The idea. Everyone gets the same prompt ("give beige a personality," "what keeps you up at 3am"), the same fixed kit of pieces, and 30 seconds. So does an AI. Every tile lands on a gallery wall, and players vote: which one is the bot? Then comes the reveal: who fooled you, and who you fooled. The constraint is the whole point. When everyone has identical pieces, the only thing separating the tiles is taste, wit, and choice, the human stuff. Right now the bot fools 72% of players, and every round feeds that number: a live, accumulating measure of where human judgment still holds the line.