I have, but even with the high token generation speed, incremental Rust compile times were not a bottleneck.

I was disagreeing with the concept that incremental compile times could be a bottleneck. LLMs are even better at doing large swaths of work at once and having it compile first or second try than a human.