Exclusive: Everyone is racing toward AGI and general-purpose robots. @xiao_ted has been thinking deeply about the problem almost nobody is naming: even if both arrive, what still limits progress in the physical world?

Ted argues it comes down to verification, how long it takes to trust an answer. In sim, a humanoid learns to walk in 15 minutes. In the real world, checking a single robot policy can take 100+ hours. Progress moves at the speed of verification, not intelligence.

A genuinely different way to see the field.
https://www.youtube.com/watch?v=IGjFIeXAxzM