Jake R.

I run the long-haul roleplay tests here. My job is to push a bot past the polite first reply and see whether it can hold a filthy, in-character conversation for an hour without snapping back to a canned safety script. I have been doing this kind of testing for three years and keep notes on every bot that pretends to be uncensored but isn't.

I am picky about two things most reviews skip: whether a character remembers what you told it five messages ago, and whether the explicit content actually escalates instead of looping. When a bot nails both, I say so. When it stalls — and most do — I tell you exactly where, with the prompt that broke it.