Cursor's agent swarm writes a 1M+ line browser in about a week, and critics find it barely compiles
Cursor ran hundreds of concurrent agents for close to a week, split into planners, workers and a judge, to build FastRender, a web browser written from scratch in Rust. The same post reports an in-place Solid-to-React migration of Cursor's own codebase that ran over three weeks (+266K/-193K lines), still needing review but passing CI. After launch, developers and The Register reported failing CI, no build instructions, commits that didn't compile cleanly and slow page loads. Build instructions were added within days.
The numbers (as reported)
| Duration | Close to a week |
|---|---|
| Output | 1M+ lines across ~1,000 files |
| Agents | Hundreds of concurrent workers |
| Solid → React migration (same post) | 3+ weeks, +266K/-193K edits |
Facts
- Company / subject
- Cursor (Wilson Lin)
- Tools
- Cursor (multi-agent research harness), GPT-5.2, GPT-5.1-Codex
- Codebase
- greenfield
- Work type
- New product (research)
- Stack
- Rust
- Published
- 2026-01-14
- Last checked
- 2026-10-18
Caveats
Vendor research experiment, and the browser was a test objective for the harness, not a product. The Register (26 Jan 2026) reported failing GitHub Actions on main, commits that didn't build cleanly, pages taking about a minute to load, and reliance on existing projects. A Servo maintainer called the code 'a tangle of spaghetti'. The JS engine was vendored from the engineer's own personal parser project. No cost was disclosed. Lin says the harness deliberately tolerated a steady rate of compile errors to keep throughput up. Cursor says Opus 4.5 stopped early and took shortcuts compared with GPT-5.2.