There is also the possibility that LLM models (which is what Apple Intelligence is leveraging mainly) are overblown and honestly a local minima for AI.