And the screamingly fast compilation speed is a boon to fast LLM iterations as well.