As we move towards a general tool for prototyping LLM reasoning and planning algorithms, we should decouple the notions of: - algorithm being prototyped - benchmark / problem set being visualized