How Vercel Saves Dev Teams $10M+ in Productivity Gains | Guillermo Rauch

It is quite easy to make money in Silicon Valley. You're just going to naturally see that people around you are building the future. Forget about like OpenAI, whatever. There's probably more AI unicorns being born today.

This week on the show, we have Guillermo Rauch, the visionary founder behind Vercel, the platform fueling the next generation of AI-powered, lightning-fast web apps.

My idea of building software is that you type something in and you press deploy and it's live instantly. Like literally no build time, no CI/CD pipelines. It's instant.

If you've used the web lately, chances are you've touched his work from websites like Nike, Adobe, and OpenAI.

Elon just talked about this, I think a couple of days ago. CEO needs to be an engineer. The CEO needs to be a developer. The CEO needs to understand the problem space very, very deeply. This truly is the world that AI should unlock in the future, is that we all become more full stack. What is your level of AI proficiency?

He scaled from coding in his bedroom in Argentina to building one of the fastest-growing cloud companies on the planet.

You have to chase your interests. You have to pursue your interests. For me, the obsession was figuring out how things work. My dad was like, can you see how far you can push this thing? Can you hack around it, change the operating system? So we always had a hacker mindset around having a computer at home. What is a modern version of that that I can give my kids?

So what about screen time of the kids then? Like, what do you think?

I would meet a lot of like tech CEOs and families and whatever, and like they all had pretty strict technology limits. And I was like, that's, that's kind of weird, guys. I'm here because I had no technology limits.

Yeah.

Yeah.

That's exactly how it feels like. Imagine if my parents had had a screen time for me.

I'm Enrique Dubogras and welcome to HD in HD. This episode is brought to you by Brex, a brand I'm proud to have co-founded and one that's shaped by the same journey many of you are on. Brex has everything startups and fast-growing companies need to make every dollar count from modern corporate cards, banking, and treasury to accounting automation, travel, and expenses. Over 25,000 companies, including DoorDash, Scale AI, and Anthropic spend smarter using Brex.

Man, thank you so much for coming and thanks for being here. I really appreciate you taking the time.

Yeah, excited for this.

So I was remembering the first time we met, one of our common investors.

Was it COVID?

I think it was COVID. Yeah, I think it was COVID. One of our common investors introduced us and said, Enrique, you should meet these guys. These are building amazing dev tools. And they're Argentinian. And as soon as he said Argentina, I'm like, do I really want to meet this guy?

You know, like, I don't know. Brazil-Argentine rivalry.

Yeah, exactly. The Brazil-Argentina rivalry. Even though I'm in a board of an Argentinian company. Oh yeah. Yeah.

So but that Argentinian company, MercadoLibre. Yeah. So their biggest, their biggest market is Brazil, right?

Their biggest market is in Brazil. But I think they do really well in Argentina, you know, really well.

Yeah, of course. And I think for Argentina entrepreneurs, it's like How the hell did that even happen? Like so ahead of their time, early to the internet. So a huge inspiration for all of us.

Totally, totally. And I feel there's like so many other amazing entrepreneurs coming out of Argentina, right? Like, you know, I think Wences was Argentinian, right? And then, you know, Auth0. Auth0, that's right. That's right. The Walla founder too. Like, you know, there's a lot of great—

A lot of movement. I think the next decade, the best is yet to come.

Yeah, for sure.

For sure. And obviously Vercel, right? Like the crown jewel of it. But yeah, man, like, so first, maybe tell me a little bit, like, you're born there. Would you grow up there? You grew up here in the US? Like, how is your life?

Yeah, fun story. I grew up in Argentina until I was at the last year of high school. And so I was like, I was 18. Buenos Aires? Yeah. First in the outskirts of Buenos Aires in a town called Lanús, and then moved to the city to go to high school.

American high school, normal high school?

Normal. Not normal in the sense of— so normal by the standards of like, you know, it's another high high school in, in the city, in Spanish-speaking everything. But it had one really cool wrinkle, which is they called it an experimental high school where teachers from the University of Buenos Aires would be trained, and some of them were actual university teachers, and they would come teach more advanced courses and materials to high school students. So kind of like experimental in the sense of like, what can we— how far can we push these kids, like what can we teach them? What new subjects can we introduce? And then try to like take those lessons and make them more broadly available to the rest of the, you know, school system. So I had a lot of upside in that. It was free for my parents and they taught me a lot of really valuable things about so many different things like chemistry, physics. Interestingly enough, it was an accounting-focused school, so like accounting, math. But I did eventually drop out because I was simultaneously going to high school like any normal kid. I was developing my sort of portfolio and online resume of open source projects.

So how do you start coding in the first place?

Yeah, so I started coding when my dad got a computer for our house. I think I was like 7 years old when we first got the computer, but it kind of came with a prompt, kind of came with a quote unquote system prompt in AI lingo. My dad was like, can you and my, you guys, my brother and I, can you see how far you can push this thing? Like, can you install software that is like harder to install? Can you hack around it? Can you, you know, change the operating system? So we always had a hacker mindset around having a computer at home. It wasn't just a computer where like, you know, you play video games. My dad would ask like, can you write your own video games? My dad himself, his skill is that He didn't know how to do any of these things himself, but he had the right prompts and questions for my brother and I. So we kind of were pointed in this direction of like, a computer is a dream machine. You can do anything with it. Don't take for granted what it comes with, what software it comes with. You can modify it. That was kind of like the—

Where do you think he got that? That was pretty like—

It was really visionary for sure. So interestingly, so both my parents went to the University of Buenos Aires. My dad was an industrial engineer. My mom was a chemistry, a chemical engineer. And my dad had the vision that software was gonna change the world to the point where he was a little bit even self-deprecating about his degree. He was like, ah, industrial engineering, that's cool. But this software thing is like way cooler. Wow.

That's amazing.

So his primary job was the fruit and vegetables exchange, like the main market that oversees all of the other markets in the US, and so in Argentina. And he would have ideas like, let's install the first Wi-Fi network in Argentina. And he would try to convince like, which is a government-based organization, to like, can we get budget to do this? And so in my early years, he would like always be looking out for like what's next and then see if I could figure it out. My brother and I, so like he would give us the prompt like, can you figure out Wi-Fi before like anybody has Wi-Fi? Can you figure out, we're probably one of the first households in this little you know, not so great town in Argentina that had DSL. I had to figure out how to make DSL work. And at the time, I think one of the big breakthroughs for me growing up was my dad got us a magazine that taught Linux. It was all about, hey, this Linux thing is coming. And it came with a CD with a free version, of course, of Linux to install on your computer. My dad was like, if you figure out Linux, I know what his like thing that he set up as an award was like, if you get Linux, I'll buy you this thing or whatever. Like, so installing Linux and taking over the machine and try to do anything with it was a big thing in my childhood.

It's amazing, man. Wow. Your dad was like really a visionary and pushing you guys to do it. There's like a lot of lessons there.

Yeah.

Do you have kids or no?

I have 5 kids.

You have 5 kids?

Yeah.

Okay. Are you doing some of those things with them?

You know what's funny? I find myself just copying him, to be honest. Like, I don't know if there's more to— like, I'm trying to find what are the, like, other things that I could do because I did— I do think it was a really good blueprint. There's a few things that I've learned. Like, number one is there has to be an innate interest to become a hacker. I don't think you turn people into hackers. So I think there was something in me about trying to overcome the limitations of the software, the hardware, breaking things into their elements, like their constituent elements, and like rebuilding the computer and all these things. It was like very natural to me, kind of like, like math brain and things like that. But I find myself giving my kids some of the same prompts of like, I look at them using an iPad and I tell them like, can you figure out, you know, like how an iPad app could be built? I ask them questions to see if they can try to reverse engineer. How systems work. Reverse engineering and hacking, I do think were really big breakthroughs for me in my early, in my early days. So one that didn't come from my dad, but that was, I think, a transcendental moment for me was I found a hacker competition online, which was made of like 20 different steps. Step number 1 was it gives you an HTML form and you have to hack it. It had username password. And you have to pass the form. It was really easy because you could read the HTML source and the password was literally in an HTML comment. So it was like a hacker exercise that almost anyone in the world can figure out. But the— it kept growing, growing and getting to like level 20 and it gets really hard. Like at one point you had to download a Linux binary and debug it, step, uh, set up, you know, breakpoints and step by step, look at the memory state, maybe some encrypted memory and try to like overcome that binary and then get to the next level of the website. Those kinds of like riddles and puzzles and things like that, incredibly valuable. And, uh, so those are the kinds of things I'm thinking like, what is a modern version of that that I can give my kids?

Cool. Yeah, no, I had a similar experience. Like when I was growing up, I was hacking Ragnarok servers is the game I used to play. And, you know, I remember being this program, I think, I think it's called Wireshark. Was that a thing? Yes.

Yeah, it's still a thing. Yeah.

Yeah. When you're like kind of like monitoring the packets coming out and like, that's what I, you know, I kind of reverse engineer because I didn't have the source of the clients and it's still super useful, by the way, like packet capture and looking at your network flows. Yeah.

Like I just last week, these are the skills that you keep forever. Totally. Just last week I downloaded this really slick AI app and I was trying to figure out how it worked. So lo and behold, you know, packet capture and like you understand like, oh, it's using this service. Oh, it's using this piece of cloud infrastructure. So those are the, like understanding how the world works, I think in a nutshell is one of the most important skills that I think for adults, kids, everybody, you know, if you can figure that out.

Totally, totally. So what about screen time of the kids then? Like what do you think? Is it?

No, I'm not a screen time person.

Meaning you don't regulate it?

I have like the basics in place. Like at night you have to start winding down and just protecting your eyesight and things like that. Like doing the things that, you know, keep kids and adults for that matter healthy and, you know, well-adjusted people. But generally speaking, I think you have to chase your interests. You have to pursue your interests. For me, The obsession was again, like figuring out how things work and tearing them apart. And it was such a joy that I would get from overcoming all those challenges. Like I remember the day that I finally made dual boot work. So because we only had one computer at the house, I had to figure out how it could boot into Windows so that the adults in the household could still get value from this really expensive thing that they saved for years to acquire. So they could run the word processor and Excel and whatever. And also I could do my own things in Linux. So figure out dual boot, like you reboot the computer and it asks you what operating system you want to launch. The amount of joy that I would get from, okay, I installed Linux and I figured out how to boot both operating systems in one computer and the amount of trial and error that it would take on some level, it became like a really healthy game for me. Where it's exciting, it can be frustrating. Like sometimes you're like crying and banging your head on the keyboard, but then you overcome it and you feel joy. So long story short, you know, if my kids can find those kinds of healthy feedback loops, I love screens for them. If on the other hand, you know, there's very unhealthy feedback loops today that I think we didn't have growing up. TikTok and— TikTok is like the most poisonous one I could imagine, but Even without getting there, like games for kids that are more like little casinos, like dopamine-inducing behavior, right? That can try to try to like con your parents into buying this like upgrade and whatever. Like, that's not good because we didn't have that. And now, like, potentially aging ourselves and like, oh, the old times were so much better than the current times and whatever, melancholy. But I do think that there was something about video games that they were harder. They were durable. They didn't have all this like in-app A/B test driven monetization.

