Mechanistic interpretability and artificial psycholinguistics are transforming our understanding of large language models. In this episode, Arshavir Blackwell explores how probing neural circuits, behavioral tests, and new tools are unraveling the mysteries of AI reasoning.