Vercel is watching developers become 10x more productive

Andrew Qu has been at Vercel for four years. He joined after building Google's AMP project and holds more patents in web performance than almost anyone. At Vercel, he runs engineering—a team that's grown to support a developer platform used by many of the world's most ambitious frontend teams.

This year at re:Invent, the conversation was all about AI. What Andrew finds fascinating isn't the tools themselves, but the second-order effects on developer behavior.

The 10x developer is becoming real

Vercel has concrete data: developers using AI tools are significantly more productive. The AI SDK, Vercel's toolkit for building AI applications, is showing remarkable usage growth. Feature development velocity across the company has increased dramatically.

The quantitative signal is in the logs. Developers who use AI coding tools are making more commits, deploying more frequently, and iterating faster. The 10x developer—long treated as a mythical creature in our industry—is starting to look more like an achievable state for ordinary engineers with good tools.

Docs are the new moat

One of Andrew's most interesting observations is about documentation. When LLMs write code for your platform, the quality of their output depends on the quality of your documentation. Good docs become a competitive advantage in a way they never were before.

Vercel has invested heavily in AI-optimized documentation. Their docs are formatted and structured to be easily consumable by models. When developers ask Claude or GPT to help them build on Vercel's platform, the answers are accurate because the training data is excellent.

This creates a virtuous cycle. Better docs mean better AI-generated code. Better AI-generated code means happier developers. Happier developers mean more adoption. More adoption means more investment in docs.

The framework wars are settled

Next.js and React are winning, and AI is accelerating that victory. When AI models are trained on the web's code, they learn the patterns that are most common. Next.js and React are everywhere—in tutorials, in production code, in open source projects. The models are better at generating code for these frameworks because there's more training data.

Andrew sees this as a network effect that will be hard to disrupt. New frameworks will struggle to gain traction when the AI coding tools are optimized for established ones. The rich get richer.

What's next for Vercel

When asked about 2026, Andrew focused on AI features within the platform itself. Vercel is building deeper AI integrations—not just hosting AI applications, but using AI to improve the developer experience. Better previews, smarter debugging, automated performance optimization.

The meta-observation is that Vercel is in a perfect position for the AI era. They sit between developers and production. They see every deployment, every error, every performance bottleneck. That data is incredibly valuable for building AI features that actually help.

This interview was conducted at AWS re:Invent 2025.