Fabien Roger comments on Reasoning Models Sometimes Output Illegible Chains of Thought