"In the experiment, sixteen AI agents, all running on Claude Opus 4.6, were tasked with building a C compiler in Rust from scratch. After formulating the goal, the human supervisors largely withdrew. The agents worked in parallel on a shared Git repository, without central orchestration or a controlling main agent. To make this possible, the company developed its own technical infrastructure. Each AI agent ran in a separate Docker container and worked in an infinite loop, automatically starting a new session after completing a task."
"The project ran for almost two weeks and included approximately two thousand Claude Code sessions. Approximately two billion input tokens were processed, and approximately 140 million output tokens were generated, accounting for nearly twenty thousand dollars in API costs. The end result is a compiler of approximately one hundred thousand lines of code. According to Anthropic, the compiler can build realistic software."
Sixteen AI agents running Claude Opus 4.6 were tasked to build a C compiler in Rust from scratch with minimal human supervision. Agents operated in parallel on a shared Git repository without central orchestration. Each agent ran in a separate Docker container and looped sessions, coordinating via simple lock files. The project used roughly two thousand Claude Code sessions, processed about two billion input tokens and produced 140 million output tokens, costing nearly twenty thousand dollars. The resulting compiler comprises one hundred thousand lines of code and successfully compiled Linux kernels and major software packages, achieving approximately 99 percent on the GCC torture test.
Read at Techzine Global
Unable to calculate read time
Collection
[
|
...
]