Cool project. Has anyone experimented with test driven development and GPT-4? I can imagine writing a scaffolding program that prompts GPT to write a function

Cool project. Has anyone experimented with test driven development and GPT-4? I can imagine writing a scaffolding program that prompts GPT to write a function that does X and passes a set of tests (included in the prompt), but then repeatedly re-prompts with type checking and test output. I bet this would succeed (in the judgement of an expert) quite a bit more often than a one-shot prompt.