Why we build autonomy instead of better joysticks

Teleoperation scales with bandwidth and inversely with distance. Autonomy scales with neither. That single sentence determines almost every engineering decision we make.

The practical consequence is that we spend more on perception and verification than on actuators. A machine that can explain why it stopped is worth more than a machine that moves slightly faster.

It also changes how we test. We do not measure whether an operator could have driven better; we measure how often the machine reached the right conclusion with the information it actually had.