On October 27, AgiBot officially released the embodied agent framework Genie Master, enabling robots to unlock new skills while on the job. The Genie G2 equipped with the framework, while working in a real convenience store, did not rely on human demonstrations or debugging; it identified the causes of failures itself, adjusted strategies, and accumulated experience, continuously improving on multiple complex and delicate tasks and achieving continuous self-evolution with no human in the loop. This is the first time the industry has moved beyond simulated environments and out of the laboratory to validate Physical RSI (embodied recursive self-evolution) in live operational scenarios. Based on the framework, after the agent issues execution guidance based on on-site observations, information about the success or failure and bottlenecks after the robot executes can be returned in real time; when a task is interrupted, the system can preserve completed progress and let the agent replan from the breakpoint, while also supporting dynamic scheduling of multiple robots.