It was like one ship, right? Like this is the game, right? Like there's no feedback cycle improvement to improve the game.

And I feel like the authors of those games were this kind of like perennial hacker that cares about the quality and the craft. And I remember there were video games that, you know, had really interesting riddles within them, or they had the advent of cheat codes. That you had to figure out how to hack the game. And so those things were coming from the early game makers that had that, you know, passion for, for building and passion for engineering more so than the new game studios that are like the passion for like optimizing every element of the game for in-app purchases and whatever. And so you have to be, I think, more towards moderation for those. I also don't reject them. I think some of them are really fun. And I also remember my days, right? Like I wasn't this machine that was just, you know, overcoming the hacking game and I would write Linux tutorials online. Gaming a lot, right? I would do open source. So I would have phases. Phases of like deep, just like using the computer for like the stupidest stuff, right? And video games and stuff like that.

What was your favorite game?

Counter-Strike, I had a big phase. Age of Empires, I had a big phase. I mean, just so many. I was kind of a big gamer growing up. And then when the internet sort of came about, like, I had a whole, like, there was a whole generation of like Flash and in-browser games. Yeah.

Yeah. And where you had like all these forums that you were part of online too growing up.

Forums were huge.

Yeah, I think it too. I don't think there was anything like Reddit maybe? Like, what's the—

Yeah, I think the closest thing today would be Reddit. It would be video game communities still. I think like, you know, the kind of like TeamSpeak type stuff of like you find your niche in the world, like you find your game like Fortnite and Discord communities. And then that leads to Discord communities and, you know, just spending a lot of time with your friends playing online. I think it's similar to that. And by the way, I have seen a successful transition of that people that grew up with those to awesome entrepreneurs, hackers, etc. Because depending on what game you're playing, like some tend to be more oriented towards this like mindset of like, can I extend it? Can I mod it? One of my kids is now super into Minecraft and I'm really happy about that because again, it's a video game, but it's also building your own world. It's a creative one. It's a creative one. There's a lot of programmatic elements to it. I've actually been meaning— this is a good reminder that I should do it. I remember someone built a— I think it was a mod of Minecraft to explain quantum effects. So you would go into a room and it would do entanglement. You would observe the state of a block and that would be entangled with another block. You could do superposition. You could do multiverse stuff. Like you could go into a room, come back, go into the same room. It's like it's in a different state. So I want to introduce my oldest to quantum computing through this Minecraft mod. And so Minecraft as a platform, I think, is super healthy. And by the way, it goes back to like the creator's sort of mindset and DNA gets embedded into the game. So Notch, the creator of Minecraft, is that archetype of hacker who— he was the CEO of Minecraft or the company behind it. And he was coding it. He was doing livestreams of coding. I don't know if you knew this, even before livestreams of coding were cool, he was super early and he would basically like livecast how he was hacking on Minecraft, the game itself, which I think is primarily written in Java, which itself was quite a puzzling and interesting choice for a video game.

Yeah, I remember that growing up there's all these like flame wars across programming languages. There's still some of it, but like—

There's still quite a bit. I think back then it was even more. Yeah. And I think this is one of the things, what is AI changing about the world? I mean, it's changing everything. But to your point, I do think there's more of a— maybe the attention window has shifted more towards the LLM than the programming language.

It's like now there's just a way to communicate what you want, right? Yeah. There's like a bunch of them, just pick one and go.

And it's whatever the agent picks more so than what they potentially, right? Like than what the human picks. Yeah, because if I tell you, hey, here's this AI that is— it'll make any, any of your dreams come true. At the end of the day, we're all working on Turing-complete programming languages, Turing machines, ones and zeros. You know, like what language you used to get there will not matter that much to you if you can get these incredible results.

Totally, totally. And I think there was also like a huge hardware component also, right? That like Before there was these things like these languages were like optimized for different use cases because the hardware relationships are different. Now the hardware is so powerful that like, you know.

I think Minecraft being powered by Java is a really good example of moving from that world where like you have the Carmacks of the world who are sweating every cycle. And then you have the emergence of like higher level, more managed memory programming languages like Java that can still be quite fast and performant.. And the fact that Minecraft was able to be bootstrapped in that context is super cool. And also I think it accidentally helped the decompilation scene. So there's entire dumps of like Minecraft source that are not from Minecraft's creators. There's like really high quality decompilations that gave emergence to the modding scene.

Right.

And so like you can create your own. So this Quantum Minecraft that I was just talking about is like, They didn't create like an API Minecraft SDK platform, although they should, by the way. And I think games of the future will. It became an extensible gaming ecosystem. So I didn't experience that exactly with playing Counter-Strike. But the closest thing to that was we had no money, zero. And so we had to like hack the games, like extend their shareware licenses and whatever. So like, the hacking and modding stuff was starting to sort of emerge and it does teach you very valuable things about software engineering.

Oh, totally. That's, you know, for me, that was exactly how I started this game Ragnarok. It was like an MMORPG that like there's a centralized server and then the client. So you could kind of decompile the client, but the server was the server. So that's why we were like on Wireshark looking at the packets and seeing how to like rebuild the server without having the access to the source code. And then one day the source code leaked and everyone was like so happy, you know, like you could get everything.

And then you realize that like, when you're on the other side, because today running Vercel, we're wanting to protect the internet. We're wanting to make it more secure. We want to make sure that the applications our customers create are secure by default. You're now constantly thinking adversarially. There's some clever, whole range of clever teenager all the way to nation state actor that we have to protect against.

With AI access now, right?

That even more. It's a war of people decompiling on one side and you know, looking at the corner cases of your protocols. And then the other side, you need this extreme paranoia because I think there was a golden era where you could assume that no client could misbehave, meaning that, you know, oh yeah, like I'm building a game. What are the odds that someone finds that there's a special packet that you can send that makes you win the game? And we didn't protect against that. But here you have Enrique, like growing up, figures out, oh, 'Oh, there's this special packet that makes me the winner.' And no one thought on the other side to check if the rules of the game had been honored. But this is the kind of thinking that if you can create that spark within your kids of, 'Oh, look, the world is following all of these instructions and you can understand it. And here's a little gap that you can overcome.' it can feel so rewarding to like figure out those hacks, like escaping the simulation, right?

Totally, totally, totally. Yeah, no, this is something my wife and I discuss all the time because there's all these things we did growing up and we're like, oh, would we, you know, we don't have kids yet, but would we let our kids kind of do that? You know, like, yeah.

I can imagine for you growing up, I was contrarian on this because I think in San Francisco, I would meet a lot of like tech CEOs and families and whatever. And like, they all had pretty strict technology limits. Yeah. And I was like, that's, that's kind of weird, guys. Like, I'm here because I had no technology limits.

Yeah. Yeah. That's exactly how it feels. Like, imagine if my parents had a screen time for me, you know, like, yeah. And same, same for you.

You're rolling the dice a bit. Like, I don't think my dad had all of like, there's also a level of accidental success too, right? Like my dad didn't have all that narrative about like what computers could do to the future of humanity. Like, and also his bias was Star Trek. In fact, like, I think his inspiration was even beyond what technology makes possible today.

Do you have siblings or no?

One brother, funny enough, he's celebrating today. He was the first employee at Scale AI. Oh, wow. Scale AI just got acquired by Meta. So my brother and I were like the two hackers of the house growing up. And then I have a sister.

What's his name, your brother's name? Ricky. Ricky, because you know, Scale AI was our first customer at Brex.

Oh, that's so cool. Yeah, I think he actually told me. So I always give credit to Ricky. He, he has such a good, like, market map of what's going on in Silicon Valley and cool tech companies. He, I think he was the first person to bring up Brex to me.

Yeah, no, they were the first customer. Alex, you know, Alex is a close friend. And, you know, because that day I remember going to look at cars at the office.

Oh, that's so cool.

Yeah, no, that's so— but you're okay. So whatever your dad did worked because you worked twice, right? Like That's so funny. That's a great story.

The timing of the conversation is so funny because like literally happened just now.

That's awesome, man. So what was the first time you felt you could now monetize this? Like when is your first time you made like your first money?

Earliest, like absolute earliest, but didn't have to do so much with coding, but it was the pre-coding era, was I would burn CDs for my friends in school. Like music CDs. So for 5, I don't know, not even 5 pesos, probably like a couple pesos, I would go to the school playground and tell you, you want music? Here's, I can sell you a CD with music. And that's because our computer at home had the ability to like, it was a read-write CD appliance. But that was kind of like, again, the mindset of like, before I got to code, I got to do really cool things that extended the capabilities of the computer. One really cool thing was like installing peripherals, right? So like adding the CD burner component to the hardware was like a really cool thing. Adding the video graphic accelerator cards. I was recently listening to a Jensen Huang interview and he was like, you know, retelling the old days of building the accelerator graphic cards.

NVIDIA GeForce.

That's right. GeForce. Yeah. A competitor of them, which I think was called Voodoo. We had those because we were always trying to add things to the computer. And in the process, we broke so many computers. I totally know that. Rest in peace. Just like figure out the voltage and all these things. So first time I made money was like, okay, we extended the computer to do this really cool thing, which is like burn our own CDs and then sell them in the school playground. I made quite a few pesos, but I quickly saturated the market. I'd sold several CDs to all of my friends and like, I also got bored with it. The next big one was, uh, and this is actually when I'm starting to code, one of the first things that I did was create websites of my own interests, which is another thing that I would highly encourage people to do with their kids. Like, uh, use v0 to like prompt something and show your kids what it's like to go from idea to a working website, your own little space on the internet. I did that in the early days with, uh, hosting, the earliest hosting providers of the web, a company called GeoCities. There was another one called Lycos, I think, something like that. You know, so much so that these are so old that I'd never known how to even pronounce them in English. Like I'm trying to pronounce them in Spanish in my head. So this would allow you to like have a little bit of hosting for free on the internet and they would give you a really clunky URL. So it would be like geocities.net slash, uh, uh, you know, some weird character, and then you could claim a path name. And inside of there, you could host your own HTML files. And so I started building websites around my interests, which were what a 10-year-old could be interested in: Dragon Ball, other, other kinds of anime, and, uh, not even soccer. Like, at this point, like, my nerd interests are starting to overcome the traditional Argentinian interests like soccer and football.

Yeah.

And so I learned from a forum that you could put banners like advertisements of MercadoLibre on the website and join their referral program.

Like an affiliate program.

