In an age when cities strain under the weight of their own growth, a team of researchers has offered a quieter solution to one of urban life's most persistent frustrations: not more roads, but smarter ones. Using reinforcement learning, they taught traffic signals across four major cities to adapt in real time, redistributing the burden of congestion across networks that already exist. London's pilot results — vehicles moving twice as fast, journeys cut by more than half — suggest that the limits of our cities may be less physical than they are philosophical, less about asphalt than about imag