Dexterity and reliability, not hardware alone, limit physical AI robots

The Robot Report argues that dexterity is physical AI's main bottleneck, not the robot hardware itself.

It says a robot that succeeds 99% of the time is not ready for autonomous work, because a 100-step workflow at that rate completes only 36.6% of the time. The article says reliability depends on a closed loop across sensing, control and recovery, and that real-world deployment generates the data needed to improve it. It also argues that the right physical form depends on the job, citing Agility's Digit, ANYbotics' inspection robots and Starship's delivery robots as examples.