That's right. And so if I had— if I was able to send someone from my website to MercadoLibre and they would sign up, that would be 5 pesos. I'm pretty sure, you know, like if someone is listening to this, which is a very small group of people that could have been around at that time, and I was blown away that it actually worked. So I think at this point I'm 11 years old and this is my full-time obsession. My mom helped me cash my first check, which ended up being like hundreds of pesos. And we couldn't believe that you could go from something on the internet to actual money until the very last second, until we were able to cash that check. Like, we're still in disbelief that this was even possible. And of course, my mom was pretty shocked.

How old are you then?

I think it was 11 years old at this point. And that didn't last because I then— I think that being a little kid is very hard at this point to like maintain attention on anything and like do something for like a long enough period of time. So I quickly moved on to other things. I think the next big step for me was Linux and open source that just like, again, took over my, my, my whole life and on some levels.

And were you good at school?

So this is where it gets really interesting. So at this point, I was regarded in our little, you know, little neighborhood, like Catholic church school, I was regarded as like the top student. I had won the, like, the little Math Olympia that they organized and whatever. So my parents were really excited for my brother and I to go do this better high school that required an entry exam, and it required studying really hard. Uh, even if you were skilled in your elementary school, the rule was you're gonna have to even need supplemental teaching, kind of like a little bit like the obsession over SAT scores here. There's a micro obsession in Argentina. I don't know if still lives on, hopefully it does, over getting into these two schools. One is around math and accounting. The other one is around literature. High schools. In Spanish. Yeah. Yeah. High schools with like this thematic focuses.

And private or public high schools? Public. Public high schools. Yeah. So like they're just like the most competitive kind of public high school.

Most competitive entry exam. And you have to study a full year. Uh, and you have to take multiple exams across four domains, math, uh, language, uh, or Spanish, uh, history and geography, at least at the time that I did it.

So you got into this high school.

So at that time I got in among the top 10 students. Oh wow. And I worked extremely hard and everyone was like, oh, he's the cracked guy that like, you know, is always competing for the top spots and whatever. Yeah. But at the same time that year I decided to take a complete break from doing my online stuff. Which at that time I was already starting to make money from the internet selling my own skills. I had discovered because of the forum relationships, I discovered a website called ScriptLance that allowed me to take little gig jobs with my own skills like CSS, HTML, PHP, etc. I was able to make some money online, so much so that at this point I'm making money in US dollars. And that's a big deal in Argentina.

Oh yeah, 100%. Same in Brazil.

So I was actually financially contributing to my household. These are very hard times. The 2000s in Argentina, financial crisis.

It's been hard times in Argentina for a while.

Very hard times. My dad was always extremely stressed. To get something nice for the house like a computer would bring our savings to zero. And, and so sometimes in Argentina you grow up with this habit of like, some years are good, some years are horrible, at least for our household. Oh, interesting. Like there's no financial stability. There's no like Ford, you know.

But how does it manifest? Like what, what is it like that, like the currency deflates, the inflation goes too high? Like how did it like feel a good year versus a bad year?

Hyperinflation, periods of brief economic optimism and changes of precedence that yield maybe like there's like 3 good quarters for, uh, for what my dad was doing at the time, followed by 3 quarters of like, there's no work and the budget is cut and whatever. So it felt very unstable looking back. And so it would translate into like, okay, this is the year that we're able to get a computer for the house, followed by the year that we can barely make ends meet, you know? So I, in one of those years, I helped quite a bit. I helped my parents quite a bit because I was actually making quite, quite a bit online. I started to learn how to get better jobs, bigger jobs. In these websites, you build up your reputation.

So these are like freelancing websites.

Yeah. So they had a rating system. And so I had a perfect— I remember being so proud of this. I had a perfect score of 5-star jobs. Like I had like 20 in a row.

What kind of job? We're just coding. Like website.

So I actually remember vividly I had that moment, like when I was, you know, 10 or 11, I had that check. Another moment that I remember very vividly is when I made— I think it was like $10 or $20 for fixing a styling bug for a client.

Oh, wow.

So they had a problem with their web, which for them was a huge deal. It's like, if you solve this for us, this, this job is worth like $15. And I did it. I remember thinking like, wow, that's so easy for me.

It took you like a day, like less than a day.

It took me like an afternoon. Correct. And then the, the balance goes up. And I remember like I was just like lapping around the house, like thinking like, this is the best day of my life. It's amazing. And this comes off many years of me doing free work for the internet. So the guy that recommended that I join this website was like, you're creating all this content online and you're helping out all these people in forums. Imagine if instead of, you know, just only helping for free, you're able to help other people that are saying like, hey, I have this problem, solve it.

Did you make a lot of online friends?

Yeah, quite a few.

I was like that too. I made a lot. I think it's not super common.

Hopefully it is, you know, in the sense of like, I don't know how safe the internet is for my kids these days because there's a lot going on. But I found communities that were like so specific too, right? So specific and like people cared very deeply about each other. I remember actually another big breakthrough for me was making sure that configuring DSL and the faster internet we got for the house, configuring it on Linux and compiling the kernel module. Because like the joke about Linux is that Linux never succeeded on the desktop because you need a PhD to use it, right? And like it's always hard, it's always breaking. That was even more true back then. I had to compile a kernel module to get the internet going.

Oh yeah, there's so many distros of it. It was like Ubuntu was the one that like everyone liked about like—

How do you solve this impossible like catch-22 of like, if you can't use the internet to get help in documentation and download the tarball, this is what the, you know, the zip file of source code of this module is called. How do you, yeah. So my life was boot into Windows, get internet access, learn, learn, learn, write things down, reboot into Linux, and try to make it work.

Yeah.

Oh my God. But I'll tell you one breakthrough here. Yeah. So this is like by the act of God. The one— the magazine that we bought had come with a bundle of software which contained this very specific software package that would allow me to actually get the internet working. And by everything that I know at Linux these days, it's actually kind of a miracle that it works because the likelihood that the CD software was going to be compatible with the version of Linux that I had and get the whole compiler toolchain to work was actually pretty low. But the other thing that happened is I befriended someone in that IRC server, that little community, that forum that called me on my landline and helped me like debug the compiler output. So some random internet stranger was so helpful that he literally called me on my landline to, to, uh, to help me get unstuck. So I think those friendships that I made online and the, the inspirations that I made, because there were people that knew so much more than I, and I looked up to them. Uh, so being terminally online, as we say these days, was actually one of the big enablers for me.

Interesting, interesting. Yeah, I know, it's crazy like how this internet growing up shapes you. Yeah. It's like, it's really a thing. Okay, so now you went to a super competitive high school and then where's college happen?

So then I became the worst student.

Oh, you became the worst student.

Like literally I went from the best to the worst because what happened is this freelancing website eventually gave me more and more amazing jobs. So two things happened. One is like, I felt like I was at the top of this website and like I was actually picking my jobs. Whenever I would bid, people would take it very seriously compared to like the other, like I could outbid people, meaning like I would actually bid higher and still get the jobs. I was helping my parents with dollars coming into the household. And then at the same time, I was developing a very strong interest for open source, contributing to open source projects and kind of starting to develop a reputation for it. And so those two things combined were disastrous for my very demanding high school. Kudos to that high school. Like, they were like hammering us, giving us very advanced content. And like, uh, you know, like the teachers were kind of unforgiving. Like, I still have one scar of a teacher that, uh, my chemistry teacher, that she thought very highly of me and she like really like was pushing me. And I remember an exam where I got 7 out of 10, and she crossed in red the whole exam. Like, I was like, she was so harsh with me. I would think, you know, like back in my little town, the Catholic school was like, they would not even study chemistry ever. So the school had those awesome aspects to it, but then eventually I feel like I let a lot of the teachers down because like my mind was just online all the time. So in some ways it was actually quite hard. And this was, at this point, my parents are completely disengaged in a sense of like, I, it like the whole thing took like, I was building a career of my own independent of the entire world at this point.

And they trusted already that if you were making money, you're gonna be able to sustain yourself so you can make your own decisions.

That's right. There was almost like they were building this trust kind of in a little bit of being mesmerized by everything that was happening. Like, it was like, why you took this job? I remember I had a client in Boston, I had a client in the Netherlands and I would just give them the updates. So what was probably hard maybe for my mom was that we'd work so hard as a family to get me into that school. And then it was just like, like the other problem was time zones. Like I would work with my European clients and like, and you know, like engineers and hackers are like also naturally night owls. So like I would stay up late and like, but it was kind of like, I'm very thankful that I was given that freedom because like it was, it was, it was also a great time for me.

That's awesome. That's awesome. Okay, so you became the worst student and then did you want to go to college at all or did you even go? Like, what did you do?

You know, I had really big aspirations for what I could have studied. I always tell people that there was a university on my radar that, that met the same profile of only a few students get in and it's like really cool. It was around nuclear engineering in the south of Argentina, in Bariloche, uh, Instituto Balseiro. Have you heard of it? Balseiro?

No, I've heard Bariloche, people go ski.

Yeah, exactly. But it's also known super randomly for having— so the nuclear program for nuclear energy in Argentina was started out in Bariloche, of all places, and they started a university there, highly specialized in nuclear engineering. And so I knew about it. I had a very strong interest for that, but it was competing for my interest in open source and all of these things. And so I ended up getting a very, very, very compelling job offer. So there was this startup based out of Switzerland who was very interested in the JavaScript framework that I was working on at the time. It was called MooTools. It was a competitor to jQuery, another popular JavaScript library.

I use a ton of jQuery. I remember the first kind of like async, you know, like management.

Sadly, we got beaten to jQuery to the like broad world popularity, but we won in very specific niches. So one specific niche is that Facebook really liked MooTools. The Facebook, the early Facebook engineers that they hired for their user interfaces group, a lot of them came from this group. They were all my friends that we were, again, these online friends. Uh, my, one of my best online friends went on to become the tech lead of React, which is the most popular UI library in the world today.

So now that you're kind of like looking at this first thing, jQuery versus MooTools, and you had to like analyze why jQuery won, like what would be your analysis?

Great question. Um, I think simplicity over purity.

Tell me more.

Every single time. Pragmatic results, pragmatic engineering, not getting too caught up on the beauty of the abstractions. So the world needs to make progress. You have to ship products, you have to find solutions. jQuery figured out how to do that better than us. They focused more on ergonomics. They focused on— another big thing was actually, this was probably a huge lesson for us, Evangelism. John Resig, the creator of jQuery, loved to talk about jQuery. I mean, sounds easy, right? But this is not intuitive for engineers. You can be the best engineer in the world, but if you don't go out and actually Ryan Dahl, the creator of Node, also one time verbalized this and it really resonated with me. If you're not out there selling and marketing, it doesn't matter how good your engineering is. And so Mootools, we're all very kind of like low-key people. Like we're doing our thing. We love our way of doing things. We had a very strong sense of aesthetics. I always respected the creator of Mootools tremendously. You know, the metaphor is a little tired, but like, I think his mindset was like, I'm gonna create like the Apple of JavaScript libraries. He was a great designer. He foreshadowed in my mind one of the most interesting job roles today, which is the design engineer. When I grew up, there was a very strong distinction of you're the engineer, maybe backend, maybe frontend, and you're the designer, and engineers are not supposed to have taste and ship beautiful things. Nowadays, the epitome of the people that work on frontend engineering are more like what Steve Jobs envisioned, which is the intersection of liberal arts and technologies, like people that are highly creative, strong sense of aesthetics, usability, accessibility, and also know how to engineer things. The creator of Mood Tools was that, super ahead of his time.

