I think that was somewhat the point of this, to simplify the future complex scenarios that can happen. Because problems that we need to use AI to solve will most of the times be ambiguous and the more complex the problem is the harder is it to pin-point why the LLM is failing to solve it.