the start of Replit, our initial seed deck kind of has this Elon Musk style master plan. It was like we build a platform, we grow it, and then AI is going to make the thing a lot more accessible because our mission was to make programming accessible.
Then we updated our mission to create a billion programmers. The moment that, you know, even GPT-3 came out, I was like, "This is the thing." I wrote this thread on Twitter about how AI agents will just change how programmers work.
This is the deck from 2015. I don't even know if OpenAI was a research lab at that time. Maybe it was, but definitely, there was no ChatGPT. But this is your master plan deck.
So, we're going to grow by building tools for teachers and students. We're going to build a simple network and AI-assisted interface that blurs the distinction between learning and building, evolve into a platform where people can learn, build, explore, and host applications.
You're talking about AI back in 2015 in your actual pitch deck. Dude, it's also clear how Codecademy was highly influential to you because I remember years ago Sean said everyone tries to learn how to code. I used Codecademy, and it was a pretty cool interface.
At some point, I kind of lost hope in courses because, you know, we have 100 days of code. We're telling users that to use our application, you need to invest 100 days. That's kind of crazy! There isn't any successful company in the world where you need 100 days to learn it.
That's when I kind of changed my mindset and said, "Okay, it needs to be ChatGPT-like. It needs to be just a prompt." We started building that earlier this year, and now that's all we're focused on.
We want to create new programmers. Existing developers, great, they have a lot of tools, but we want to go after the citizen developer. Everyone is a developer, and I think that's what you're talking about.
You go from like 100 million developers in the world—well, I think it overstates the number. It's probably more like 30 million, and then you 10x that. So, what does the world look like when anyone with an idea could make something?
One of my favorite books is "The Sovereign Individual." The thing I really was excited about is this idea of ideas becoming wealth. No longer have the bottleneck of making something—that's where we're headed.
This is what you're talking about, Sean. It's the time for the idea guy, and maybe that's tongue-in-cheek. Maybe the way to talk about it in more precise terms is that people who find these gaps in markets, people who have expertise in certain areas, can tell that there's inefficiency, and they can create an AI application that can immediately plug that.
I saw this video on Twitter the other day. It was of a snake that got its head chopped off, and it floated around and bit the tail of its own body. Are your employees thinking that they're sort of doing that to themselves?
Are they like, "When you make jokes or when you talk, you know, like you don't need to hire all these programmers to do all this stuff?" Are they like sitting there with their hands in their pockets like, "Uh, does that mean us?"
I always wanted the company to be super lean, and for a long time, we were like 10 people. But now we're like 70 people. That's still nothing!
I'd rather not hire a lot more people because I think, again, the efficiency for programmers. Look, citizen developers are going to go from zero to like say 10x, but also existing software engineers are going to go from 10x to 100x, right?
They're going to become more and more productive the moment we automate all of software engineering. I think that's sort of like the moment of AI. I think it's a little far away. The reason I say this is because once you automate software, then the agents can rebuild themselves.
You go into this loop of increasing intelligence. Every version builds its next version, builds its next version, and so this is what they call intelligence explosion. That would lead to the singularity, right?
It's a pretty crazy time when we automate all of software engineering. I think it's coming. I don't know if it's 10 years or 15 years, but I think that's the time where the world really radically changes.
Have you met anybody in the tech industry that blew you away, either personally or maybe read about them? Maybe met a friend of a friend told you a story?
I saw a picture of you with Jensen. You know, you've met Paul Graham. I know that you're connected in the AI circles. You met Sam Altman. In addition to building the tech, I love the characters, and I love the stories.
That's why every Elon snippet of how he runs his companies goes viral and [expletive] like that. What are your favorite kind of inspiring stories or crazy stories that you've either experienced directly or read?
Yeah, one of the curious stories when we were raising from a16z, Mark invites me to breakfast at like 10:00 a.m. at his house. So, I go there, and I expect like I'm going to talk about the business.