But like, what would be even like, you know, just so let's dive a little bit deep on this, like pragmatic versus aesthetics, like what does that mean in practice?

So one concrete example was we were very caught up on bringing classes as in object-oriented programming to JavaScript, which, which was a very awesome feat of engineering. Because one of the classic dunks on the JavaScript programming language is we can connect it to Minecraft and Java. Java, the grown-up language, has enterprise-grade abstractions. And I'm using this term kind of provocatively, like classes in object-oriented programming. JavaScript is this loose, more, you know, we call it technically like prototypical programming language that didn't have that way of organizing the codebase. And at the end of the day, it didn't matter. So we brought classes to JavaScript, and that was an immense feat of energy. I remember our IRC chat conversations about this were so intellectually stimulating. I would just consume the content of the backs and forths of like, how do we hack the language to add this thing? And actually we had quite a bit of success appealing to, again, the connoisseurs of the language. One of our early successes was we had a company called Cloudera to use and standardize and MooTools. And I remember talking to their tech leads and they were super excited by the fact that we brought this more professional abstractions to the language. But out here, jQuery was focused on pragmatic solutions. Oh, you want to do Ajax? You want to create a more interactive thing? You want to do animations? And we had— jQuery was very focused on shipping iterative improvements, making the code ergonomic, not thinking so much about classes. And the, and the quote unquote CEO, the creator of jQuery, was a great teacher, evangelizer. Another big thing that now I'm thinking about is their focus on backwards compatibility. So they would just ship very frequently and never break and never have to run migrations in the code. So I think I did pick up a lot of really good lessons with me that I carried till today. The backwards compatibility thing just became an extreme obsession for me. We still support apps in Next.js that were written 10 years ago in version 0.1. Wow. And, and always give people an incremental path forward. So, so much of that battle taught me amazing things and evangelism and talking about your software, like literally what we're doing now, like going out to the world, going to conferences, teaching, teaching, teaching. Education is the best kind of marketing.

Wow. No, that's awesome. So it almost feels like they a little bit more focus on the user and how the user is using it versus on the engineering, how— That's right. —users should use it kind of thing. That's right.

And people that are listening to this, they probably can relate this inside their companies. There is a class of engineer who sometimes even subtracts value because they get caught up so much in the how the thing is engineered and the aesthetic beauty of the code. And other people are looking towards the stars. And how do we get to Star Trek?

Yeah. How do we ship?

Yeah. And how do we build products and how do we move the world forward? Yeah. And you have to find the balance between the two.

Yeah, 100%. 100%. That's so interesting. Okay, so going back to the story, you've— you got an awesome job offer from Swiss company. That's right. And what was the company? It was called Pokken.

So they were building a hardware device. The best to think about it, it was, you know, that feature today where like you can bump two iPhones together to exchange contact information. We're building that with a hardware device that you could carry around with you. It was really cool, really innovative. And the web portal was going to be a really important piece of the sort of like social networking aspect that they were building. And they wanted to use Mood Tools. Because they recognized correctly that a lot of the social networks of the future, just like Facebook was recognizing, we're gonna need a lot of interactivity, real-time capabilities, dynamism, the kinds of things that only JavaScript can enable. You know, there's so many ways to do servers and backends and whatever, but JavaScript has this distinguishing feature, which is that it can run in the web browser. Yeah. And with technologies like Next.js, you can run it on the server side, you can run it in CI/CD pipelines at build time, you can run it at the edge of the network. So it's the most universal language. I'm a huge fan of English, having been from a Spanish-speaking country, in that it's universal. It can get you anywhere. You can do business, you can have fun, you can sing. So JavaScript is like that. You can use it for everything. And they recognized that power. They invited me to come to Lausanne, Switzerland. I have a funny anecdote, which is they thought that the expert engineer, the expert front engineer that they worked hard to recruit, they were going to pick me up at the Lausanne train station. The CEO, I watched him do like 3 double takes because he found this little kid. My parents had to do an authorization for me to fly out of the country. Because I was underage at the time, but I was the, you know, this expert engineer who's going to come help move the project forward. So when I realized that, and at this time I have two pending classes that I need to pass to get my high school diploma, which are really funny. One is a history of Argentina. Advanced, by the way, to their credit, this was like a deep course. Like, it was like Argentine history and philosophy. It was a very advanced course. That's why I couldn't pass it easily because I know the basics of Argentine history, but like this, there's so much content to learn. And the other course was Portuguese. Oh, really? Yeah. That's funny. And the funny thing is these were the two, uh, mandatory but optional classes in the sense that you could pick between IT and history and English and Portuguese. And given my background and expertise, I could have made my life easy and pick IT and English. Instead, I do what I frequently do, which is like, how can I push this to 11? Oh, wow. And I chose Portuguese, uh, and, and history. So, cause I was like, okay, I already have mastered English by this time. I was working online, right, and stuff, working online. Uh, we wouldn't do that many calls, although I had been on a few Skype calls at this time. So my, my, uh, written and spoken English were really good. And obviously my computer stuff was pretty, pretty tight, especially for like high school level stuff. And so I made my life harder. Uh, but by the time, so this is the summer, I need to go get these classes done. I worked really hard to get Portuguese. If any Brazilians are listening, you should be proud of me. I learned enough to pass that one and then I couldn't, the history one I just couldn't do. By the way, Portuguese I barely, barely passed. Portuguese is actually a very difficult language.

It is complicated. If you go to Brazil, can you speak some?

I mean, now it's just all portunhol. I'm very proud that one of our principal engineers and one of the early members of Vercel, Mateus, is from Brazil and I can understand when he speaks and speaks to his parents, his friends, like I can get it. So I can say like, I understand enough to get around. But this is the pivotal moment of my life. Like, do I continue with the traditional education system or do I embrace a life of like being able to go to other countries? I'm starting to get invited to conferences, I get more job offers. Facebook had reached out, which was extremely exciting because again, they were recruiting a lot of people in the Mootools team. I think it was a tough sell for them that I was, I think I was like 16 or 17 at that time. And I'm in Argentina. I'm not in Menlo Park or anywhere close, but I was getting, you know, more notoriety in the world. And so I had to choose. And obviously I chose the world of, you know, pursuing my passion in frontend engineering and then eventually entrepreneurship in Silicon Valley. Is there any regret? Zero. Zero. The only regret is, I mean, people, as much as it's fun to say you're a high school dropout, I think like I should go back one day and get my diploma. Oh, you didn't even finish the high school? No, no, I'm— that history course.

Oh, that history course bombed you?

That ruined everything. Wow, that's amazing. And these people, again, I respect those teachers so much. They took each class so seriously. Like, they would never let you get away with anything. It was perfection or nothing. So there was no way around the intense scrutiny. And one of my skill sets, so the thing that helped me get into this school among the top 10, one of my skill sets was memorization. Oh, interesting. So I'd memorized the entire textbook of history and geography. So the score was from 0 to 50. I got 50 out of 50. I got every single point in history and geography because I'd memorized every single word. So my thesis was for history, it was like, oh, I just memorized everything. But the amount that I had to study was easily like 20 times. So I am making this shape with my fingers because in Argentina, like, books with photocopies of books, like you, like, like Xerox, like photocopies of books. And just the content was overwhelming and I just couldn't do it.

So, uh, that's amazing. I think you ought to be, I'm going to ask ChatGPT after, probably one of the most successful high school dropouts ever.

I think there's quite a few here in Silicon Valley, but I think college dropout is very common. High school, I don't know if there's that many. Yeah, but again, so very thankful to this school in that they were giving us so much good content that I feel like— and this is a hypothesis that I would love to, like, see play out in the world more broadly. What if, for the world, we raised the bar in high schools? Forget college. We make high school 10 times more interesting, more focused on what your passions are. We bring in content from university-level courses. And yeah, we make— and with AI, of course, I think you're going to be able to get unstuck way more easily. And I'm not going to discredit the entire university system, right? Like, I think there's still another, you know, few years where you need to harden and sharpen those skills. But I do think it'd be— based on my experience in Argentina, I think I would love to see more schools like that. And I'm sure there's more here in the US, I think. Because my youngest is 8, I haven't yet researched enough. But, um, you know, interestingly enough, you asked me like, what are the things that now I try to pass on to my kids? I'm actually a little bit more relaxed on the elementary school phase because I feel like I was able to take on so much more pressure when I was more around like 12. So when I was 12 is when I did the intense studying. For getting into high school. So I would, I would go to elementary school in the, in the morning, and then I would study the whole afternoon and sometimes even into the night. So this is the time of my life when I'm able to pull like really long days of intense concentration. And so you have to find what is that time for your kids that you could, they could actually pull this off. And my hypothesis is that it's, is in their early tens, teens, rather than like when they're like 6, 5, 7.

At least that's kind of how I'm looking at it now. Interesting. Yeah, no, I do feel that people do get interested though in like different ages and different stuff, right? Like some people, you know, you meet, there's someone who works with us that like start coding when they're like 7, you know? Right. And then, you know, some people take college, you know, and a bunch in the middle.

I think I've seen everything work at this point, man. It's like, it's funny because I have successful engineers at Vercel who started super late. I have some of the cases like mine started like sometimes I meet people that had the exact same experience. It's actually fascinating. I met, maybe it's happened a handful of times, but like literally the exact same, you know, hacker challenges and timelines and times that they discovered Linux and open source. It's kind of fascinating. Um, so there's a little bit of a blueprint. Totally, totally, totally.

So how many, how many jobs until you started Vercel?

By some definitions, none. None. In the sense of like, I think my only job was the one in Switzerland, realistically. And for a summer, I joined my dad's team at his company to like do like some like programming stuff and whatever, but it was more like a summer job. So my only real job that I couldn't hold for very long due to my personality was that startup. But my take on jobs is the thing, is the following thing. If you have a job, you have to be the best. You have to give it your absolute best. If my job is to clean this room, I'm gonna be so obsessed. Like, you're not gonna find a freaking, like, piece of dust. If I'm working for that company, I'm gonna ship something for them. I'm not gonna quit, like, 3 months in. They're betting on me. They're flying me to this thing. So, like, I wanted to complete a project, ship. And so in that process, I realized I do have a talent for like starting things from scratch and moving really fast. I felt like I was shipping really fast. I met two colleagues there who like we really gelled and we could work really well as a small team. And I was like, when we— when I came to Silicon Valley in San Francisco, I was like, wait, is that what a startup is? Is you and two people? And you can get really intensely focused together and you have complementary skill sets.

