通过要求智能体加速代码，写出比最先进库更快的 Rust 代码

Writing Rust code that's faster than state-of-the-art libraries by asking agents to make the code faster

By giving agentic LLMs explicit, pass/fail performance constraints and anti-cheating guardrails, Max Woolf got them to iteratively optimize Rust code to 2x-20x faster than state-of-the-art libraries across ML and everyday software domains.