The art of reverse prompt engineering

I've been playing around a lot with [reverse prompt engineering](https://home.pickaxeproject.com/prompt-reverse-engineer), which I define as showing a GPT model an output, and then instructing it to write the prompt that can create that.

If you're operating in the OpenAI playground, you can do this with just simple prompt engineering.

If you're working in a coding environment though, you can feed it multiple outputs, produce the prompt that would produce each one, then cross compare. Then instruct GPT to write a prompt …

aipromptprogramming coding current engineering environment gpt multiple openai openai playground playground prompt simple