And you still get paid to do that.

And you get paid to do that. And all day you can pull— goes back to like those days of like being up all night trying to like, like find a breakthrough in my Linux days and whatever. And you can do that all day and all night while you have snacks and, uh, you know, Diet Coke and espresso. It's like, that was like the definition of me, like of a good life. Yeah. And so, and that's what basically Silicon Valley showed me that was possible. So I get into this coworking space that that company had rented out and I start meeting other startups. I was very lucky that some of the other startups are like ended up being like cracked successes. I was like, my coworking mates were the people that started Instagram and Twilio and a lot of other like big websites, TaskRabbit, like all of those. Were in like this little coworking space on the Embarcadero here in the city.

So how do you move here to the US?

So the company in Switzerland opened their basically like quote unquote USA HQ by renting out that coworking space.

So they sponsored your visa and everything for you to come?

No. So this is probably the hardest time in my life. Going back to like, I tend to play hard mode. I had to figure out, okay, I'm gonna quit that company. I shipped some really good stuff. I'm gonna quit and I'm gonna start my startup with these two friends of mine that I met at this company and then another one that I met at the coworking space. How do I get a visa? You know, at this time I had already made some pretty significant contributions. So I was already a mutualist developer. I was starting to get invited to conferences. I had letters of recommendation from people, but getting to basically be able to stay in the US and start a company was one of the hardest challenges of my life. Because on paper, I was horrible. I had dropped out of high school. I was from Argentina, which is not among the like highest reputation countries for the immigration system in the US, right? Like I think, you know, Germany is probably way higher when, if you're an immigration officer and you have an application, like this might've been one of the strangest applications that they probably have gotten, right? So no college degree, not really extensive work experience, right? Because I'd only had the Swiss company. So I actually had to endeavor in proving to the world that I was worthy of this visa, which is called Alien of Extraordinary Ability. 01. 01. And that's what I have till today. So that was for me, like one of the— I embarked in this journey of, you know, accumulating evidence. One of the hard things was I had built a few open source projects that became extremely popular. Which ones? One is called Socket.IO. Oh yeah, I heard of that. Yeah. So Socket.IO, it still surprises people because Next and Vercel have gone so much further that, you know, like people don't know of my past, but Mood Tools and Socket.IO were two very important projects in the ecosystem. And that got me, it was kind of became like a mountain of evidence, like for the immigration system, because these things get really, really big. Socket.IO ended up powering like Microsoft's products, a lot of startups. In fact, it even powered research into submarine vehicles that were remotely controlled through Socket.IO. So for the uninitiated, Socket.IO is a real-time communication system focused on cross-platform, cross-device support. So it enables two computers, devices, et cetera, to communicate in real time. And it makes it really, really easy for developers to create real-time systems. So it was starting to get written about a lot and it was on engineering news websites.

Like if you wanted to build like an online chat, this was the default, right?

Exactly. Which chat is so back now. Like funny enough, companies like Perplexity and Notion were born using Socket.IO. Oh wow. So I'm very proud of the fact that the project continued to sort of thrive over the years. Now it's more like you don't really need it as much. People still use it, which is actually kind of cool, but it's less necessary these days. We've baked a lot of the technologies that Socket.IO enabled have been baked more formally into web browsers, into web servers and things like that more natively. But it was a pioneering project. So that project really helped me a lot. I got invited to conferences. I gave a pretty interesting and I think novel, maybe like somewhat disrupting talk at a conference in Europe, in Berlin. So I was able to come to the US and say to the USCIS, here's my application. I've done all these things. I wrote a book on Node.js and JavaScript, but definitely like hard mode in terms of immigration.

This episode of HD and HD is brought to you by a brand that's close to my heart, Brex. When we started Brex, it wasn't just about creating another financial product. It was about solving the real gritty challenges that founders and startups face every day. Building something from the ground up is hard enough without having to navigate clunky legacy banks with endless fees. That's why Brex is the financial stack that scales with you no matter where you are on your journey. From maximizing runway to earning yield on your cash, Brex was designed to make every dollar count so you can focus on what really matters, building your dream. So whether you're just getting started or you're ready to take on the next big thing, Brex has your back. Check it out and see why the world's most innovative startups like Anthropic, Scale AI, and Robinhood trust Brex.

And now at this point, okay, so you were able to get all the evidence, got your visa, like what was the first idea that you guys had? And like, And how is the— what came first, Vercel, Next.js? Like how did the two like interact in the beginning?

Yeah, so I'll give you the first part of the story very quickly. So it's so funny in retrospect. And this is why I recommend people that they have good founder idea fit. My first idea was revolutionizing education through modern software platforms, which is kind of the silliest idea if you consider that I knew nothing about education. I was trying to sell edtech products to the school system here in the US. In the process, we built an amazing software platform. It was so innovative. It was real-time. It had real-time file sharing capabilities. It was a website for teachers and parents and students. It was a whole platform. And we realized and we were highly inspired because we'd seen how Instagram had succeeded. They were our peers in this coworking space. They succeeded by stripping down their product to the fundamentals. A lot of people don't know this, but before Instagram, they were a more complicated app called Burbn. And Burbn had a file— not file, it was a photo sharing feature. And then they realized, what if the whole app was just photo sharing? So a lot of success is in what you choose not to do. In what you choose to delete and strip down to the essentials. So we stripped down the edtech platform and we did a hard pivot into more of a collaboration through file sharing, like really easy, fast linking. Like I could like drag and drop a file, immediately get a link, send it to you even before the file is completed uploading. And the company was a little old at this point. And so we were a few years in, we didn't have a ton of funding. But we did have this really, really, really cool platform. And I remember one weekend they showed it to the founder of WordPress, Matt Mullenweg. And I remember like his expression was like, holy crap. Like I could see how we could take this technology and bring it into WordPress and together we could build the future of WordPress. So we ended up selling the company to WordPress. Oh wow.

And this was the first company you did?

Yeah. So my first company was a pretty, you know, people call it an acquihire these days or whatever. Like it was a small but quite meaningful for me exit. You made some money. I made some money. I went from having zero to having more money than I knew what to do with. That's amazing. Which for an Argentinian dollar constraint, the first time, it's like—

Like you could like not work for a little bit and it's totally okay, right? Yeah.

But because I always do things in hard mode, I didn't do that. So what I did is— How old were you when this happened? I think I was 22 or 23. 23. Wow. Amazing. So when I left WordPress, the very first thing I did is I started angel investing. And so I literally put everything back into the game.

Oh wow. Everything. Any good outcomes from that?

Yeah, well, Scale AI, which just closed yesterday, was one of my first investments. Wow. Okay, so that one you did well. But I'm happy for my brother because he was the sweat equity part of the deal. And, but actually, so I do say, and this is going to sound weird, but it is quite easy to make money in Silicon Valley. You just have to legitimately build here and spend a lot of time and energy and become a really big part of the community here. And you're just going to naturally see that people around you are building the future. You don't have to be the Oracle of Omaha to find the opportunities here. And it's amazing that we've built an ecosystem that it's so wealth creating. I mean, the future is still, I mean, people talk about like the emerging tech hubs of the world, but I believe that SF, especially with AI, has this unreasonable advantage. Like there's probably more like AI unicorns being born today. Forget about like OpenAI, whatever, like even just today, I talked to a cracked entrepreneur today who is building an AI startup on Vercel. That I thought today, like, holy crap, like, it's a thing that kind of already exists, but there's a whole new take with AI. And like, I was like, that's amazing. Like, and again, that's why I call it easy. Like, this is what you get to, you get your pick of the litter in many ways. So it was kind of a weird thing to do at the time. My dad was very against it because I left 2 years worth of vesting on the table. Which is probably like, I can't remember exactly the number, it was like millions of dollars. And I left, I didn't tell him this because I think he would've been like, he would've called the cops on me or something like Britney. But I started investing in a lot of startups. And I mean, I do have, looking back, I did have the privilege of like, I was single and ready to mingle. And I had like, I was in a position where I could, risk it all again. And so Bitcoin was also another early thing that I invested. Wow, that's amazing. Yeah, I didn't hold it until now. I had buy, sells, buy, sells and things like that. But again, going back to Silicon Valley and Hacker News, I remember I found Bitcoin through Hacker News and it's like people were like, oh wow, you can create digital money. It's like, wow, amazing. And there's another funny story of being here at a party and running into Brian Armstrong when he's starting Coinbase. So that's the beauty of this place is like you literally like you go to parties, you go to meetups, you go to this and that and you're meeting the builders of the future. So started doing some angel investing, but I couldn't shake the idea even while I was at WordPress, I couldn't shake the idea that the future of software was going to change very fundamentally, that the idea of static websites and content were going to give way to a whole new slew of applications that were highly dynamic. I, what I tell people these days is that the web has been on a journey. The web was initially for documents and static websites. Later on, much of it thanks to React and Next.js and the things we and our peers and partners are building is becoming a dynamic medium for applications. In fact, so much so that ChatGPT, the thing that everybody talks about has changed the world, started out as a website, started out as a Next.js website. So the medium of the web gave rise to an application platform. And what I'm excited about these days is that we're going from dynamic to generative. A dynamic application is almost like a poor, poor man's LLM-generated application. Imagine a world in which any website that you open was generated just in time by an AI. And that's the future that now we want to enable with Vercel. But even the idea back then was very similar to the idea today is like, how do we enable people to create? I was a little frustrated. I don't know if TikTok was around at this time. I was actually a little frustrated that I could see the world moving more towards consumption. Everybody's just becoming a newsfeed-consuming drone. And I had an idea that we could make it so easy to create applications and deploy them to the web that we could inspire more people to be entrepreneurs and startup creators and again, participate in the value-creating side of the world.

And so for the non-technical audience, can you explain a little bit what Next.js is?

Yeah. So creating a web application is actually pretty freaking hard. I'm, you know, I'm going to set the bar high. Like if you wanted to create the next ChatGPT, right? And again, ChatGPT had Next, at the time it didn't exist, but creating all of the tools, setting them up and configuring them was very, very difficult. Not only on the, again, the tools that you use to build the application and the languages and stuff, That's what Next.js kind of covers, but also the cloud. The cloud was— I was— when I left WordPress, WordPress at the time was on-prem, so they had their own data centers. I had seen firsthand how difficult it was to maintain your own infrastructure. But then I looked at the cloud and I didn't see a massive simplification. I was like, okay, it almost looks like Now you're maintaining data center, but a virtual one in the cloud. And it felt frankly like tedious and distracting. So if your goal is to create an application, why are you spending all of your time configuring stuff? And I lived it firsthand because as soon as I left WordPress, I wanted to cook on my next idea. I literally didn't even take a vacation with the exception of a short trip to Yosemite. And I was like, okay, I'm going to create my website. And like, I spent weeks and weeks and weeks configuring stuff and setting up the cloud and whatever. And so that process actually distracted me entirely because I had another idea at the time. I was like, okay, this idea is even more important. The fact that even for me, who's been programming since I'm like 7 years old, it's so hard to bring an application online. I'm gonna disrupt this.

