A new paper led by Mikhail Genkin was published in Nature. This work develops a flexible modeling framework that dissociates the dynamics and geometry of neural representations on single trials and infers both simultaneously from experimental data. Applying this approach to recordings from the premotor cortex, the study shows that individual neurons have diverse tuning to a decision variable computed collectively by the population. These findings reveal a unifying geometric principle for how the brain encodes sensory and dynamic cognitive variables.