OpenAI fed this thing a steady diet of internet forums as training data. You'd better believe that it knows how to verbally destroy anything and everything, at every conceivable level of sass and/or vitriol.