I feel that like Heroku was like really good and it kind of just got worse after that, you know, like VWS.

A lot of people actually told me that at the time. I remember one of the first people I pitched was like, but doesn't Heroku solve this? And I told him like, look, like there's a huge opportunity here. Like it's gonna be the next 10 years of development. And one thing that Heroku I think had missed is that you still had to create your own tools. You still had to figure out the framework part and it felt like they were not fully integrated. And I had aspirations that like far exceeded what the state of the art were possible in terms of deployment time. One of the biggest obsessions of my life, I think one of my character traits that helps for this kind of business is impatience. My idea of building software is that you type something in and you press deploy. And it's live instantly. Like literally no build time, no CI/CD pipelines. It's instant. And so if you show me, well, look, I can do it in 10 minutes, I'm not impressed. If you show me I can do it in 100 milliseconds, you have my attention. And so we saw an opportunity to bring an order of magnitude speed up maybe two orders of magnitude. We have clients that are some of the world's most renowned brands of the world. Like we host some of the largest e-commerce websites on the internet and they tell us we could deploy only once a quarter. We could deploy maybe once every couple of weeks. The Vercel way is you're deploying software instantaneously. To any number of users every minute of every day. And of course, this comes as a little bit of culture shock sometimes to the world's most established enterprises. But the reality is that you need to move at the speed of thought, especially in a world where there's an AI model dropping every week. The chessboard is being reconfigured on the scale of weeks. So that's why Vercel has been so successful, is because we We didn't know AI was coming, maybe not, certainly not at the pace that it came and in the shape that it came. I couldn't have predicted that. But what I knew is that anything that I'd done successfully in my life required extreme iteration velocity, that you can ship and try and ship and try. And we've been able to attract incredible talent to the company, like our CTO, Malte Ubel, built these systems at Google. For 11 years. He— Google had independently discovered that in order to empower engineers to keep up and innovate at that pace, you need to have the best possible deployment infrastructure and the best possible integrated frameworks. Facebook had discovered the same thing. In fact, Facebook invented React, which is the sort of like the engine of Next.js. They discovered that, look, if we need to iterate and we need to body Snapchat and we need to body TikTok and all these things, we need to iterate and we need to deploy and ship really fast. So Facebook became known for just how fast you could check in a code change. In fact, it's still an inspiration, I think, for many companies at the scale that they're operating with millions of lines of code and thousands of engineers. If you want to ship an experiment to a billion users, it can take you 10 seconds. So they have incremental compiler pipelines with very smart caching systems, some of the things that we're democratizing with Vercel. And you can ship to any number of users any, any app that you want in seconds. And so that's what's made Vercel work, is that yes, there has been other ways of creating the cloud and configuring the cloud and whatever, but like the fact that you have a platform that has already been thought through by some of the most successful engineers of the internet. And you can just like, by the way, this is something that we give everybody. A student that wants to deploy their first web app, they can go to Vercel and create it there in seconds for free. Or we now have 2 of the top 30 largest websites on the planet hosted on Vercel. So the other thing that I'm quite proud of also on behalf of our team is the breadth. You can build any kind of software on this platform as long as it's accessible through the internet. You can build it.

And how does the, you know, the front and back end work? Is it like all in one?

Like how does that whole setup work? Yeah, we started out very focused on the user interface side of things. I mentioned that, that one of the things that Versaillians kind of tend to gravitate towards is how things feel. The look and feel, the performance. That's kind of one of the outlooks that we brought to the market is like, focus again, kind of going back to my lesson of like jQuery versus Moodle, right? Like we're very focused on the product outcomes. That's why I said iteration velocity. I didn't say iteration speed. You might have an engineer at your company or you might know people in your life that move really fast and go nowhere. It's a sad reality of the world. Like you have to move fast in the right direction. And so the point of view that we bring to the market is let's work backwards from building an extraordinary product experience. This is why we abstract out the cloud and the infrastructure and AWS, we are coming to the market with a strong point of view. If you're a Fortune 10, 50, 500, 5,000, you name it, even a startup, you should not be focused on how the low-level implementation details work. You should be focusing on, is this what your customer wants? Is it converting? For brands like Under Armour, We've lifted their conversion rate extremely significantly and we've made their engineers more productive. And so we're very focused on the business outcomes and the developer experience. It's really interesting actually. Like we're a cloud that appeals to developers because of the ease of use, but is very, very focused on what will your mom or my mom think? Like what we want them to think is that they access these applications and they're like, They're fast, reliable. Maybe they're doing shopping during Black Friday, or maybe they're actually adopting an agent or the latest AI application. And we want those to be, you almost just know that they were built with Vercel technology because of how good they feel. That's awesome.

If you're like now investing in some like dev tool startup and you need to give some advice on how to succeed in this industry, like what would the key things be?

That one is one of my big ones, which is that duality that you have to keep in mind that you have a customer that is the builder. And this is true of anything, right? Like imagine that you're selling raw materials for construction for a building. Like you're not just focused on like your buyer buying the raw material. You're focused on, you're thinking about hopefully, I mean, if you want to be successful over the long term, you're focused on the person that is going to live in that house or have an office in that building and they want to have an amazing experience, right? They want to have materials that feel good, that are durable, that— so the world of software is not that different. It just moves at an insanely faster pace, right? And so that duality of personas that you have to think about is something I advise a lot. The other big one is, look, AI reset the board. It didn't matter what you thought about the world, your reputation. I mean, you might have some assets that you can leverage in this world, of course. Like, I'm not going to say like the competition started and we're all at zero, but you have to rethink from first AI principles. So the journey that we've been on is we realized that frameworks are not enough in this world. Frameworks make, like Next.js, and this requires almost like an ego death situation. Frameworks got us here, they're not gonna get us there. Frameworks make it easy for people to create. And we've been really successful at that. We have 1.4 million monthly active developers in Next.js. In fact, one of the biggest selling points of Next.js is the community that you're joining. I was chatting with the CIO of a Fortune 10 company this week and I was telling him like, look, The reason that Next.js is so compelling for you is that you go to an LLM and it knows Next.js. You go to the job market and the people know Next.js. You might acquire a startup. Do you want to integrate the technology fast? It's Next.js. But Next.js has reached millions of people. I believe that AI-based code generation, what we're doing with v0, which is our web agent, is going to bring us from millions to hundreds of millions of people because the language, the Next, as Jensen Huang says, the next programming language is human. You can go to v0 and instead of having to learn Next, instead of having to learn JavaScript, all the things that I had to do, you can just type in English and you get a working full-stack application in seconds. And so Vercel had to go through the process of inventing AI applications so that now we could take the lessons and teach them back to our customers. So if you're building a new startup today, first of all, you have this amazing advantage. You can build v0. You can build an agent instead of building something that might be more, you know, for people that are more visual. The metaphor that I would give them is think about how different the interface is between Adobe Photoshop, like the classic. If you imagine what Photoshop looks like, what I picture is menus and tools and nested menus and columns. And I actually never— I could never learn Photoshop. I didn't have the patience to learn. I'm more of a math brain and code and whatever. I never learned the incantations of clicks and drags and whatever. And now think about ChatGPT with image generation. That week where we're all building Studio Ghibli photos, you could say to ChatGPT, build me this image that looks like so-and-so. People are using it to create logos. Your next big startup that you wanna launch, you're probably iterating in a chat-based UI like this. There's Vercel customers like Scenario AI They help you create game assets with— again, you don't go to Photoshop, you go to a conversational interface to create game assets. So think about that advantage that you have today. And if you're listening to this and you're a big established company, think about your— you should be thinking about your— my days are numbered until I come up with the agent version of myself because customer behavior is going towards this new kind of interface. Sometimes there is no interface at all. Sometimes it's actually you are deploying an agent that gets called by another agent in another interface. This is why people are so excited about MCP and MCP being almost like the new HTTP. Today we announced there's 4 customers that have built very successful MCP servers on Vercel. One of them being Zapier, one of them being Solana, one of the most successful blockchain companies, So sometimes there's no interface. Why? Because I might be using my phone and I talk, and that is going to call an MCP server to bring the value of those companies into the new agentic world. And so I think it's exciting to be creating new companies today because of this precisely. You're going to be creating dev tools or startups or infrastructure or user-facing products. That have completely shifted what that frontend looks like for consumers and would allow you to leapfrog either the incumbents or your competitors, depending on how quickly you can adopt these new methodologies.

And what do you think, like, you know, if you look ahead and you think about how programming is going to work in the future? So you mentioned that you look at cloud and at cloud people are still managing that infrastructure, right? But despite Vercel, there's still a huge amount of people that that is their life, right? Do you think that is, which direction do you think is gonna go? Who do you think is gonna end up managing their own infrastructure versus who's just gonna deploy?

Great question. So I think the world never works by replacing atomically everything that existed with the new thing. You know, there's still FM radio. There's still a fax machine somewhere. Totally. So the world evolves by exponential growth curves. What's happening is not that there's no more DevOps, but it's that the companies that offer no DevOps and complete infrastructure automation like Vercel are growing exponentially. And so because they're growing exponentially, they're going to become a substantial part of the future. And what I would like to say to people that like infrastructure is apply to Vercel. Because I want to hire the most cracked infrastructure people so that we build the infrastructure so that no one else has to build the infrastructure, so that we're not all rebuilding and reinventing the same wheels. It's like, at the end of the day, it's like building concrete, it's like building highways. That's what we're doing, right? We're in the business of building those foundations so that other people can build on top. And so I would love for that world to sort of like, okay, what is the next problem to solve? So chapter 1 for Vercel was, I call it like the self-driving car of infrastructure. Like, can we create, can we give the world autonomy for infrastructure? To your question about code, I think chapter 2 is, turns out you don't have to write code so much anymore. So bringing autonomy to code. And so with v0, we're starting to experience this. We already have 3 million users that are just typing in English every single second. There's 6 new app generations on v0. I could have never imagined that scale in the old school, old school, like I'm going to sit down and create an app world of the, of the developer world. The pace at which AI-generated code is growing is going to be the next exponential. And just like I said, I mean, I'm not going to say, you know, writing software by hand is the fax machine. I think that'd be a little too extreme. But what's going to happen is that we're going to see that AI-based developer platforms are going to cast a massive shadow on the more manual techniques. You're still going to require expert coding for the frontier. So the frontier is, for example, solving problems that have never been solved before. Some of the infrastructure challenges that we have, for example, at the scale that we're operating, we cannot take chances with AI-generated code. I was actually chatting with a partner today who was telling me he had an SEV 0 because one line of code that was AI-generated snuck through into a critical system. One line of code. It was a deletion of a line of code. So this is a world that we live in, right? Sometimes you delete a line of code and it seems inert and it passes your tests and it passes staging. It actually passed multiple levels of— I was so fascinated. I was eating popcorn while he was telling me this story. And so there is a level of engineering that we do at Vercel in which actually LLMs are not that productive yet. There's certain programming languages where it's not that productive. There's kinds of frontier code that is not helping, but the world doesn't write frontier code exclusively. In fact, I would argue that like 95% of code that gets written has been written before in some way, shape, or form, maybe not the exact form, which is why LLMs are so successful. LLMs, the metaphor to think about them is that they're like almost like combining a database of everything with a slight permutation on the data so that when you ask something, it has all of this knowledge. Some of it is fuzzy, kind of like how our memory works, but it creates a recombination of all of that knowledge that can be so timely that you're like, holy crap, that just saved me a week. And so because the world evolves by, again, there's a frontier, but then there is a lot of lag. And there's a lot of repetition, AI will make us exponentially more productive. And so I'm excited about this world because the number of people that are going to be able to contribute to software or create new software we've never seen before in the industry. Silicon Valley and the world got caught up a lot in like, how do we bring more people into the industry? What inclusivity program can we build? AI, turns out, is the ultimate inclusivity program. And it's not an inclusivity program, it's literally technology. Technology solves all problems, right? And so I talk to enterprises these days where what v0 has done for them is that everyone that was not in the engineering department could never speak the same language as the engineering department. I have a funny customer in New York. I was chatting with the CEO recently and he told me, Engineering is like the government. I feel like I have to petition to them. That's so funny. It's like a bureaucracy. I have to convince them. They're very powerful. And this is the CEO. The CEO felt like he was overpowered by the engineers who say, no, it's too complicated. It's too difficult. Imagine if you can start doing the work that they do, even though you don't know how to code. And this is not because you're going to like tell them like, move over. I'm going to ship. I'm the CEO. I'm going to ship., but now you actually start earning more credibility. You start having more meaningful conversations. You start exchanging prototypes. You start sending links. And so that's kind of the broad impact that I think AI is bringing to coding is it's a true democratization effect. It's the pun that we like to say is everybody can cook in a Ratatouille sense.

