News

AI agents rebuilt curl in Rust for $8,600 — its creator shrugs

What's the deal? AI coding agents converted 185,000 lines of curl's C code into 340,000 lines of Rust in a little over three weeks — and curl creator Daniel Stenberg is unmoved. One team behind a November rewrite claims it spent $8,600 to pull it off.

Why now? This autumn brought a wave of curl-in-Rust projects, starting with a from-scratch reimplementation called urlx and followed by a rewrite that reused the curl name outright. It sits within a broader trend of AI-assisted Rust ports, including efforts around the TypeScript compiler and GitHub's Copilot runtime.

Writing code once was never the issue. Stenberg argues the feat is unsurprising: AIs have trained on curl for years, and the project is heavily documented with thousands of test cases. "The C code sits there and all the AI needs to do is convert it to rust, then polish until the tests work," he wrote.

Behind already. During the same three weeks, the curl project merged more than 300 commits and shipped over 200 bugfixes, plus several vulnerability fixes — none of which the clones have. "Will these clones keep up?" Stenberg asked, projecting another 200 fixes in the following month.

Not a security win. Stenberg notes the rewrite is no simple safety upgrade. Of 81 vulnerabilities confirmed in curl during 2026, only 14 — 17.3% — stemmed from mistakes a memory-safe language like Rust would likely have avoided. A Rust curl would still carry 67 of them.

What could go wrong? A snapshot of a live codebase goes stale the moment it is taken. Anyone running code derived from an earlier translation would miss upstream fixes unless someone carried them across by hand or ran the agents again.

The signal: The demonstration shows agents can reproduce mature, well-tested software fast. But Stenberg's point is narrower and harder to dispute: a working community of users, contributors, and maintainers — not the language — is what keeps a project alive. "I don't believe in rewrites into any language," he wrote. Whether these clones earn real adoption remains unproven.

Read more: daniel.haxx.se

Image credit: "Daniel Stenberg on Curl" by LornaJane.net, licensed under CC BY-ND 2.0

More top stories