Static, Dynamic, Generative: What’s Next for the Web? w/ Guillermo Rauch

Welcome to Syntax! Welcome to a brand new episode of the Front End Happy Hour podcast. Welcome to this week's JS party.

Live from Ship Shape Studios, this is Whiskey Web and Whatnot with your hosts, Robbie the Wagner and me, Charles William Carpenter III. That's right, Charles, we drink whiskey and talk about web development. I mean, it's all in the name. It's not that deep. This is Whiskey Web and Whatnot. Do not adjust your set.

All right, welcome everybody to another episode of React Rung React and Ramblins.

I did it again. That's okay. Anyway, it's even better.

Yeah, it's not live stream. I could always re-record it later. Anyway, I'm your host, Charles William Carpenter III, here with my special guest co-host Aaron Francis.

I am standing in as special guest co-host and happy to be here.

Well, that's guest co-host. It's in the name.

I know, right?

And I'm here with my guest, uh, Guillermo Rauch.

Is that right?

New guy. We just met him.

Yeah, I'm listening to him.

Yeah, jump on the podcast and hang out with us. So Guillermo, or G as you may like to be called, for those who don't know who you are and what you do, mind introducing yourself?

I'm Guillermo Rauch. I started Vercel, formerly known as Zeit.

Yes.

Many, many years ago, I started the Next.js project. And many moons ago, I started, uh, open source project called Socket.IO and Mongoose. So I've been around the JavaScript block for a while and I'm really happy to be here at React Miami for the first time.

Yeah. Thanks for joining us there. Thanks for fixing my pronunciation. You did kind of snipe my— I was going to make the Zeit reference because I have been a user since back in the Zeit days. Thank you. I would now my stuff to the internet. It was mad.

We should give people a badge if they log in as an account that was creating during the Zeit days. That would be pretty cool.

That would be very cool.

Yeah. All right.

I think honestly, and I'm going to betray myself, I've never used React in my life, but I think I would get that badge.

Really?

Because back in the day, I think I deployed with Zeit.

Cool. Yeah.

I signed up for everything back in the day because I used to— I just wanted to get my usernames everywhere.

It's very important.

Yeah.

It kind of matters. Like, I want to be Chuck Carpenter on the internet everywhere.

If I'm not Raj G on the internet on everything, I freak out.

Yeah.

So we need an agent, an AI agent that goes out into the wild, into the internet, and just signs up for everything. Just squats on domains and handles.

Yeah, it's perfect.

It's not a bad idea.

I love it.

It's a good career option for you. You should like—

even if the whole for sale thing doesn't work out. Yeah, well, you've got a good plan B.

You can make millions of dollars doing this. I'm an ideas guy. No, I mean, frankly.

Yeah, yeah, no, I get that. I get that vibe from you. All right, so we start this show typically with whiskey, but special for this one, we're doing rum. And today's rum for this— yeah, I'm not sure if you're familiar with it.

Flor de Caña.

Flor de Caña. This is the 18-year-old expression. I don't know a ton about rums, but I do know that aged rums are a little bit tastier than your like normal mix it with Coke kind of.

Anything that's aged sounds amazing, right? Right.

I agree.

Yeah. Except for software. There's this quote of software ages like milk, not wine. Yeah. Have you heard it? I have. Yeah. It's the best.

Yeah.

The latest software, we want everything to be aged.

I agree. What about your podcast host? Do you like them aged? Because I'm not getting any younger.

I love them experienced in the arena for many, many years. You ask amazing questions.

Oh boy. Well, I'm about to underwhelm you, Seb. But here I am now pouring Seb rum. I would guess— I know this is a pretty decent brand.

It is from— here you go.

Thank you, sir.

Thank you. Jumping on Flor de Caña. I mean, sounds Cuban, possibly. Nicaragua. Oh, okay.

Yeah.

Product of Nicaragua.

Product of Nicaragua. Flor de Caña.com. Yes. We should look up how that's built. Mm-hmm. Yes. Are they using Next.js? Yeah. That will influence my business. If not, I just spit this, spit this out.

Yeah.

It's the barrier to entry is that even your spirits marketing site must be built in Next.js.

Yes. In fact, one of our big clients is Diageo. Oh yeah. They own a lot of, uh, alcohol brands.

About all of them, half them.

It's fun to, I mean, we're joking obviously, but every single company in the world is going through digital transformation, right? Like, yeah. This company's probably been around for decades and decades and decades, and now they have to start thinking about, well, like, what's all this direct-to-consumer thing? Are we gonna get this intermediated? Are we just gonna be this little icon or logo in the big shops like amazon.com? And how do we rethink and reinvent ourselves in the digital age? And now people are thinking about that in the context of AI. You know, I started my company thinking about developer experience and just being a geek, solving my own problems. But it's immensely rewarding that now we get to solve huge problems that we could never anticipate for every company in the world. I mean, that's kind of the power of the web. It's the power of React. We just talked about it on the panel. No one knew how big these things were gonna get. At the same time, we're only at like 1% of the web. Like if you even measure like how many websites and web applications are built with React.

Oh my gosh.

We're all in the, you know, hype, like, cloud Twitter X world of we take it for granted, but it's still relatively minuscule in the grand scheme of things. And AI is making it seem even more minuscule. You know, only single-digit million developers that know React. There's hundreds of millions that know how to, you know, use Microsoft Excel and things like that.

So, yeah. So the ubiquity that can kind of come over the next decades, it's like, there's a lot of opportunity to that.

Yep. And I think you're right. It does feel like on Twitter, X.com, the everything, right?

Market saturation feels saturated. Yeah.

But yeah, when you look at the real stats, you're like, wait, WordPress does what? We got a long way to go.

We got to keep going, guys. Or sometimes they're not online. Like, that is true. By the way, Floridacanias representatives are listening. Like, I don't mean to make any assumptions about their business, but like, imagine how many big ideas they probably have. About cool things they want to do with software or the internet. And the limiting factor today is access to technology, access to developers, recruiting the right people. Yeah, it's still— it's techy.

Yeah, right.

All right, before we keep getting too serious, let's remember the first part of this show, which is we're going to try this alcohol. And you may not be aware, and for anyone who's joining just on this episode, we have a highly technical rating system. It is from 0 to 8 tentacles. 0 being terrible, throw it out, spit it out. 4 being like middle of the road, you would have again, you wouldn't really seek it out. And 8 is clear the shelves because if I see it, I'm gonna buy it.

And every single one.

Have you already rated an 8 before?

Ever?

Yes.

Not here. Not here. Not today or whatever else. Uh, not in the rum arena.

Actually, I don't think I ever have. Could a rum even be an 8?

That is sure.

That was my question.

That was my question.

I don't think so. You don't think Not for me personally.

Yeah, not for me personally.

That's true. I don't think I would clear it and say I'm never drinking anything but rum. So that's kind of tough.

Yeah. So in Argentina, the most popular drink is Fernet con Coca. Con Coca.

That's right. Oh, I know.

I've had that. Well, tell me, because I've never heard of this.

No.

