Cloudflare Engineer Rebuilds 94% of Next.js API Using Claude in One Week

A single Cloudflare engineer has rebuilt 94% of the Next.js API using Anthropic Claude AI model, demonstrating the growing capability of AI coding agents to handle substantial engineering tasks.

Steve Faulkner, Director of Engineering for Cloudflare Workers, completed the experimental project in about one week, spending roughly $1,100 on Claude API tokens. The resulting project, called vinext, is a drop-in replacement for Next.js built on the Vite build tool rather than Vercel Turbopack.

We honestly didnt think it would work. But its 2026, and the cost of building software has completely changed, Faulkner wrote in Cloudflares blog post.

The motivation was practical: Next.js has a deployment problem when used outside Vercels platform. Cloudflare had been contributing to OpenNext, an open-source adapter project, but found the approach fragile - reverse-engineering Next.js build output leads to unpredictable changes between versions.

Faulkners alternative approach was to reimplement the Next.js API surface directly on Vite, rather than adapting Next.js output. Early benchmarks show vinext builds production apps up to 4x faster and produces client bundles 57% smaller than Next.js. Customers are already running it in production, according to Cloudflare.

The project is open source on GitHub at github.com/cloudflare/vinext.

The experiment adds to a growing body of evidence that AI coding agents can handle significant real-world engineering tasks - not just code generation demos, but substantive reimplementation work that integrates with existing ecosystems and serves production traffic.