Magnetic skyrmions are nanoscale spin textures that can behave like particles and move and diffuse in real space. This makes them promising information carriers for future memory and computing technologies. A new computational study in npj Spintronics shows how geometry and topology can guide asymmetric thermal diffusion of magnetic skyrmions, offering insights for unlocking unconventional computing paradigms, including future stochastic and neuromorphic computing devices for more efficient artificial intelligence (AI) applications.