So Fernet is a liquor from Italy. Argentina had a lot of Italian immigration, a lot of influence in the culture, the language. My great-grandfather is from Italy. And so somehow became a thing even though it's like a pretty niche alcohol.

It is. It's an amaro and it's a dark amaro.

Yeah.

So it's a very herbal.

Yeah.

And then mixing it with Coke sounds insane.

The combination with Coke, not Diet Coke. No. That's actually like too much.

And it has to be the right Coke too. Like our American Coke won't work. It just doesn't taste right.

It's not as good. I'll say Mexican Coke is probably better. Yeah, definitely. But I'll tell you, the combination of those two things elevates Fernet. It takes it actually from a 1 possibly.

Yeah.

For me at least.

Oh, okay.

To an 8.

Yeah.

Whoa.

Yeah. It's in a corner.

It's a big swing.

Yeah.

I would say, Aaron, A, we should try it tonight. I've had it. I used to go to Argentina a bunch when I worked for National Geographic.

I have friends in Córdoba.

Oh, wow. Which, so, like, Córdoba likes to say that they invented the drink.

And there's a huge dispute. There's probably a Wikipedia— you know when there's Wikipedia articles on, like, disputes?

Uh-huh.

Yeah. The X versus Y debacle. Of course. That's, like, who invented this? And there's a bunch in Argentina with Uruguay. Uruguay claims to have invented yerba mate. I love Uruguay, but I mean, there's no fucking way. They claim to have invented tango too, that Carlos Gardel, our tango guy, like Maradona of tango, was born there. There's a lot of myths out there, by the way. Yeah, but you want to go on a podcast?

That's nice. Yes, that's another podcast spin-off. Let's just dig in the truth.

But anyway, so what I was gonna say is Fernet Cola can go to an 8. Because of the cola. Yeah.

So you like cola is what you're saying?

Yes, yes. And we're— this is pure rum.

Yeah, pure rum.

So it's gonna be hard to get into an 8.

Well, let's try it.

Let's find out.

Here we go.

Yeah, there you go. I smell it first, is just what I do. Hmm, it's a little woody smelling. I'm gonna prime it.

It is woody.

Yeah, I don't know what they age it in. It's usually used barrels, which is why you have to go up to like 15, 18, 20 years with rum to kind of get some depth to it. It's a little licorice in there. Am I crazy?

I just don't think I'm getting to an 8 on rum.

I just—

I'm not getting to an 8, but I'm actually surprisingly not on the 1 realm.

No, no, no.

Yeah, because do you drink a lot of hard liquor?

I'm above half for sure.

I'm about 4.

I'm in the second half. Yeah, I'm at— I'm in a 4 to 8.

We're narrowing. Okay, narrowing.

You probably put this in my head, but I have a slight, like, cola syrup flavor to it. I think you got a little, little syrup.

I'm gonna rule out 8. Okay. I'm gonna rule out 7. Okay.

Fair, fair. Sounds like you're going first and I'm here for it.

No, no, no. This is good. Keep going.

This is a 6.

Okay. Okay.

See, you have some context, Aaron, because you had the last one and you know how you felt about it.

So I think I'm gonna be so bold as to agree. I think this is a 6.

Yeah. Yeah. Yeah.

This was good.

I gave the last one a 6.5.

Oh, you went into the decimals.

Yeah, you can do whatever you want.

We don't care. I didn't know that was allowed.

It's allowed.

If you want to do floating point, you do floating point.

Yeah, you can get all weird with it. It doesn't matter. 6.25 stars. Totally fine. I feel like I like this one better. It's got a little heat to it, but less actual burn going down.

It's just smoother.

It's a nice sipper.

Respect.

Yeah, so I kind of feel like it was better than— I'm going to say 6.75 for me.

6.75?

That's high. Florida, Kenya. I won't say it.

Look at Florida, Kenya. They've cooked. Yeah, they have.

I am enjoying this.

This is going well for them.

Yeah, it is. Yeah, this is going to be a massive marketing opportunity.

Florida, Kenya, run.

They're literally hundreds or just thousands of developers.

Family legacy. By the way, the mount behind it almost looks like Mount Fuji, which makes no sense.

No, not at all. Vesuvius, maybe? That looks like Vesuvius somewhat.

Carbon neutral certified.

Yeah. So they're doing the right thing, I guess.

They're doing the right things for the environment. Yeah.

This is one big ad for Florida Candy.

Yeah.

I want to talk about this whole, you get a special Zeit badge. Again, kind of—

That's good content.

Y'all should absolutely do that.

I'm extremely grateful to the people that have been with us since the early days because, I mean, this is not to pat myself on the back because it came at an incredible personal expense, but startups are almost impossible. Yeah, startups in the infrastructure space even more so. Impossible squared because you're solving basically a conundrum, which is infrastructure is a thing that is mission critical. It has to be rock solid, it has to be trustworthy, it has to be secure, it has to be reliable. You have to be able to bet your life and the life of your children on this thing. Seriously, as the world, as we were talking about, like, becomes more more digital first, you know, during COVID I always tell the story to our colleagues that almost every single vaccine appointment emergency thing that had to be created, it was shipped on Vercel. There was no time. Yeah. And the whole healthcare industry had to become online overnight. Not unlike e-commerce. People talk about this in, in the, in terms of e-commerce and how it like, it grew like 3 years worth of growth in like a year, et cetera. Right. But it also happened with healthcare. So to create a platform that can earn the trust of so many people and that you can literally bank on. We also host a ton of financial services companies and crypto companies. So that could not have happened without the brave early adopters truly that bet on the promise of, okay, let's make this thing 100 times easier than it used to be, but it's not going to be perfect overnight. And so every time someone tells me, I was with you in the Zeit Days, I think, thank you because you had a huge part in building this. And so that, that really deserves a special badge. We're gonna make it happen.

I don't know. That's a fun, it's a fun idea and that would do very well.

Yeah.

On Twitter.

That would be great.

I think a lot of people would love that too. And, and the industry has had a lot of pressure to make people more of a generalist. And so you have less access to specialists who can create CI pipelines with Kube. Kubernetes and do all these crazy things in order to have observability and yada, yada, yada. Like a way for someone who's a JavaScript developer to get their projects online and then have that positive experience and bring it back to their employers and be like, look, we can, we can actually just do stuff.

Yeah. And opening it up to as many people as possible. So, in the early days, we used Kubernetes, which you just mentioned, as the actual control plane. And we discovered very quickly with our exponential growth that a cluster can only get so big.

Huh.

And we started running into operational limits that Kubernetes themselves had never seen because no one had opened up an entire service to the wild west of the internet where everybody could deploy. Yeah. You know, the smallest unit of deployment was the cluster. So you might create a cluster just for yourself, your company, your friends, your, you know, labs at your company, research teams, et cetera. Etc., but no one had created like the global cluster of the whole internet. Yeah. With that everyone can sign up to and start deploying with one command, which by reducing the friction, you massively increase the number of deployments that get created. Yeah. And then we had the idea of, well, what if every Git commit got automatically deployed? What if we made it worse?

What if we went, what if we pushed it even further?

Yeah. So everything about like running everything hot. I mean, that's what startups are like. You run yourself really hot. You run everything hot and the computers especially. Yeah.

