Scientists have developed a geometric deep learning method that can create a coherent picture of neuronal population activity during cognitive and motor tasks across experimental subjects and conditions.