Interesting, I like this one. This seems like something GPT-3 might be able to do with some fine-tuning.