In my experience code reviews involve a fair bit of back-and-forth, iterating with the would-be committer until the code 1) does what it's meant to and 2) does it in an acceptable manner. This parallels the common workflow of trying to get an LLM to produce something useable.