For as long as engineers have wrestled with the invisible turbulence of air and water, progress has come slowly and expensively — through wind tunnels, trial and error, and hard-won increments. An international research team has now built HydroGym, an open-source machine learning platform that teaches AI agents to actively shape fluid flows, drawing on reinforcement learning to compress decades of computational struggle into something faster, cheaper, and transferable across industries. Demonstrated first on simple channel flows and then successfully applied to airplane wing simulations — cutt