Just front the GPT call with a function that gathers a bunch of example posts. Feed them into GPT and then tell if to write a comment in the style of those that

Just front the GPT call with a function that gathers a bunch of example posts. Feed them into GPT and then tell if to write a comment in the style of those that you just posted. That way, it doesn’t matter whether the particular user was in data. Also, that way the model has no excuse. Because you’re not asking it to imitate a specific living person. You are simply asking it to match style to a bunch of examples.