A language is LLM-optimized if there’s a huge amount of high-quality prior art, and if the language tooling itself can help the LLM iterate and catch errors