On Sunday, Amazon's Zoox began charging for robotaxi rides in Las Vegas.

A purpose-built vehicle, no steering wheel, no pedals. The technology worked well enough that a regulator let it carry paying passengers.

The interesting part was not the car. It was what Zoox's CEO, Aicha Evans, said about the moment.

However hard it was to build the technology to get to this point, there's a new hard that's coming, which is paid operations. It's different when it's free versus when the customer is going to pay.

That single sentence is the whole story of autonomous mobility right now.

For a decade the hard problem was the driver: could a machine perceive the road and control the car safely. That problem is not finished, but it is visibly being solved. Zoox, Waymo and others are proof.

The new hard problem is everything around the driver. Charging the fleet, cleaning it, repositioning it, monitoring it, recovering it when something goes wrong, pricing it so the maths works, and meeting a paying customer's expectation every single time. That is not an AI problem. It is an operations, energy and logistics problem, and it does not improve on the same curve as the software.

You can see the gap already. In the same period that robotaxi technology kept maturing, Tesla's paid robotaxi miles actually fell by around a third quarter on quarter, even as it added cities. Capability went up. Profitable operation did not follow automatically.

This is why I keep saying the autonomous driver will arrive sooner than most expect, and the autonomous business will arrive much later.

The demo proves the technology. Paid operations prove the company. Evans clearly knows which one she is now facing.