Guillermo Rauch sat down with Sonya Huang and Pat Grady of Sequoia Capital for the Training Data podcast, published August 5, 2025. The conversation arrived a year into v0 — Vercel's text-to-app generator — reaching three million active users, and covers how AI is reshaping software creation, who gets to build things, and why the web itself is undergoing its most fundamental transformation since the move from static documents to interactive applications.
It seems like the coding market is being turned upside down. The frontend web developer segment looks like it's changing the most rapidly. What are you seeing?
Whatever LLMs are good at, people get really excited about. A couple of years ago when ChatGPT came out, one of the first things we noticed at Vercel was that ChatGPT is extremely good at writing React and Tailwind code. We got really excited because we saw that this was almost like a generational shift — not just a better framework, but something that opens up the top of funnel to every person on the planet. All you need is your natural language.
That inspired us to create v0, which is a text-to-app or text-to-frontend tool. Ever since, its growth has been astronomical. The biggest thing we've seen is that AI coding tools are now attracting what I'd call dev-adjacent profiles — designers, marketers, basically anybody with a keyboard. I sometimes make the joke: instead of yapping into your Teams chat, just yap into v0 and you're creating more value.
With v0 as it exists today, what are you most proud of, and what do you feel is not quite there yet?
Most proud of is the reach. Over three million builders, with high engagement and retention. Teams are using it. Fortune 10s are on the enterprise tier. This is an agent working out in the real world and providing value — not a demo. I'm also proud of how much we've extracted from building v0 and given back: we've open-sourced the underlying framework, the AI SDK. We recently released the v0 model to invite entrepreneurs to build their own agents on top of it.
The other thing is quality. Getting reliability right took real work. Anybody who used early ChatGPT realized it could write code that looked impressive, but getting it to the point where it reliably produces functioning applications required extensive fine-tuning. We trained a custom code application model that sits in front of the frontier model, entirely in service of reliability. You come in, you create a generation, it has to work.
“AI products have built-in success metrics. If you look at the first coding AI product, Copilot, its creator immediately had one gigantic metric: acceptance rate of the completion.”
You've been known for quality and developer experience for years. How do you systematize that? Are there metrics for taste?
One of the things that excited me about building AI products is that the feedback loop and the metrics come built in. If you look at the first coding AI, Copilot, what it did was propose a code completion. The creator of that product immediately had one gigantic metric: acceptance rate of the completion. You come back to the office the next week, it's 52 percent. You replace the model — what are we seeing? That compass is built in.
We have that in v0 in spades. The fact that people want to deploy applications into the real world is a very high signal of engagement. But the more subjective thing is taste — how do we embed all the best practices we've learned over years of creating products on the web? Little things, like making sure that on iOS the Safari theme bar color matches the background of your page. Most people wouldn't notice if it's off, but when it's right it's delightful. In the past you'd build frameworks and education around this. Now we can embed those learnings directly into the model.
How do you see Vercel's role in the world? How do the AI products fit into where the company was before AI?
Chapter one of Vercel: infrastructure on autopilot. We gave people the best possible developer experience — that was the Trojan horse for automating the cloud. Chapter two is the post-framework era. We automated infrastructure; now can we automate writing the software to begin with? With AI being able to generate code, we're opening up this frontier of automating it all, putting the human in the driver's seat from a creative perspective. The question we ask when you go to v0 is: what do you want to ship?
But it's broader than v0. As you look at all the other areas of our business — firewall and security, observability, the foundations of our cloud — we are disrupting all of those with AI too. Imagine an agent that instead of reporting that your workload is experiencing 500 errors can actually give you a pull request with the solution. That's the direction. The fully self-driving infrastructure. I think we're practically almost there, depending on the domain.
“The web is going from being static to dynamic to generative. Everything will be ephemeral. You won't even notice it's an ephemeral app.”
Why has ChatGPT been such an effective source of new customers for Vercel?
We shared recently that signups sourced from ChatGPT have been growing exponentially. When we created the AI SDK — you can think of it as the Next.js of AI — we noticed that if you ask frontier models what the best way to deploy a React application is, they choose Vercel a lot of the time. This is all of the training data, all of the writings and GitHub issues and solutions — it's chosen Vercel.
We had an anecdote at the AI Engineering conference in San Francisco: people came to our booth and told our colleagues that they'd learned of Vercel because ChatGPT told them to use it. This changes how marketing is done. You're going directly to your AI buddy to learn about the world. The kinds of content you'll write in the future are different — people are no longer doing keyword-based searches for certain problems, they're asking questions. I think it's a great outcome for the internet overall.
Do we end up in a world of disposable apps? What happens when everyone is creating lots of apps all the time?
I've always fought the idea that an app is a permanent thing. The idea that you download and install, that it's your responsibility — this is your puppy now, take care of it — that's insane to me. I'm Team Web, where everything is instantaneous. And I think it's even more ambitious than ephemeral apps. Everything will be generative just in time for everybody. You won't even notice it's an ephemeral app. Everything will be ephemeral.
We're in a race: if the quality of generation continues to improve — quality, performance, security, reliability — then it's going to be an unwinnable battle for the downloadable, installable, procurable software model. I'm Team Fully Generative Software forever. The web is going from static to dynamic to generative, and Vercel was already investing for years in that transition. We were working on streaming websites, on serving dynamic content just in time, which turns out to be exactly the infrastructure you need for LLMs.
In the world of AI, what's overhyped, what's underhyped?
Underhyped is easy: LLMs are so capable. There are so many great applications, so many agents to be created and shipped. Even if we froze everything as it exists today, the amount of value the world will get, the number of successful entrepreneurs we'll see building on these platforms — it is a genuine platform shift. It requires that you rewire your brain, which is hard to do. But people who are just being born into this and never learned the old world are going to see an enormous amount of upside.
Overhyped? I take such a long-term view that it's hard for me to find something truly overhyped. I've been bullish on crypto through multiple winters because ultimately all money will be digital and all software will be AI. If you take a very long-term view, it's very hard to say anything is overhyped.
“In the next five years we'll see the biggest transformation to the web in its history. Even five sounds like a lot — in three years, kingdoms will collapse and AI-native companies will rise to unprecedented heights very quickly.”
When do you think we'll get to the utopia of the generative web?
In the next five years we'll see the biggest transformation to the web in its interfaces, in the habits of people, in the number of creators participating. Even five sounds like a lot. I'll say in the next three years we're going to see kingdoms collapse — companies that were born on the internet but haven't been able to make those adjustments fast enough — and new AI-native companies rise to unprecedented heights very quickly.
The rate of improvement is so fast that whatever breaks the norm just takes over the world. You're already starting to see this with startups: the pace at which they're growing is unlike anything before. I think anyone getting into building software today is in an amazing position. They're leapfrogging the past generation. The past generation has their gray hairs and hard-earned lessons. But people getting into software today have incredible access to everything we've all learned collectively.