Well, continual usage doesn't really help with fingerprinting. They already got that.

I'd say, if you're going to do this (you shouldn't) the prompts should be fed by another LLM, and just let them create garbage together. I'm sure OpenAI likely uses user prompts for training, so this fucks up their training data.