On October 29, during the main forum of Zhiyuan Innovation Day 2026, Yao Maoqing, co-founder of Zhiyuan, emphasized that embodied AI is now encountering more pragmatic challenges. The criteria for assessing robots have shifted: success is no longer defined by their ability to perform impressive demonstrations, but rather by their capacity to operate reliably around the clock in real-world settings, respond effectively to unexpected disruptions, and deliver precise ROI (Return on Investment) calculations. Yao argues that the next stage of embodied AI advancement should not focus on accumulating isolated capabilities. Instead, it should ensure that every real-world operation generates valuable data, builds operational experience, and feeds these insights back into the system. Only by establishing such a closed-loop process can robots evolve from “one-time successes” to “sustained excellence,” transforming embodied AI into a measurable, replicable, and scalable tool for boosting productivity.