Now the question becomes what happens when those LLM become very good at abstraction?