How long was the brutal personal phase? Like how many years was it just like waking up and eating glass?

Every, every day is chewing glass, but not the kind where the systems are melting. You know, every company at every scale, even AWS, still has days where like, I don't know, this region had a, meltdown or this thing happened. So those are exceedingly rare these days, which is amazing. 'Cause like I always tell the team, the worst thing that could happen is something security related. The next worst thing that could happen is something durability related, like data loss. And the next worst thing that could happen is something availability related, which is like uptime.

Yeah.

So those things are exceedingly rare, but we don't want to downplay it as like the, the chewing glass happens because you're setting up for like bigger and bigger and bigger challenges every single time. Bigger customers. We just onboarded recently, I learned from our engineers, our top 3 most trafficked properties on Vercel are all new. Meaning like, wow. Out of like, they came out of nowhere and they quickly rose to the very, very top of the platform. And one of them, I, I'm not gonna name them, but they're like a super hot AI company. And every single time there's like an exponential increase in like usage or traffic or launches and releases and all these things. So there's always like an element of like fun and also like pressure that comes with it with this job. You know, at this point, like when you're an entrepreneur that's been in this arena enough, you're signing up for this.

Yeah.

Early days you're more like surprised about like, ah, my friend Nat has this really funny tweet about like, We didn't do this because it was easy. We did it because we thought it was gonna be easy. Amen. Yeah. So you're, you, the first few years are like that. You're like, ah, so there's that other aspect of this. Oh, and oh, and this gets harder because of that. And, but now like, I feel like there's no more surprises. No, that's good in a way.

So now it's just operational.

Yes.

You know, things are gonna be hard, but you can pull them off.

That's right. No more like, that's right.

What the hell are we even gonna make it?

That's right. That's right. And the other pressure that comes with startups that are at our scale is every successful startup in many ways is a one-trick pony. That's right. Like you wanna be focused. Yeah. You wanna be like one thing. The problem with that is that as you get bigger, the world wants more out of you, which is actually really fun. So the classic thing that happened with Vercel is people see that we make something in the cloud really easy. And then they're like, well, but can you make that other thing really easy? And that other thing really easy and that other thing really easy. Can you, can you make my data pipelines? Yeah, right. Like a lot of people kind of love the approach that we take on things like, well, I have this other hard problem at work. Can you also make it easy? And so the challenge becomes, okay, how thinly are you going to spread yourself across all the possible universe of things that you could tackle? If you're not ambitious enough, you're kind of going to fall behind. The world wanted you to do more. Hey, Vercel, like I thought you were going to ship like 100 different products this year.

If you do too many, you're like, hey, Hey, Vercel, like, are you just shipping slop?

What's going on here? Yeah.

So you're just trying to cover too many edges at that point.

Yes.

And that is an interesting force to deal with in its own right. I think my guess is that every entrepreneur struggles with this one on some level of, if I do too many things, I'm a jack of all trades, master of none. Yeah. If I only do one thing, you know, my growth, whatever, will slow down. I'll disappear. Deploy my customers, whatever. So that's, that's a fun one.

Yeah. And you're accepting a certain level of responsibility to your customers. You start to like absorb more and more of those areas and, you know, I'm now just trusting Vercel with handling my traffic, handling, you know, attacks and, you know, DDoS attacks.

It's a great example.

And then another one would be like around security, right? Cybersecurity stuff in general, not just DDoS specific. Yeah. So at what point, like, do you intend to sort of accept that responsibility for like knowledge and ownership into my project? You know, how much should I continually trust Vercel to do the right thing for me, or should I get informed and cover those edges myself?

That's, this is a great one because we live in a massive ecosystem, right? Like the threat model is nearly infinite. Think about what happened recently. There was a very popular project in the JavaScript tooling ecosystem that got pwned and someone deployed a crypto miner. Luckily it got discovered and sorted out very quickly, but npm itself is a threat vector, right?

Mm-hmm.

We go above and beyond in protecting our own contributions to the npm ecosystem, but the reality is my customers are using a lot of the things in the npm ecosystem. And I truly feel like I'm responsible over that because we are all here responsible for the ecosystem that we exist in. And so if I, if I go and tell a client in a bank, hey, this TypeScript thing is amazing. Next.js is amazing. Go all in. Deploy the future of your digital interactions with your millions of customers whose trust you've earned for 100 years onto this ecosystem. And they're like, oh shoot, I was using LeftPad. And leftpad deployed a post-install script and the post-install script like extracted all of our keys and, and we're not a business anymore. So I feel like when you're in this world of developer tools and infrastructure, you have to assume the end-to-end responsibility. Cyber attacks on the traffic side, meaning like volumetric attacks, DDoS attacks, et cetera. Yep. We've been fighting those for the 10 years that this company has existed. What's more recent and new is that we started to productize the visibility. Like literally today I got a tweet of someone saying like, oh my God, thank you Vercel for your firewall because it prevented these 3,000 threats and it blocked these 40 sources of traffic. And I remember telling the team, we have an internal principle at Vercel, which is seeing is believing.

Hmm.

We've been blocking those cyber threats for years. Yeah. But if you don't see it, you don't believe it. So for those that are in the audience thinking about building products that have an, quote unquote, an invisible component, 'cause I always think it's funny that infrastructure is invisible. So it's very hard for us to size up the investment of, quote unquote, raw materials that go and human resources that go into it. We're here in Miami, I was like admiring all these beautiful buildings, the skyline, whatever, and like, You can more or less say like, that must have taken like thousands of workers and hundreds of millions of dollars, if not billions of dollars in investment. Right. Infrastructure is invisible. In fact, we recently did the math of like how big our racks are inside of AWS. It's actually pretty bewildering now to think about how much physical space Amazon has automatically allocated for us with robots and people. Oh, but we never had to think about it and we never visualized it.

Yeah.

And so when you think about cyberattacks and protecting against cyber threats, the lesson for us was like data and knowledge is power. We have to make it super transparent to the end user what's going on. Yeah. And what's going on is actually pretty sick. People will like indie makers or solo entrepreneurs will build something and Vercel, you know, welcomes them with open arms because we love the idea of like getting, feeling the pressure of the people that need to ship. 100 times a day and want to extract every possible ounce of power out of the platform. But then those people that are building like their dream project get attacked by freaking like botnets. The most recent that I thought I heard about that we researched was a set of DVR cameras all throughout the United States that are super high bandwidth. Uh-oh. So they're constantly uploading really high quality video in real time.

Yeah.

And so you can get so much internet throughput out of those that when we were mitigating this attack, we were like, who the hell, like, harkens to this power? Yes. To attack, like, sometimes it's literally like a tiny website. Sometimes it's like an extremely strategic target. Yeah. But everybody— basically what we've learned is that everybody needs and deserves the same level of protection. Yeah. But the sort of like giving you access to the transparency of the data and the twisting of the knobs, that's more recent in the Vercel sort of journey.

Huh. So you talked a little bit about building product and some of the advice you have for people building product. I wanna go back to what you were saying earlier about you do one thing well and your customers are like, hey, do this other thing. Do databases.

Yeah.

