Every time I ask the latest models to write something, it defaults to LLMisms like contrastive negation and long unnecessary lists.Is there a way to desloppify AI writing, or is this just the new status quo?