Keenon's Embodied Community at WAIC 2026: How Far from Visible to Trustworthy?

I noticed an interesting detail: at the WAIC 2026 booths, Keenon Robotics' embodied robots weren't surrounded by engineers debugging them like in previous years, nor was anyone standing nearby with a remote control. They were walking around on their own, serving tea and water, and avoiding pedestrians—at least superficially. But as someone who deals with factor data every day, my first reaction wasn't 'Wow, amazing,' but 'This conclusion needs data support.' Is the sample size sufficient? How long does it run when generalized to real-world scenarios? My risk model tells me that any system that 'looks autonomous' hides unknown tail risks outside of historical backtests.