Okay, welcome. G, welcome. Good to have you here. I'm sure you need no intro. Probably most of our customers as well use Vercel, but I'm going to give you an introduction. And if you use Supabase and Vercel...
Yes, that's pretty dominant. We're in the right room.
Yeah, I'll give you an alternative introduction. So, you're the author of some of the most popular open source tools—we care about open source—socket.io, Next.js, and Mongoose, which is one of the most popular ODMs for Node.js. And so that's what I wanted to ask: as someone who spends so much time with Node.js, how do you have such a high tolerance for pain?
Well, that was the thing that I was trying to do with Mongoose. I think the barrier to entry to every database is the developer experience. I think something that you guys have done magnificently is you didn't just give people Postgres and it's like, 'All right, here it goes, choose your instance size.' I'm checked out. I think the modern database experience has to be full stack, right? You have to give people the driver, the higher level developer experience, the examples, the integrations with frameworks. We're just chatting backstage how to further improve the integration between Vercel Fluid and Supabase. And so I think the way that you can think about it is developer experience is eating the world, specifically the world of infrastructure. And at the time I was a random community member trying to make the developer experience better. I think the bar is so much higher these days, right? You have to, as a company, as a community, own that layer of the developer experience.
For sure. And then what about open source itself? I mean, Next.js is, I'm guessing, one of the most popular open source tools in the world. What do you think about the tension between providing an open source tool and commercialization? How do you think about it?
That's such a great example. We started off with that example of my ODM because over time I didn't really have the capacity to maintain it and so the company took it on themselves. My point on balancing open source and business is actually a very capitalist one. If you think about capitalism and why it's so great, it's because we're all incentivized to continue to grow the pie. We don't think of the economy as a fixed pie. We wouldn't think about it as a zero-sum game. And so the way that I think about Next.js is: if I grow Next.js to be the default way to build web applications and Vercel takes a slice of that pie, which I'm really happy to share with other players in the ecosystem, then that gives me two things. One is the capacity to fund the project. And this is no small thing, right? Because otherwise, for projects that are so critical to the infrastructure of the internet, you want to be responding really quickly to security CVEs and security issues or incidents. You have to be really good at keeping up with the change in the landscape of technology. So maintaining open source is clearly a full-time endeavor. And so if there's a company that is incentivized naturally by the progress of that open source project, I think things sort themselves out. I talked to a lot of early stage founders that ask me this question and my advice is always the same: don't think too hard about it because if your project grows and it becomes valuable to people, there will naturally be an emergence of services and infrastructure that are valuable around that open source project.
Yeah, there's also this kind of tendency that happens with products, as you pointed out. DX is kind of now the next frontier. Whereas when you're designing something in open source there's a little bit of this design by committee. So you're probably one of the best product minds on the planet right now, and one of the things that you've talked about is this progressive disclosure of complexity. How do you think about building products and reaching people but also allowing them to kind of have this unbounded experience?
Yeah, there's a couple things there. Number one is I try to make a distinction between the developer experience as far as the API and the developer experience as far as the workflow. So to also wrap up on the previous question, I think something you guys have done so well is Supabase actually gives me a sense of how my database is evolving over time. Developer experience is never a one-shot, one-time thing. Especially if you position yourself in the critical path of a production workload. You're going to have to be monitoring it. You have to be understanding, for example, which are your slow queries. You're going to have to be thinking about the health of that project or that piece of infrastructure. So, I think there's a place for companies to add a ton of value on the workflow. Nowadays, I think of developer experience as an open-source good. And I think at Vercel we've done a really good job of maintaining—Next.js developer experience is fully open source but when you ship to Vercel we enhance the entire lifecycle of bringing this thing to production.
Nice, very cool. Okay, so...
Sorry, also I'll finish on one other thing on the progressive disclosure of complexity. I'll say it's sort of become my main philosophy of how I build products. So on one hand you have to make your product fit within the needs of someone that is experimenting or prototyping with something. If you think about the success of your product in the context of vibe coding for example, why is V0 and Supabase such a good fit? Well, because in the early days of prototyping something, the tool lends itself really well to fitting the needs of what I call day one. The day one experience is something I still care very deeply about. Vercel is obviously operating at a very large scale today. Lots of enterprise customers, governments using the product. And so I have to think about what I call the day 1000 experience. You don't want your customer to graduate out of your platform, right? But day one is so important because it's even hard for us to size up what the total size of this developer market is. Especially in these days with AI, I don't think we could have anticipated what has happened with vibe coding. All of a sudden, everyone's creating a gazillion Supabase instances per day, right? And so that is a function of a founder. Typically, it's a founder that is obsessed with the day one experience. And what ends up happening is other platforms that are looking to integrate you will actually judge the value of your day one experience. When we were integrating Supabase tightly into the Vercel ecosystem, one of the things we looked at was database spin up time. It has to be instantaneous. It doesn't have to feel like you're spinning up a piece of infrastructure. It has to feel like you're storing a file in the cloud. That's how fast the expectation of a new database is today. And secondly, the developer experience. If an agent is going to be the user of this day one experience, the SDK and the APIs need to be so good that it's sort of like three lines of code or a handful of tokens to get it started with it.
