Ask HN: How do you make AI writing usable?

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?