Sixteen Claude AIs Build a Rust-C Compiler to Boot Linux Kernel

TL;DR
Anthropic ran 16 Claude Opus 4.6 agents on a shared Git repo to autonomously build a Rust-based C compiler; over two weeks and about $20k in API fees, the system produced a compiler that can build major open-source projects and boot the Linux kernel across x86, ARM, and RISC-V, and it even runs Doom, with a 99% pass rate on GCC’s torture tests. The effort relied on autonomous agent teams, task locking, and bespoke test harnesses, but required substantial human scaffolding and faces notable limits (no 16-bit x86 backend, buggy assembler/linker, and degraded coherence beyond ~100k lines), illustrating both the promise and caveats of current multi-agent coding.
Topics:businesstechnology#ai-agents#c-compiler#claude-opus-46#gcc-torture-test#linux-kernel#technology
- Sixteen Claude AI agents working together created a new C compiler Ars Technica
- Anthropic's newest AI model uncovered 500 zero-day software flaws in testing Axios
- Anthropic Releases New Model That’s Adept at Financial Research bloomberg.com
- Anthropic’s Claude triggered a trillion-dollar selloff. A new upgrade could make things worse Fortune
- Anthropic releases Opus 4.6 with new ‘agent teams’ TechCrunch
Want the full story? Read the original reporting
Read on Ars Technica