The Day Autonomy Stopped Being a Demo

The first time you ride in a car with no driver, the thing you notice is how normal it is. You get in, put on the seatbelt, and the car does the driving — signals early, leaves a gap, does the speed limit. Nothing about it is impressive, which is exactly what makes it impressive.

The demo is a performance; the product is ordinary. We've been promised self-driving cars for twenty years, and every demo was the same shape: a car driving itself around a closed course while an executive smiles for a headline. A demo you can always tell. Then you take an actual ride, and it's boring — and boring is the milestone.

Watching a machine do something is observation. Putting your body inside it and letting it do the thing to you is trust. The striking part is how fast that trust forms — one intersection, two turns, and by the fourth block you've stopped watching the wheel because there's nothing to watch.

That's the real milestone for any technology: not when it works, but when it becomes unremarkable.

Everyone credits the autonomy stack. The sensors, the maps, the learning. But the reason the ride feels ordinary is the boring part — the thousands of edge cases solved so the car can just drive. The pedestrian who hesitates. The truck double-parked. The cone in the wrong lane.

The demo is the car driving itself. The product is the car handling the world, which is 90% edge cases. The last 1% — the exceptions, the almost-never-happens — is the entire difference between a demo and a product.

The field-notes version of this — from the ride itself — is on WanderingCTO.

Think this argument fits your event? Tell me about the room — the calendar is selective.