OpenAI proposes a second neural net to catch ChatGPT's code mistakes

A large language model catches many coding errors, but it also has a propensity to hallucinate bugs that aren't there.