Do WAF, do all of this other stuff. Do you personally, or does Vercel as a company have a framework for deciding like, hey, our customers are constantly asking us this. We definitely want to do it, or we definitely don't want to do it. Like, how do you determine— how you don't stay a one-trick pony and get passed up and how you don't spread yourself too thin?

Yep. I love frameworks, clearly Next.js, et cetera. When you grow a certain size of a company and business, you start working on the framework for the company, not just a framework for the developers. And so the, the, my job today is mostly, I still think a whole lot about Next.js and V0 and things like that, but I think even more about the framework to create a company that can absorb all that input and produce high quality output. It's like the render function of the products and experiences that we put out into the world. Yeah. So what are the inputs? The state are the things that a customer requests. Yeah. Innovations that we think up. And then what gets sort of filtered down and passes as props is the most important function you could possibly write. Too rigorous of a function will make it such that you're too much of a machine. You're not flexible, you don't take risks, right? But you do have to have certain filters. So one filter is things that we have used ourselves internally to great success have a way higher likelihood of being productized and seeing the world. Yeah. So a good example would be if I create a really amazing thing with D0.

Yeah.

I'm way more likely to greenlight the project because I'm like, hey, like if I see myself using this, at the time we actually created it into a meme, which was called no random acts of AI. We can't just like be excited about AI and put things into the world that we're not proud of. So we wanted to create a product that had embedded our taste opinions, that we used it internally, that we're proud of, that whenever we demo it to others— another principle is before something goes into the world, you have to be able to sit down with a customer or with a peer and you demo it. So Apple actually borrowed this from Apple because their internal mantra when creating the iPhone was like, we're going to demo every week, demos over memos. And that's how we evolve the product. Demo, demo, demo, input, input, input, demo, demo, demo. And so constantly being demoing and watching other customers, you know, you can call them design partners. That's like the trendy term in, in the Valley these days. People using the product and you watch them. Mm-hmm. Because sometimes we tell ourselves a lot of lies in our heads of like, yeah, the product is really good. The signup really works. The funnel is perfect. And then you watch someone do it and it's like that meme of the guy trying to drink water by licking the bottom of the glass. And so that's another one of the sort of like internal processes you can call them, or principles that we follow. The other one is, I borrowed this one from Satya from Microsoft. He calls it brand permission. Okay. I am not joking when I say that people ask me to sometimes build skateboards with Vercel logo or laptops. Or like I've heard product requests in the wildest categories, like, yeah, build me a forklift, a Vercel forklift, you know what I mean? And so I always think about like, all right, like I'm gonna announce it to the world, you know, Twilio also would do this, like the PR, write the press release today. Vercel announces the Vercel forklift to speed up construction worldwide. Like, I don't think you would have the brand permission. Your customer base would be like, What? Yep. I've been asking you for like Qs. Mm-hmm. And you should be doing forklifts.

Wait, wait. Tell me Qs are coming.

Yes. Well, I can't commit nor deny, but confirm nor deny, but this could solve a problem for me in the near future.

All we can confirm is they're coming before forklifts.

So that's all we can say for sure. That already makes me feel better. That's a good framework for anything about it. So people call this adjacencies. So it has to be something that is within your constellation of reach and knowledge and proficiency. Yeah, Satya had a great metaphor here of they've been working on productivity tools, Word, things like that. Things like DocuSign would have been a perfect adjacency for them. And perhaps sometimes they wandered off into doing weird things. Yep. And then customers have had that reaction. So I really like learning also from like the other like entrepreneurs and builders have been building the generational companies that you learn ultimately from a lot of mistakes. Yeah. Jeff Bezos also talks about, you know, if you're not tolerant of failure to some extent, you'll become so rigid that you realize that ultimately one success pays for 1,000 failed experiments. The loss of what people call like venture capital, which is like one startup returns the fund, is also true for your own investments. One good product returns from a, for 1,000 failed experimental products.

I think that's fair.

One feature, I've seen this a lot in our products, one feature or one integration. So like the GitHub integration with Vercel paid for all, every other integration. I've written so many bad integrations. I've misprioritized and whatever, but that one paid for everything. Yeah. Because everybody now goes to GitHub, they want to deploy, they install the Vercel integration, they never think about it again. Yep.

Yeah.

And that's another area where like I, in the early days, I kind of underestimated the power of integrations. And this is true for everything you do in the world. Like try to integrate with others, try to partner. You're not doing anything solo in this world. You're always partnering with the community, with the ecosystem. And that's why also I think React has been so successful.

I don't wanna make you dig up traumas, but what are some other like experiments that you tried and realized, whoops, that ain't it. And then are there any that like eventually came back to, to pay off?

Deploying Haskell and COBOL is funny one. Is that real? Yeah, yeah, for sure. In the early days, I think the common trap for entrepreneurs is more is more mentality. But the way that you justify it is not because you didn't listen to enough Steve Jobs videos. It's because the world pushes you to be ambitious. You look at public companies, you look at your aspirations, you look at your idols, and they have massive market caps. They support every language, they support all of it. And you might get tempted to be like, okay, I'm gonna win if I say yes to everything. And the version for that for us was that you're gonna be able to deploy everything. And I would literally spend time sometimes creating examples for those languages. I'm not kidding. Like we had a bunch of Haskell examples, COBOL, and part of it was to show the flexibility of the platform. But in some cases you're like, man, like you can just read the room. No one has that need. Just work a little closer to the customer. I think part of my job is to think ahead, of course, to look around the corner, like what's the next thing? Try to live in the future so that I can bridge that gap and be the one that takes an L for the team sometimes. And you know, Silicon Valley is like that. A friend of mine was saying yesterday, like Silicon Valley's like a big lab and everybody kind of lives in the future. We have way more self-driving cars.

Yeah.

But I was telling him this funny anecdote of when I arrived in Silicon Valley, I was an early user of a lot of the things that today are some of the most successful technologies. I was an early user of Uber. It was called Uber Cab and it was only black limos. Yep. I was an early user of Airbnb. I lived in one of the first listed Airbnb units in San Francisco where the co-founders had stayed to eat their own dog food.

No way.

So talk about using your own product. And I was an early customer and like super fan of a service called Washio, which would, as the name implies, wash and fold my clothes.

Oh yeah, I love that.

At an unreasonably low price.

Wow.

With a huge amount of like low latency, like it would take no time and it would perfectly iron my clothes and like leave them. Sometimes it would even take them up to my apartment. And I would just like have to like take them from my door to like my closet. Like my life had been perfectly automated, but unfortunately that category didn't work out. So this is the concept, right? Like you try to live in the future.

Mm-hmm.

Yeah. You try to live in the future where every luxury and everything that you would want to automate that sounds like you're watching The Jetsons.

Mm-hmm.

Is possible. Yeah. And then you work into, okay, how can I bring this to reality to everybody in a system? Sustainable fashion. And so that's kind of the attitude that you have to take. But if you position yourself way too out into the future, if you try to like swing way too hard, you might miss. And so that's kind of like the meta lesson in that kind of like you have to balance, I call it today, you have to balance ambition with realism.

Yeah.

