“OpenAI has explicitly trained its Codex models to be proficient at performing code reviews.”