another story about the lack of chain-of-thought reasoning traces in frontier models...