Inspired by the human eye's accommodation mechanism, this paper proposes a supervised-evolving learning algorithm and designs a neuro-metasurface focusing system. Driven by on-site learning, the system rapidly adapts to changing incident waves and environments without human intervention. Adaptive focusing is demonstrated in various scenarios with multiple wave sources and scattering obstacles, showcasing potential for real-time electromagnetic wave manipulation in applications like achromatic beam shaping, 6G communication, and intelligent imaging.