with the speed of llm/ai improvement, this too maybe steamrolled