Yeah, LLM + Reasoning Module is only as good as its weakest part. And we had more than one AI winter around the development of reasoning modules. I think you're 2x correct -- we need both and the exercise to create it is worthy of :p