Now, on the progressive disclosure of complexity, what happens at day 100 or what happens at day 10,000? If I'm not mistaken, and I don't know if you've upgraded from this marketing tagline, I remember a tagline you had like 'start on a weekend and then scale to billions.' I think that to me goes at the core of this day 10,000 thing. A lot of startups have died or failed to scale because their thing is really attractive at that weekend moment and then you begin to throw more data at it, more workloads, more requirements, more guardrails, more enterprise features and then you end up graduating out of the product. So it's extremely important that you think about this also for all your products.
Is that how you—well, how do you think about like V0 and Vercel? Do you think of it as an on-ramp and just part of the journey? Kind of like day zero essentially to—
Day zero is actually awesome. You know, as day one, we're day zero. We're even before day one.
But that's by the way how we refer to Supabase. You got this tricky thing in databases and in tech, you know, called the duckling syndrome. Whatever you choose as your technology is going to be one of the things that you stick with. Whatever you learn on, you stick with through your career. It's an investment. So the problem with databases is you don't ever migrate. And our thing was, well, if we want to become a database company, we have to be a day zero product. You have to think of us before you even know what you're going to build. You have to choose us.
Yeah. So day zero is awesome. Sounds like we both are contending for it. I mean, we're sort of making it right now. Day zero AI or something, but totally aligned on the philosophy. And part of this day zero mentality is you have to stay on your toes because the world is moving so fast. Runtimes, frameworks, integrations, agents, MCP—congrats on the Supabase MCP improvements. We also would have not anticipated that we had to adapt our DX, Vercel MCP, Supabase MCP, to this new ecosystem of APIs. And I think we have to work on this day zero thing every single day.
Okay, let's talk about your fundraise. You just raised a Series F at 9 billion. Congratulations.
Congratulations to you as well.
You shared in your blog post you want to move from pixels to tokens. So what does that mean?
Pixels to tokens. I think of the past 10 years of Vercel as building the most beautiful pixels, hopefully. So shadcn just launched more components today. I don't know if you checked it out, if you've been on X today. We're very excited about pixels. I love pixels. I'm the pixel guy. However, increasingly much of the software that we'll be creating and shipping will kind of have invisible interfaces. There will be a lot of work that is going to happen in the background. The Vercel chapter one was all about the foreground. You're interacting with websites that are really fast. You know, select.ssup.com is hosted on Vercel. It's beautiful, super green, and pixels matter a lot to me and we'll continue to invest in them, right? But, you know, in the future I'll say, 'V0, build me a SaaS company. I want to make 50k in MRR. Make no mistakes.' And so what's going to happen is the agent will start cooking. I'll go grab coffee. And all of that work that happens behind the scenes, all of that compute, all of the tokens that are going to flow between all of the steps between tool calls and computation, etc., will be happening in the background. And so this is kind of the foundation for what I've been calling the AI cloud.
When you think about AI, you might be thinking, 'Well, is that GPUs? Is that models?' I'm actually operating at a higher level of abstraction, which is: you get to pick whatever model you want to use but you're coming to us because you want to either build something that's AI native, so it's like brand new with AI. You know, a lot of people are building really cool products, like database.new is an example of an AI native product. But there's also a lot of companies that are thinking, 'How do I modernize myself? How do I disrupt myself with AI?' And they're also going to be using this set of products and experiences that we're creating.
If this becomes part of the experience, is this kind of still like the day zero? Do you basically—where do you see developers fitting in after the pixels and the tokens have been generated?
I'm team developers forever. So the experience that I'm optimizing for—the best metaphor is Next.js to Vercel hosting and CDN, etc., what AI SDK is to AI gateway and sandbox and Fluid, etc. And what that means is there's still a developer experience. This developer experience might not start out with you building a front end. You might actually just start out with the agent. You might integrate it with some other conversational interface. You know, some people might start by deploying their agent with an interface on Slack, email. The world is probably going to move into a lot of event-driven stuff, which is actually really cool. So one product that we recently launched is Vercel Agent. Vercel Agent fires up anytime a pull request happens. So this is why I call it event-driven. The human didn't actually ask, 'Hey agent, please go review my PR. Make no mistakes.' You just submitted your PR. Soon we'll fire up the agent when a build error happens, when you have a stale pnpm lock file, another event that triggers all of this massive amount of background computation and intelligence. So Vercel Agent started with using the AI SDK, deploying it to Vercel, maybe use some backend framework like Nitro, Hono, Express, whatever, and then maybe you'll eventually build a front end for it. So I'm not saying pixels are not relevant. I'm saying the value will accrue in the future towards the accruing of intelligence, and intelligence will be one-to-one synonymous with how many tokens are flowing through your programs.