And I would err sometimes on like, don't underestimate how big a small perfected thing can get so that then that becomes your launchpad into the next small perfected thing and the next small perfect. And a good example, I think today, and I owe a huge amount of kudos here to Lars Graml and Jerry Palmer and our team. AISDK started out as this really small wrapper and multi-RTTO for web streams are really freaking hard and all of the LLMs stream tokens. And we started out with like, you know, let's create a little toolkit to like literally wrap these LLMs. And now it's becoming an instrumental foundational piece of the AI ecosystem. You know, it might get as big or if not bigger than Next.js itself. And it started as a small thing that we were gonna perfect, but we did have all these questions of like, all right, that's an adjacency. Do we have brand permission? Can it get big enough? You know, are we gonna distract? Is it worth investing? So like, just because you're making progress and you get bigger and you have resources and you have cracked colleagues, et cetera, it doesn't mean it gets easier because you have all this like fundamental philosophical questions.

Mm-hmm.

You can always push it further and push it further and push it further. And is that gonna go over the edge?

And then there it was. Yes.

Yeah. This just in, whiskey.fund is now open for all your merch needs. That's right, Robbie. We're hearing reports of hats, sweaters, and t-shirts, as well as a link to join our Discord server. What's a Discord server? Just read the prompter, man. Hit subscribe, leave us a review on your favorite podcast app, and tell your friends about our broadcast. It really does help us reach more people and keeps the show growing. All right, back to your regularly scheduled programming.

If you had to give this past, let's say, 5 years of Vercel an era, what would you name that era, and what do you think What do you think the next era for Vercel is going to be?

That's a great one. I think the web has gone through a couple eras in rapid succession. I call them like static, dynamic, generative. Okay. Static was, and by the way, this is a weird one to think about because the web started out with like full dynamism, CGI bin, put a Perl script in a folder, plug the human directly into the machine almost. Telnet, SSH, pipes, Bash, Unix. But to scale the internet, we made it static. The web was starting to go into a couple directions, static site generators, Gatsby, Jamstack, et cetera. Right. And cache control, CDNs, and then client-side JavaScript. We proved to the market that you could have the luxury. This is kind of like the Wachio example or like the Uber example. I wanted really badly to democratize the technologies that the Googlers and the Facebookers had access to. As a kid in Argentina, this is a wonder of the web. I had been able to reverse engineer how Google Search worked, how Meta worked, how Amazon worked. And I found this common pattern of like, these are highly dynamic personalized systems that use a lot of server-side rendering and are highly scalable. Of course, they have billions of users. So the last chapter of Vercel was, can we give every single developer on the planet that luxurious developer experience that before you had to get a job at FAANG to get. And so it was the transition of a web that went from static to dynamic or hybrid. The future of the web is not that you're gonna get data from a database and you're gonna permute it into some kind of markup. Of course it's gonna continue to happen and there is a lot of value in, in doing that. But now we're starting to democratize the other aspects that made Google special. I even call this the unbundling of Google. Google got really big and popular by basically reading your minds. You know, you started typing in an autocompleter. This is another thing that got me extremely freaking like JavaScript built.

Yeah.

When Google launched Google Suggest, you would start typing and then they tell you the future. What the hell is that? In fact, I wrote this article that I referenced today in the panel, the 7 Principles of Rich Web Applications. And my last principle that was starting to be future forward, I wrote this in 2014 thinking about the next 10 years of Vercel, right? And to your point, this is kind of the chapter that we're now completing. The last principle was called negative latency. So the only way that you can get negative latency is either by creating a time machine. We might get there someday, you know, reversing entropy and faster than light communication and wormholes and whatever.

Like maybe let's do forklifts first. That sounds hard.

I don't know if we have yet brand permission for like Eisendrosen wormholes.

Yeah.

But the other technique is machine learning. And so the unbundling and democratization of Google is that all of these technologies that they invented, like transformers and generative models and all of these things, are now APIs that anybody can call. Like, truly everybody can cook. And so the web will become generative in a number of ways. One is when people want to build for the web, they're gonna be able to generate all of that code on some level. And it's the luxury of sometimes being able to abstract yourself away a little bit and, and think about like 10 years in the future. It is kind of crazy that we all got accustomed to memorizing shit, looking it up on Stack Overflow, and then typing it manually.

Ah, the good old days.

Yeah, exactly. Like, you know, we, we're doing a lot of that still, but it's kind of wild that we thought that the way of the future was like typing fast. When I was a kid in Argentina, there was this software that my dad got. Everything in Argentina is soccer-themed. Spoiler alert. So we're here in Miami, Messi, etc., right? The president of Argentina right now. That guy. Yeah, exactly. Okay. National hero. Our president won not only because of his new ideas and whatever. He used to be a goalkeeper. I did not know that. I didn't know that. In order to sort of achieve cultural dominance in Argentina on some level, you need a sports, a sports background rooted in soccer. Huh. I don't think it's as true in the US, although there is a social status that comes from like, were you awesome at football or baseball or whatever?

I could kind of see that because we have no like modern era massive wars. But if you look at some presidents, They were war heroes.

That's right.

That would definitely help them. So you need to be perceived as some kind of conqueror. Yes.

Hero. Yes.

To the people.

And so the software that my dad got me was Campeón del Teclado, the Keyword Champion. Yep. It was a typing game and you would score by typing really fast. Nice. And I crushed that game. I think, I think my dad was smart there to like, he found something that like had overlapping interests.

How did he do that? And I get my, kid computer literate.

Yes.

Without making it boring and whatever. Genius. Yeah.

And, uh, he tried it again with my sons and it doesn't work. No. It's like, dad, first, they live in San Francisco. Yeah. They don't know what soccer is. No.

Yeah. You're bringing shame upon your nation by that also. Yeah. But they don't know what soccer is.

That's true. 'Cause you, you don't have an MLS team, right?

Yeah. They, well, we have a San Jose.

Oh yeah. First Earthquakes. Yeah. That, which is an original MLS side, but they've gone away and come back a couple of different times.

If we were in Miami, I think they would be into that.

Oh yeah. I mean, who wouldn't?

I, I blame not myself. I blame our surroundings. They're, they're more into, well, I mean, they're into a lot of like tech stuff too, but other sports of course. So anyways, typing really fast, having good memory. Yeah. Good recall, pattern matching were the fundamental skills of how do you get really cracked at, I mean, not, not the only skills, but a contributing skill.

I think it's useful.

And so now we're generating a lot of that and it's, I think I always call out to people and I remind myself daily, you know, you have to not become too attached to your ego in the age of AI, of this skills, the individual lanes that you developed proficiency in. Yeah. Because if you get too attached to that, It's just, we are one model generation away from being like—

Absolutely crushed.

Yeah. So then the future of the web is more people contributing to it, more generation. And also the other like science fiction, but I think we're gonna make it real, is that the web itself becomes fully generated. That the next time you return to a website, there is a new version that within certain guardrails, because we have to be careful about it, you know, like Skynet 2.0 and whatnot. Yeah. Within certain guardrails, the web should become fully generative. Yeah.

