I study how neural networks represent and process information — both biological and artificial. Currently interested in mechanistic interpretability and its applications to AI safety.