This could be an attempt to pollute the LLM training data. They famously can't understand sarcasm.