I mean, the, the con— actually, I worked for a company called Acquia years ago and worked on a, a digital product that used like marketing segmentation to have dynamic content that was individualized. So that concept has been around for a little while. And so why wouldn't that just continue to evolve and probably get a lot faster and smarter with things like AI?

Yeah. And you know, you could imagine a world where v0, when you ask it for, well, create me the landing page for this, it's not just creating like the one-time iteration of that. Yeah. But it's embedding in the code the runtime v0 component that behind the scenes knows all of these technologies that us nerds know about and like ISR and partial rendering and all this junk. Yeah. That people won't have to know about, will leverage those foundations, but then with the, you know, marketer, business person idea, person will get to do what I think they've expected the web to always be able to do. Yeah. Which is, oh, can we launch 100 permutations of that? Or can it just get better based on my feedback or my customer's feedback in a more direct way? Hmm. So yeah, I, I think the TL;DR is that the web gets easier and more accessible to more people and ultimately will realize, I think it's, it's ultimate potential. I think maybe to getting even more philosophical here, like a lot of people, and this is not the world that I would want my children to grow up in, you know, political issues about blocking TikTok aside, we've become a consumer society. Yeah. We consume more than we create. For sure. And I think it's our responsibility as creators, DevTools makers, infrastructure platform makers to turn that around. Let's turn our kids and their grandkids into builders. Maybe not all of them. Like I, You know, like I don't fully subscribe to the marketing, like there's gonna be 5 billion developers, probably not, but there should probably be 10, 50, 20x more than we have today.

Yeah. Yeah. I could definitely see that being the case. We're gonna continually evolve and as long as we are giving the ability for our next generations to actually learn the foundational information to like, you know, champion the fast typing, which, and that's not what you really need anymore because now you're doing it in a different way., you know, a speech-to-text way, but you understand what you're saying and how it'll translate that. I think that's an excellent future.

Yeah. I had an amazing trip to Japan recently. Speaking of Japanese whiskeys. Yeah. I went to Okinawa and I went to the Abacus. Did you see Miyagi? I did not. Oh, okay. But I'll follow up. I actually have not been to any martial arts or sumo. I've been wanting to go, but not yet. Okay. But anyways, the, I, I went to the Abacus school. The abacus is the first computer that humanity has created to basically automate in a primitive way, but in an effective way, arithmetic, right? You can do math that is much more ambitious than what you could do in your head. Yep. And we watched these kids that were literally, man, like just freaking alien. A 4-year-old doing like addition and multiplication with like 9 digits. Wow. Like he watches a, a screen flash and he does abacus in his head. Oh, it was hilarious because they had the abacus on the table. Yeah. But they didn't use it. No. This is like, like out of an anime, you know, like they play basketball or whatever and like they do like alien things to it. This is it. Okay. Because the abacus is sitting there and in their heads they're doing like mental virtual abacus. Yep. And the lesson there and the reason we were there was, look, that is an important skill. Like why wouldn't you want to, go through the endured pain that these kids have gone through. I'm, I'm a huge fan of that. Yeah. But at the same time, you have to have the realization that that's not gonna get you all the way. Mm-hmm. We have calculators, we have AI. Yeah. We have not just bicycles of the mind right now, we have freaking rockets of them. We have reusable rockets of the mind. Yeah. And so funny enough, Okinawa is the least privileged prefecture in Japan, so we had an opportunity to sort of bring ideas from Silicon Valley and inspire them with like complementary skills. Yeah. You don't just want to become that one thing, right? Yeah. But it was just amazing and it inspired us for what the future of computing will look like and how it'll transform humanity.

At this point in Vercel's history, what are you like, what is driving you? Because like from the outside you've made it. Right? I know you don't. I'm sure you don't feel that way, but from the outside it looks like, all right, he's made— he's got a big company, like they're crushing it. You've talked a lot about democratization of the web, which is pure, noble, and right. And that may be the answer. Is it that? Is it you get a high off of the framework, building the framework that is the company? You know, you're a big framework guy. Is it seeing Vercel IPO, or is it something completely different?

So one of the amazing things about building a company like this is that there's never a dull day. As I said, there's never a day that's not hard. There's never a day that you're like, oh, I made it, I'm so relaxed. So in case people in the audience were worried about that, like what's awesome is you get more reach, impact, etc., and then you get tougher problems to solve, higher ceilings to break through, and other kinds of scale problems. As I mentioned, you go from this problem of like the framework for the tech to the framework for the company. And you become more ambitious in the process too, right? Like new doors open. It's like a fun video game in which there's more bosses and things like that. Along the game while you're struggling, like almost like when you play like Pokémon on Game Boy, there's a lot of people that help you along the way. There's the people that like cheer you on. I remember playing Pokémon a lot. It's like, there's always a guy in those RPGs or whatever that you talk to and like they have nothing to add but cheer you on. And so there's, there's that, there's your mortal enemy of the day, you know, like whatever, Matthew Prince or whoever. And like you, you battle him here and there. Yeah, but you're always in a journey in a good competition market. In capitalism, you learn from your rivals and you grow each other and whatever, and ultimately you pwn them if you're Ash Ketchum, ketchup, whatever it is. And so there is the competition that I really like. There's the fact that you go around the world and you make a lot of people happy and they're like, The stories that you learn about the company are fascinating. At scale, the things that people will tell you, like, this is a funny one. Like we hired a, a great product manager at Vercel and he told me that his entire career was, was made by, so the funny thing is my last name is Rauch and there is a little town in Argentina called Rauch. Oh. In which there's only cows. Hmm. Like every time I would pass it in the highway, it's like, I was like, whoa, cool. Dad, is that ours? And he's like, no, there's nothing there. And maybe your grandfather had it one, one day. But this guy had grown up there. I didn't know this. And he wanted to get out of the life of like the rural Argentina, et cetera, struggle. Yeah. And he went to college for computer science and his big breakthrough was he started learning Next.js. Mm-hmm. And he started contributing to the examples. And a company in Germany found him because they couldn't find Next.js developers and they were like, hey, let's go to the committers. Yeah. Who's been contributing recently. This guy had made an open source contribution. This is also like full circle to the power of open source and like a contribution that you make to a repo might change your life. Yeah, absolutely. So he went from Rauch, Argentina, which by the way, sounds dope, Rauch, but it was very poor to being in Germany with a work visa and a new life and whatever. And eventually he found his way to Vercel. And, you know, hearing this story is just like, just so amazing. And it was all of because all these threads connected. And so that's another thing that's like so amazing. All this emergent things that happen, things that people deploy in the platform. We've done so much with, there was recently massive floodings in Argentina and all of the disaster relief efforts that happened that we sponsored on the platform that people could create. Projects and apps to like help people. Actually, that one is one with vZero. Like they, they had to move so fast. Mm-hmm. To create this like interactive map with real-time data of where people could get help and they had to bytecode it. We keep creating platforms and you keep seeing all this emergent phenomenon and it's just interesting and, and fascinating and, and hard, but, but rewarding. That fuels the fire.

So it makes me wonder, 'cause I don't know where that town is. What town are you from?

From? It's called Lanús. Okay. Which is not short for Lee Anus, not to be confused. Oh, okay. So it's, it's in the outskirts of Argentina. It's famous for having been the, where Maradona was born.

