Vercel’s CEO & Founder Guillermo Rauch on the impact of AI on Web Development and Front End Engineering

Our CEO and Cofounder Lukas Biewald recently sat down with Guillermo Rauch, CEO and Founder of Vercel for a wide-ranging discussion on the future of AI, web development, and front end engineering. Below are some key takeaways from their conversation. You can watch the full interview here.

Expert AI Agent in Web Engineering

Lukas asked Guillermo about Vercel’s new product v0, an expert AI agent in web engineering that generates code, UI, and responses based on simple open-ended prompts. Guillermo shared how v0 is surprisingly good at translating a pasted image into working CSS so developers can avoid the toil of trial and error of a visual glitch they’re trying to fix. Guillermo described how for many of their front end developers the two primary tools they now rely on are v0 and Cursor.

“In a month we did more UI generations than in the previous 12 months. So v0 is growing exponentially and people are really appreciating this sort of like copilot experience but for generating web UI”
—Guillermo Rauch

Releasing Every Day

Guillermo shared how v0 helped Vercel realize that in the era of AI they have to release every day. The AI world is simply moving too fast. If you think you can operate on a yearly or even quarterly release you’re NGMI (not gonna make it).

Guillermo shared their obsession with iteration velocity by dogfooding their own full stack Vercel platform—front end, back end, Next.js, AI SDK, Postgres database—and spreading their GPU usage across 3 data centers and having 20 to 50 feature flags being tested at any point in time.

“So if you go to Twitter to v0, you're going to see that that account is just the product changelog. And every single day there is a ship”
—Guillermo Rauch

Open Source vs Closed Source

Lukas observed that Guillermo has been a long time champion of open source in his career but was only using closed source models in his business. Guillermo talked about the complexity of understanding the open source versus closed source debate on AI models and referenced Linux as the canonical open source example.

For a model to be considered truly open source it needs to share not only the weights, but also the datasets and the instructions. Guillermo also made the point that performance was paramount and that OpenAI, Anthropic, and Gemini still were the frontier models from their own evaluations.

“We're actually using only frontier models at the moment, excluding Llama. So I would love to use Llama, but in all of the evaluations, like it wasn't quite up to par”
—Guillermo Rauch

To watch the full session, click here.