Explore how roboticists train policies in simulation and make them work on real hardware. This lens covers environment design, domain randomization, system identification, adaptation, and the stubborn gaps that emerge when robots encounter real contact, sensors, and uncertainty.