Is the problem you're thinking of LLMs, or cloud LLMs versus local ones?