Ah, your other national hero. I was gonna say, really the primary one, I would say.

And he was born in an even rougher part of town as I was. His didn't even have any paving whatsoever.. And so there is this famous video of him kicking a ball that he had assembled himself with like used worn out clothes. Like he didn't even have money for, for a soccer ball. So we grew up there and then eventually moved to the city. Yep. Because I got into a high school that basically like kind of we're talking about like open source. It's like a launching pad for a lot of Argentinians that if you're willing to study really hard, this school is free. Oh, you have to like win a competition almost to get in. And then I actually be, I got so distracted by Mootools and open source React and all these things that like I quickly became the worst student, but a good open source contributor. We'll allow it. It turned out okay. That became the next launching pad. So kind of like the Rauch story, like the open source is a global launching pad. I hope that becomes a legacy of our companies. Of everyone that's here, React, Laravel, everyone's been participating in this. You sometimes don't know how big of a launchpad this thing is becoming for people in other countries, especially. Yeah. Where they don't have a clear path to universities. They don't have a clear path to jobs. They don't have a clear path to mastery. Yeah. And, uh, and open source is one of the big channels of our generation. Yeah.

It's a learn by doing platform where we lack things like apprenticeships and whatever else.

That's right.

Which I think would be very beneficial, especially in, you know, our career paths. But you know, that is open to anyone and the quality is there. And if you have a good community, then you're getting good feedback and mentorship through that as well. Now I've got another important follow-up question to my previous question, which is Boca or River Plate?

Well, I have to go with Lanús. Okay. Because it's actually, they punch above their weight. Yeah.

They're so good. I like racing.

So, you know, that would've been, but Argentina has all these clubs. Like that's Racing Club. Yeah. And River is River as in river. I don't know why they're like, one gets the Spanish accent and the other doesn't.

It's strange. And I'm sure there's some historical context. And like you said, Argentina has had a ton of immigration into it. Yes. And so there could be a good mix there.

So my dad had determined that the household was Boca fans. Okay. To, to add some challenges for us in elementary school because we're like, wait, why are you not with the rest of the town? Yeah. No, we're Boca. We're special. Yeah.

But the future, you were gonna be in town eventually.

River, they're called the millionaires. They have the most fancy everything. Yeah.

It's the racetrack thing or something, right?

Yeah. And the part of town is beautiful. Polo matches. I actually recently looked up why they're called the millionaires. It's something with immigration. Yeah. They, Lucked out. Oh no. And they lucked out with exporting a few soccer players. Like, imagine if you had landed the Messi pass to Barcelona, but not when he was 14, but when he was like 21. Oh, right. You would have made generational money for like 10. You would have owned half of Argentina. Yeah. So I went to River Stadium last time I was in Argentina and it was just incredible. Like everything was pristine and the show, I actually highly recommend it to anyone that visits Argentina. Go to the River Plate Stadium and watch a live— like, if it's like a very hot match, very contested, even better.

Yeah, I was going to say it's also safer. I've been to the Bombonera and people are like, what do you do? It was daytime and the police like rolled up on us and like, what are you doing here?

Yes, you should go. Yes. 3 blocks that way. Yes. Yeah, it's gotten so much better in River compared to Bombonera. But it is funny that to get better, The ratio of citizen to police officer was absurd. That speaks to the potential of the riots because the ratio was like 1 to 20. Yeah. Yeah. So they get passionate.

Well, passion is a good thing.

Passion is a really good thing.

All right. I've got one last off-topic question for you. On topic, but slightly off-topic. I know you and Taylor Atwell are friends. What is your favorite thing about Laravel?

Many things. One is, uh, about PHP in general, which is a huge inspiration for Next.js because there's two brilliant things about it. One, which I think some people will dispute, and one I think, which is objective. The objective one is that the lifecycle of PHP in which a request comes in and that determines the computation unit is what eventually AWS Redis covered with Lambda and the concept of a function and the invocation trigger being the request is a very meaningful bit in the huge entropy of distributed systems. The fact that the request can inform load balancing and placement and scheduling of systems, that it basically tells you something needs to be done. And when this response completes, we can do cleanup of sorts, so we can do deallocation of resources. Yeah. Is the right mental model to think about computation in the context of web services. Hmm. And PHP, because it was focused on web services and by the way, this is why focus is so important. There is PHP CLIs. I've seen the artisan stuff, et cetera. But the fact that they focused on web services actually was made it so successful. They actually nailed the computational framework. They were like, okay, if you want to use PHP, the first thing you need is mod-php and Apache. As opposed to almost every other programming language and runtime environment that said, well, what do you want to do? If you want to do this kind of system, you have to run it as a daemon. And here's 200 options with init.d and systemd and Kubernetes and whatever. So I think PHP picked the right battle at the right time, even ahead of their times in sort of like that request-oriented invocation model. And the other one that's more controversial because I think some people don't believe in it, but I'm obviously drinking the Kool-Aid hard for 10 years now, is the file system routing.

So like, I love file routing, by the way.

I love file systems. I feel like I'm on the verge of being vindicated in the form of LLMs. So LLMs are becoming very good at file systems. In fact, one way that LLMs have been able to agentically distribute work and explore systems and not saturate context windows has been by using files. In fact, O4, O3 mini, whatever came out yesterday, if you use Codex CLI, it's been post-trained to examine file systems and use ls and grep and find and all of these things. So file systems are actually a very meaningful part of the infrastructure of the digital world is what I would say. So kind of like Guillermo Rauch had to get good at figuring out streets and stoplights and turning signals. It couldn't just like say, YOLO, we're just going to do whatever.

This is the shortest path through your house. Exactly.

Exactly. Do that. Yeah. That's the file system today. And, uh, and that's why I'm saying it's been vindicated because when I started Next.js, the inspiration was PHP, but it's a gift that keeps on giving. Yeah. Is the gift that like you open a project and you can reason about it. You can find the conventions, you can explore the routing system with opening up the folder. And so that was a huge inspiration from PHP that you would throw in a PHP file in a folder and it would immediately become routable. Yeah. I mean, CGI-BIN had also done it, of course, but, uh, for me, the inspiration was PHP. Yeah. And now on Laravel specifically, I feel like it's very similar to Next.js where I think it proves that runtimes are too primitive. Of course, they're immensely valuable and necessary, and we wouldn't be here without Node. I love what Bun and Deno are doing to push Node and push the community forward and in their own ecosystems. Yeah. But I find like the sweet spot is the framework. Mm-hmm. I think you reach the highest number of people and you just give them the most amount of value. So obviously kudos to Laravel for finding that same sort of balance or place where they could add value in the BHB ecosystem. And also the other, I think this is not gonna be surprising to people because it's so obvious that it's created a very complete story. Mm-hmm. And we're talking about, you know, how enterprises ask us for more features. They expect us to do everything for them over time. And I think that's great. I think you're gonna find that customers over time want complete solutions, especially the more mature they are. I think when you're more in the experimental phase of the innovation curve of a community, the community wants, oh, give me this piece. I want to experiment with this other piece. If you think about React, think about how the community was breaking it apart into all these things. Here's the state manager. Here's the CSS in JS solution. Here's the router. Why? It was immature. And not because those pieces are not necessary, but like the next 100 million people are probably not going to be interested in having 12 different options for how to do CSS in JS, right? And so this is why I think we're now entering the stability phase of the ecosystem and there's going to be more complete solutions.