We spend like two or three hours talking about politics and the world and all sorts of things that are interesting to him. I felt like this guy is more than just a technologist. He's like a philosopher.
Right now, he's going out, and he's talking about this stuff. His Joe Rogan interview went super viral, and he's been always having these interesting ideas about the world.
The interesting thing about a16z is his partner Ben is sort of the executor, sort of the executive. He wrote "The Hard Thing About Hard Things," where he teaches you what it means to run a company. It's painful, it's hard, and what it means to hire executives and what it means to scale a company.
So, you have this duo of the doer and the philosopher, and I think that's really amazing. I think they have really big plans. I would just hate the philosopher to be like, "Are you going to do anything? What are you talking about politics for right now?"
It's got to be the worst to be the doer in the doer-philosopher relationship, right? You know, I think Sam was interesting to kind of meet and talk to because he's very effective.
The first time I met him, or maybe not the first time, but he was on his computer as I'm talking. I'm talking, and I was like, "Yeah, we're fundraising. I went to talk to a16z. I'm a really big fan of Mark."
He was typing on his computer. "Okay, I introduce you to Mark." When you send Sam emails, he replies pretty quickly with a couple of words, a couple of sentences.
I saw how effective and fast you can be, and that really—I'm not like that. I'm trying to be more like that, but I'm someone who really values the quietness to think about ideas and to think about strategy and things like that.
So, I'm not always on top of communication. It actually makes me a little overwhelmed. But I think seeing these people at least inspired me to be a little more like that.
You tweeted out the story that I loved about—you said the most gangster story in Silicon Valley is Steve Jobs buying Pixar for $5 million, investing $50 million, and operating at a loss for a decade.
So much so that he had to cut personal checks to make payroll and somehow turning it around to a $7 billion exit. Why did you like that story?
You know, there are people who are overrated in Silicon Valley, and I think there are people who are underrated. I think people think about Steve Jobs in terms of the flashy things: the iPhone, iPod, you know, coming on stage and doing that.
The thing I like about the Steve Jobs story is when he was lost in the desert for 10 years. He left, he was fired from Apple, and then he created two companies that were failing the whole time—like NeXT Computing and Pixar.
They were literally failing. They didn't do anything. They weren't selling. He was just investing more and more of his money. I think he was going to go broke, but he kept going for 10 years.
How do you do that? I'm a person who, like we talked about in my story, I want to be able to go the distance. I think going the distance is an advantage for entrepreneurs.
Pixar became this hugely valuable company. It goes from making no revenue to making billions of dollars and going public over a couple of years. NeXT Computing saved Apple. Apple was having a problem with OS.
Intel, you know, they had the chip before. I don't know if they made it internally or something like that, and then everyone was moving to Intel, and Intel was the best computing chip. They wanted their computers to be fast, and they needed a new operating system.
They tried to buy—they went to the market, they tried to acquire a company, they couldn't find a great operating system, and NeXT Computing had a great operating system. That became Mac OS.
So, they bought—I didn't know that! I thought NeXT was just a failure. I didn't even realize it actually contributed. I thought they just bought Steve back, Aqua hire, but it wasn't an aqua hire.
No, I mean, Objective-C, for example, you know, NeXT Computing was really obsessed with this idea of object-oriented programming, and they innovated a lot on what that means.
It is based on Unix, but it has a lot of interesting features on top of that. So, Apple saved Apple because Apple was otherwise not going to be competitive without the new chips, right?
Well, dude, I know we kept you half an hour over. I apologize for that, but this was amazing. This was one of my favorite episodes in a long time, and I'm not just saying that. You can go check all the other episodes. I don't say that at the end, so this was awesome.
Thanks so much for coming on. Where should people—Twitter is the best place to follow you? Yeah, Twitter, @omj on Twitter, and the Replit handle on Twitter as well, just @replit.
Dude, thank you very much! You're the best! Of course, my pleasure!