While there is 100% improvements to be made in the “how do we review code as fast as ai writes it” area I think the comparison of it and compilers is a false equivalence
Reviewing LLM code is much more similar to reviewing another humans code than a deterministic compiler