That's amazing. No, I think that's absolutely true. And I think the amount of software probably is going to get written in the world is going to increase exponentially, exponentially grow, right? Like, so where are we in the bottleneck? Like, what's the bottleneck today between that and that vision?

In terms of AI adoption for building software, I think that's happening extremely quickly. You see the rise of the AI IDEs. First, we saw the rapid adoption of Copilot, then you see the rapid adoption of Cursor, Windsurf, Claude Code. I think the appetite for adopting these tools, unprecedented. We're seeing it, it's unprecedented on an individual sense in that it's actually making the internet fun again. When I was, uh, in my childhood playing with GeoCities and building my own little websites, like everything felt novel and exciting and new, and you could create websites that were unlike any other website that existed out there. Then we went through the plateau of productivity and like, and everything becomes business. Now we're back to like everything's exciting and new and everybody can do anything. And so individuals are adopting a lot of these tools, like literally paying out of pocket. But we also are seeing incredible penetration within enterprises in even highly regulated environments, healthcare, law, finance. There's no customer that I've talked to that is not interested in adopting these tools. In fact, I have a funny anecdote of investment bank type of company where 6 months ago they told me, we're going to adopt Next.js, we're really excited, we're adopting Vercel, but don't say the AI word in the office. Like the regulatory environment and like the vibes are just not right. We're actually going to give a workshop at their office. Now we give a subsequent like chapter 2 of the workshop and It was the exact opposite. The leaders of the company were like, we need v0, we need AI, we need models, we need LLMs. Can you give us more? The conversation had shifted entirely into AI. So I think it's interesting because it gives you a little glimpse of the lag times. In Silicon Valley, we adopt everything day zero, you name it, like blockchain, AI, etc. In the enterprise, there was a little bit of a lag, but I now have not seen any category, even government. We've had amazing conversations with government bodies where you hear the CTOs of these entities publicly state, we're adopting AI, we want AI, we want to help service the public with AI. And I think a big motivation, by the way, is the world doesn't want a singular AI to rule them all, like the Eye of Sauron. Like ChatGPT is awesome, but I think we all want to have our own intelligence out there in the world. Jensen calls it token factories. A token factory is that in the old world, you were producing software downloads, zip files, whatever. In the new world, you're going to be producing tokens. Maybe whether you like it or not, because if you're not producing tokens, you're going to be outcompeted by another token factory. And so these are kind of the things that I think are driving this interest.

But if we look at Silicon Valley, at the bleeding edge, I think the founder of Cursor did a podcast or tweeted the other day that we're still a little bit limited by, for big code based on context and for a 10 million line code base, you need 100 million tokens and the models kind of get confused. Where do you think in your view is like the bottleneck for kind of getting to this vision even here at the bleeding edge?

Yeah, I think one of the unique advantages from my point of view at Vercel is because we build the frameworks, we build compilers, we build infrastructure, we don't have to solve every limitation of AI for legacy code languages, et cetera. So we're not so focused on like, okay, what does a 10 million line codebase look like? We can break down the problem in ways that the AI can actually make more sound progress. Mm-hmm. We can actually guarantee that the code even works, which, like, with LLMs, there's no guarantee. Yeah. So we have, we have trained models on top of the foundation models that massively increase the likelihood of success of the code diff. And then we have a runtime environment. So like when we run the code, We use our tooling, obviously, and we're able to actually, these are all stochastic systems and probabilistic systems, but the probability of the code actually working is much higher. We've done a lot in the security space to make sure that the code that we produce is secure. So again, v0 is taking a different vantage point into the problem, but I think we haven't seen so many of those limitations. But I'll tell you, limitations is a very emergent technology, of course, right? So there is paper cuts, everywhere. But the wins so far outweigh the things that you could nitpick about the technology. In some ways, it kind of reminds me of like when ChatGPT came out and people were like super concerned about hallucinations. Yeah. When was the last time you heard someone be concerned about hallucinations?

Yeah, no, it's kind of out of the topic.

You know, the reality is that with Search Grounding, which like brings, you know, fresh data and content into the prompt, the hallucination problem has broadly been eliminated. And even when it's there, it's very hard to say that your own, let's call the support chat problem as one of the examples, your own support team members at scale are hallucinating. Humans make mistakes. It's the same problem as self-driving cars. In fact, the best metaphor that I found for v0 and this idea of autonomous infrastructure and code is like a self-driving car. Waymo got to a working prototype, what, 2 years in, 3 years in? If any, anyone is like that actually knows this is listening, like you can call me out, but like few years in they had a working prototype. How long did it take them to get to dominating a city? I think they first tried it in Phoenix, Arizona. Took them a while., but then you master that environment. So this is kind of what our approach with v0 is like. We're mastering the environment of this full stack, little sometimes, but full stack applications on Next.js. And what we monitor is the incident rate. You could call it like Waymo had like incidents or the disengagement rate, which is that Waymo has a mechanism so that a remote teleoperator can gain control. Tesla Full Self-Driving has the same system. Like you can like watch the car do its thing and then you can grab the steering wheel. Coding tools are gonna have this same thing that you can monitor very closely. Like when are my customers disengaging? In fact, before the podcast, someone tweeted at me, they had v0 feedback and where to send it to and said, please send it, please reply to me now 'cause they just wanna read it. And I'm sure that he's gonna talk about he's going to tell me— I'm kind of predicting because I get many of these— he's going to tell me about a scenario where it had to disengage, like v0 had to, like, could not get that far. And so just like Waymo made progress one city at a time, I think we're going to, like, conquer this. And once the system learns, it learns for everybody. Once it learns for our millions of users, it's going to learn from the next 10 million users and the next 100 million users. So I don't see this stopping anytime soon. I think we're going to conquer the codebase size problem, the complexity problem, the hallucination problem, the security problem, the performance problem, you name it. And again, that doesn't mean that there's not going to be humans controlling the systems. I think there will be. But at this time, I think it's just exciting to think about just how much better on autopilot these systems will get.

And so what are the principles, you know, when you're thinking about building Vercel? I think one of the things, for example, you talk about your jQuery battle times was, you know, this kind of like pragmatic versus pure. Yep. You know, like building more on that, like what else now that you're building Vercel and V0 and Next, you know, are kind of some of your core product principles to make sure things are going to get adopted?

Yeah, one value at Vercel is called iterate to greatness. You never reach greatness. You never reach the perfection state. You have to iterate. We call it ITG for short. You really have to break down problems into pieces that can be digestible. You have to, you know, sometimes we're thinking about a major infrastructure project that it might take us months, years. And I always challenge the team. Like I ask them, what is the ITG? What is the thing that we can ship today? What is the ship, the thing that we can ship in the next hour? And there's always something. A great example is before we built the AI SDK. So for context, Vercel has open sourced a lot of the foundational components of v0 because we want other people to, one, compete with us, build other website encoding agents, to build agents across other domains. Like legal and finance and construction and you name it, logistics. And so our operating model at Vercel is we are patient zero. This is another important framework. We're patient zero of the technology. We launch technology when we like it. It might seem arbitrary, but we have a lot of smart people at the company that have come from amazing companies. We have young people, old people, you name it. We have— we, we are like Rick Rubin says, we, we're confident in our taste, uh, and we want to feel the pain. And so once we are patient zero and we have, uh, um, another principle of ours is to be truth-seeking. Once we've reached truth, we put it out there into the world, and it allows us to speak about things with a level of groundedness that I wouldn't have and a level of self-assurance that I wouldn't have if I wasn't customer zero. And so, but even before we get there, I always remind teams, look, I know that we're being patient zero to create this thing. Maybe there is an ITG that's possible on maybe it's documentation, maybe it's an example. So I'll give you a very concrete example. We created a very little MCP framework so that customers can create their own MCP servers. And we're actually now quite confident because we dogfooded it and we worked with 4 amazing design partners. Before that existed, the thing that I would tell my colleagues was, all right, we don't have a framework, we don't have anything. How can you still hack it together so that we can make progress, so that we can learn. Maybe we're not going to put it out into the world as a recommendation, but we're going to say like, look, this is a Vercel Labs prototype, you know, go knock yourselves out with this. So I think companies sometimes get too caught up in like multi-quarter planning, multi-year planning. And sometimes you actually, the way that you learn this is that sometimes you do that thing that you thought was so small. And maybe you thought of it as a hack and then it ends up having insane product market fit because people tell you, oh my God, you just unblocked me. So I have so many like war stories of this where like we're building a really big product and then an intern comes along and because we have that value, you know, adds a checkbox here in the product that produced 90% of the value of that big product that the more experienced engineers were building. And so I'm, yeah, I'm a huge, huge fan of, I learned this from one of our colleagues at AWS who said, I asked him one time, he's like, I'm in awe of how you guys have created such reliable infrastructure. But you guys at re:Invent, you also ship a lot of innovations. Like what's the secret?? And he told me, look, like the only secret is reducing scope. You have a big, big, big vision of the future and you don't want to compromise on the essentials. You don't want to compromise on quality. You don't want to compromise on security. You don't want to compromise on reliability. You don't want to compromise on durability. You can't get hacked. You can't lose data. So the only thing you can do is shrink the surface of the product. So that's kind of something that's always been stuck in my mind. And like over time, what I wanna do and my ambition is all of the things that I just talked about remind me of what led me to create Next.js. So Next.js was rooted in an article called "The 7 Principles of Rich Web Applications." And I have another funny anecdote related to Brazil. I introduced that, essay at Brazil.js in front of thousands of awesome Brazilian engineers. And I made a really bad joke that did not land with the audience. I think Brazil had just lost 7 to 1. You went there. And said, oh, it's the 7 principles of rich web applications. You guys should know about 7, something like that. And silence in the entire room. That should come from Argentina though. That's right, that's right. It was just I went too far. But now I'm thinking about what is the framework that I can create but for building more innovations out of Vercel. So turn Vercel into that sort of innovation machine that has principles. Because without principles and guardrails, you just have like, I like another little framework that I like. I listened to Satya at an event recently say, he called it brand permission. You know, companies can innovate in a million different directions. And of course you need restraints. You have to pick your battles. You're not gonna like do 20 different product innovations at the same time if your team cannot absorb that demand, right? Or your team cannot meet demand and have the capacity to do it. But the other thing is that you need brand permission. Like I can't tomorrow launch the Vercel plant. We're gonna go into the flower business. You know, it's like, people would be like, what? Like, it doesn't make sense for you guys. I thought you guys were gonna go into, I don't know, AI for this other thing. And like that would be like brand permission, right? And so this is more advice for when companies are operating at scale. I sometimes even as a user get kind of befuddled, but like big brand, why do you go there? Uh, I remember, I think Apple recently launched, uh, a party invite thing and I was like, but why? Like the meme, why? Like there's amazing. I think at the same time that Apple did that, Android and Google, I think it was the Android App Store had just declared Partyful, which is a fantastic app. Um, App of the Year or something like that. Like it was in their highlights and, and Apple starts competing with them. And I was like, why? It's like, how do you do fellow kids walking into a room like Mr. Burns? Uh, not your place, not the right time. Not like, it's just, again, that, that goes back to like the brand permission thing and, and whether it makes sense, you know, at some point these companies get so big. And they get so rich. Peter Thiel has spoken about this, right? Like Google doesn't know what to do with their money. Like the famous, there was a famous conference where he said like, they're just too cash rich. And if you pay a dividend, it means that you didn't find a way to reinvest it into your business, et cetera, et cetera. And so I think you can find ways that are true and authentic to your mission in the world that that are more tasteful and productive. So that I think is very important for companies that are getting bigger to keep in mind.

