running inference with no prompt just letting the model go wild on its own makes me really wonder about what sort of dogshit data they're using to train