Apple's Minions research project splits AI inference workloads across connected Macs using Apple Silicon. This creates a shared pool of graphics unit capacity. The project leverages idle hardware to distribute processing.