Comment by vagab0nd

Cool trick. But is this better than reinforcement learning, where the LLM decides for itself the optimal thinking time for each prompt?