Yes, please. That'd be awesome.

It is what people want as well. And I think Laravel has done a masterful good job at giving you a very predictable, stable set of tools. Obviously more on the backend side. Sure. Clearly. But it's, it's an inspiration for us in that sense as well.

Cool. No, that's really, I, I didn't know that PHP played a part in the inspiration for Next. That's very cool. I do like PHP's create the universe and destroy it on every request. It's awesome. I love it. It's very easy to reason about. It's very reactive, right?

Like React. And Jordan was just talking about this in the panel. React took that approach of like, there's some state over here, let's create the entire virtual DOM or even the illusion that you're creating the entire application every single time. And I find that that's the model that ultimately you'd be surprised about how frequently things get actually start over from scratch. Just totally in the world of distributed systems because things are always failing. And they have to be recreated from scratch. Yeah. You don't want to piggyback on the accumulation of state. In fact, one of the great things about functions is that they don't make assumptions about having been running for 2 years. Yeah. And accumulating a lot of caches. It has forced Vercel to do an enormous amount of work into making things fast when they're cold. Yeah. I would always tell the team, I love that we have this challenge. Because in the real world, things do get cold. You know, there's traffic spikes, there is, you know, during the weekends, most of our customers during the weekends like lose a lot of their traffic. Yeah. And sometimes they have unpredictable spikes of traffic where like, oh, oh, I didn't tell you Vercel that I was gonna do this Super Bowl ad that I paid $7 million for. Yeah. But you told me they were gonna scale and yes, of course I'll rise up to that challenge. And that means reconstructing the world from scratch in many, many different ways, failover. Mm-hmm. You know, at scale things are constantly— the, the CTO of AWS always says, you know, ultimately something is failing all of the time. Right now as we're sitting here, some computer just died. Oh yeah. There's also the categories of failures that our, our infrastructure engineers have trained themselves to create monitors for, which is like unpredictable degradations. Mm-hmm. Which is worse than like, I, I'll take a segmentation fault, I'll take a, computer that stops responding over a computer that starts being, it spikes up to 100% CPU for no good reason. And at the scale of the number of computers that we're running, we don't even have the bandwidth to investigate it. We just need self-healing systems. And so embracing the wilderness of the universe is a fantastic engineering principle that I think, you know, runtimes like PHP kind of came into from the early days.

Mm-hmm. All right. Now it's time for the most important question of them all. Let's say you couldn't work in tech. There is no tech. What would you do otherwise?

And you don't have to have the skill now.

It can just be an interest, a passion, or whatever else.

Yeah. So when I was a promising young student in Argentina, because I had these two faces, like the dark side and the, you know, almost like competing priorities between my dad and my mom. My mom was a chemical engineer. Super cracked at math, loved academic recognition. My dad almost like wanted to ignore his academic achievements. He became an industrial engineer and he said, you know what, software is the way. If I could go back in time, I would have not gotten this degree. You know, almost like self-deprecating in a way. Yeah. And so I started my life as more on my mom's side of, you know, winning the math contest at school. School and studying really hard. And, and then when I was about 13, 14, I went off the rails really badly. But I had almost like a fork on the road. So one of the dreams and visions that I had is in Bariloche. Wow. Argentina, the south of Argentina. By the way, if you come to Miami, there's a Bariloche Alfajores place here that I bought alfajores that I took to the DAX party last night.

That's what that was. Yes. I saw you walk in with that.

I saw that and I was like, Bariloche, is this from— Bariloche. I'm like, so I could have ended— it's either go north to Silicon Valley, like really north, or go very south. Argentina is a very long country, to Bariloche, or the area where there's an institute called Balseiro, which is one of the most reputable nuclear engineering places in the world. Yeah. Okay. In fact, a friend of mine who's now the chief of advisors to the current president in Argentina is a nuclear scientist from this institution, and we're recently talking about it. It's like a small world because he also had kind of a fork in the road. He went that route. I went the tech route. And so I still long for it, to be honest, because in order to get to where we need to go in terms of becoming a multi-planetary species that has superhuman intelligence, we're gonna need unprecedented amounts of energy and compute. Mm-hmm. In fact, there's almost an equation that has been implicitly discovered, which is the transmutation of energy into intelligence, right? Like if you think about test time compute, it's becoming very, very clear that it's not that we can't have unbounded intelligence. Our limiting factor has become energy. Yeah. And so nuclear is gonna take us there, fission and fusion. You know, there's another parallel universe. Yeah. In which Rauch G is studying, is, in nuclear science and trying to like come up with a reactor of sorts. And I think there is also like an intersection of like, you know, how do we get more out of improved designs of computers? Yeah. And so I think parallelization is clearly the way. I think math models and vector math and GPUs, you know, there's this Austin Karpathy tweet about like, how weird is it that intelligence is being created from gaming cards. Mm-hmm. Like we would buy GPUs to render Grand Theft Auto, sick FPS. Yeah. Or Call of Duty. And now that's behind the intelligence age and the intelligence revolution. But I mean, a big part of that is the great parallelism that comes from like these plans get put onto the GPU grid. Yeah. And so uncovering new forms of highly parallel computation in quantum computing and you know, newer designs for ASICs and GPUs is one of my other big interests slash parallel lives.

I gotta say that's a way nerdier answer than I expected, right? It is.

Most people say like, I don't know, I'd be a painter or a farmer.

Right. I thought he said, when he said Bariloche and he talked about the plane, he was gonna say, I wanna, I would've been a base.

There's like an, there's an implicit lesson here that he's like, if I didn't create this massive company, I would go be a nuclear scientist and create a reactor. I think maybe some people are just built differently. Maybe that's what it is.

React, there's a connection into React. Like React is like this like reactor, right? Mm-hmm. Uh, it's all connected in the universe.

Yeah. You can't escape it. I would ask you if there's anything you wanna plug or where people can find you. I'm pretty sure by this point they know where that is.

But say it anyway.

Where can people find you? Rauch.io next. DMs are open. There you go. Uh, always happy to talk to people.

All right. Perfect. And thanks for sitting down with us. Uh, thank you. Any parting thoughts or all good?

No, I mean, had a lot of fun. I'll be back as promised.

Yep. Yeah, we'll make it happen. Don't worry. Otherwise I'll chase you for years.

You know how this goes. All right.

Chuck never forgets. Thank you for doing this.

No, that one sucks.

Now hit one. That was awesome.

Sweet.

Hello. You've been watching Whiskey Web and Whatnot recorded in front of a live studio audience. What the fuck are you talking about, Chuck? Enjoyed the show? Subscribe! You know people don't pay attention to these, right? Head to whiskey.fun for merch and to join our Discord server. I'm serious, it's like 2% of people who actually click these links. And don't forget to leave us a 5-star review and tell your friends about the show. All right, dude, I'm out of here. Still got it.