That's amazing. And what's, for companies who want to get developer love, like what's the secret for that?

I think number one, Elon just talked about this, I think a couple of days ago, or at least I saw it go viral a couple of days ago is, I think the CEO needs to know what's up. CEO needs to be an engineer, the CEO needs to be a developer, the CEO needs to understand the problem space very, very deeply. People will quickly lose respect for a leader that doesn't, has not walked the walk, right? And I use my products every day. Maybe it's a little bit of vibe coding in the week and then maybe something more ambitious during the weekend and that kind of thing. But this applies to like every industry, right? Like Boeing, I think, is known for having changed from CEO that was an engineer to CEO that was, I don't know, a spreadsheet aficionado or whatever. That can also be taken to an extreme. There is people that bring other really important perspectives into businesses. We all can't just be assembly code experts. And I think this truly is the world that AI should unlock in the future is that we all become more full stack. I think we all become, uh, you know, you can develop more of your creative side. You can develop more of your technical side. I think people will spike in certain, um, uh, areas. It's kind of those like charts that you see that have like multiple skill sets and like where your character is stronger, right? But we all have some strength. In all of the dimensions. I don't think it's going to be acceptable that you hire— we're starting to see this in how Silicon Valley and other big tech companies' screening processes are evolving for screening for AI competence. You hear people saying like, well, like, what is your level of AI proficiency? Like, it's becoming a synonym for, or a proxy for, level of curiosity, level of the ability to develop new skill sets. So I think the company of the future will be people that are constantly upskilling themselves in many ways. You know, keeping up with what AIs are coming out of market is really important because with one tool that doesn't require, I think, extensive mastery. So I'll give you one concrete example. There's a company that allows you to convert text to 3D models and images to 3D models.. And I've been using that because I wanna explore what's possible. If you had asked me, hey Rauch, learn Blender and create your own 3D model from scratch, no chance. Zero chance. But now all it takes is 2 clicks. And that's why I was emphasizing the, the need for companies to rethink their interfaces. Maybe not even 2 clicks. It was 1 click to sign up or log in with, with my Google account. And then drag and drop a file, and now I have a working 3D model.

Yeah, there's a, I think Josh Kushner, who's a good friend of mine, tweeted this thing that I thought was really good, which is the internet democratized information, but AI is democratizing skill. Yes. Right? Which I think is like very much to what you're saying to the 3D model and everything.

It's exciting to be a human, contrary to like the demise of human fears that we're all gonna be kind of insanely skilled.

Yeah, no, that's going to be amazing. And as a closing question here for us, if you look through the story, right, like all the way back from when you were like hacking in Argentina to building out one of the largest tech companies, what do you attribute to luck? What do you attribute to skill? And what were some of the mistakes you kind of made?

I mean, luck immediately what comes to mind is that I was born in a household with loving parents and I mean, not a lot of means, but like a desire to like be better and overcome challenges and good role models and aspirations. And I think that's luck. Like I didn't choose, I guess maybe there's some weird religion that thinks you choose your parents, but like I didn't choose to be born there. So I'm happy about that. Lucky in that I ended up here. Did I find my way here? Did my skills get me here? It's hard to say, but like there was some element of luck in like the different path that I took that landed that plane here and whatever. Probably some element of luck in that I got some immigration officer that looked at my application and said like, yeah, yeah, I get it. Like it's all about GitHub and code and repos and stars and whatever, but seems legit. Whereas I've actually heard really like bad nightmare stories of just like people catching the immigration officers on the wrong side, walk up on the wrong side of the the bed, their foot, whatever. And like, they just like will reject you until you die, whatever. So there's an element like, because I consider such a privilege to live here. And I think America is like an unfair advantage in many ways. It's such a great country to live in. And so that's an element of luck that I'm here. I feel, I mean, this place is so special. I feel that like, you could rebuild yourself from scratch. I don't know if starting completely from— imagine you wipe Memento style, you wipe your whole memory, you delete your network and you have to start over. You can still achieve greatness. Everybody here can achieve greatness. And I don't know that you can say that in every country in the world, right? And so there is an element of luck in where you're born, what countries you find your way into. And obviously there's an element of skill and merit and whatever those countries evaluate for. Those are the things that come to mind. On skill, I'll say I found my niche in the world at an early age and I deepened it. And in fact, my dad not knowing programming languages was actually kind of against the choices that I made because he would think, oh, but aren't other people doing Java or C#? That's right. And I was like, no, no, trust me, trust me, bro, on this one. Like, I'll go with this quote unquote toy language that if you, and again, this goes back to the MooTools thing that like, if you get really crazy with it, you can like evolve it into something that's far more powerful, which ended up being an amazing bet because JavaScript evolved a lot. It wasn't always powerful. It wasn't always fast. It didn't always have type systems. Then it became sort of, it became quote unquote TypeScript. And so that bet was like 10 years ahead of its time. Same with Socket.IO and real-time communication. I think betting on the web as an app platform versus as a way to transmit, you know, what it originally was designed for, which is hypermedia. In fact, one of the founding articles for the web was called Worse is Better because the web was designed to artificially, but like by design constrain its power so that it could scale better. And that principle is called worse is better. That principle kind of got tossed out of the window once you bring very powerful scripting languages on top of hypermedia. So I think skill set, just generally speaking, in making the right bets at the right time, which is why I think part of my brain has gravitated towards investing and angel investing and those things because I think with software, you have to think with choosing platforms and programming languages, you're an investor. You don't realize it, but whatever you choose to specialize in, you're an investor. I was an early adopter, not as a direct engineer, but I started investing early on in the Rust programming language. It ended up being a very successful off the bat. I invested in startups that were focusing on creating infrastructure with Rust. I focused on— I made some early bets on rewriting a lot of our compiler infrastructure to Rust, which is kind of a crazy bet to make. And now it's one of the most successful programming languages for writing infrastructure. So I encourage people to see themselves more as investors than as— you can call them like coders or just like purely scientists or engineers. It's like you're making investments. And what was the third one? Mistakes. I mean, mistakes all of the time. If anything, I would, one of the things that Vercel makes it possible that it was very hard to do without Vercel is containing the blast radius of experiments or mistakes. For most companies in the world, this is why companies a lot of time don't ship. Companies sometimes ship once a quarter, like that consumer electronics company I was talking about where their market cap, brand notoriety, et cetera, incredible. So they become fearful, they become paralyzed. They didn't figure out what Zuck figured out with move fast, with stable infra or move fast and break things, which is that you should actually have the mindset of a scientist that is experimenting with the software. You should, at any given time on v0, I can open it, I can open a special version that loads for the staff that works on v0 that shows me all of the experiments that are currently running, all of the feature flags, all the things that we're hypothesizing will be the next big thing, including models. We do a lot of A/B testing of model checkpoints and model improvements, et cetera. So what you want to do is you want to accelerate the rate of discovery discovery. And some of those things are mistakes. But what we do is we contain the blast radius of the mistakes. We say, okay, the first thing we do, going back to patient zero, we ship to ourselves. Sometimes it's painful because I'm giving demos to people and my version of Vercel is buggy as hell. Why? Because the team is shipping things just for me. And I love that because we're making quote unquote mistakes that we're repairing and learning and repairing and learning. Increasingly, agents will do— Vercel agents will do the repairing for you, which is fucking amazing. But embrace the error loop. In fact, the biggest thing that's made v0 more successful at writing code that's high quality is the error loop. When v0 spots an error, it fixes it itself. The very first version of v0 would just be like, okay, here's the error. And we would be like, what? Is this AI? It's just an error? No. Now we get into what we call an agentic loop. The agent notices the error, repairs, creates another error, repairs. The world is full of mistakes. I think your entrepreneurial journey or life, you'll make a lot of mistakes. So my advice here would be try to contain the blast radius. You don't want to make a mistake so big that— Naval has a tweet on this, like you go to prison. Like that would be the worst mistake you could make for your life or like your health deteriorates or your loved one's health deteriorates or whatever, like, but you need to embrace a certain level of risk-taking and mistakes. So it's amazing.

Guillermo, this was fascinating, man. It's like so interesting to hear your story from the beginning. So I really appreciate you making the time and coming to do this.

Thank you. It's super fun.

Thanks to our friends at Atomic Growth for helping with production and distribution.