What was said
At Tesla's 2019 Autonomy Day, Musk announced that by 2020 over a million robotaxis would be on the road, earning money while their owners slept. In the same period, autonomous trucking was the other consensus prophecy: long-haul truck drivers were widely cast as the first occupation to fall, and Andrew Yang built the core of a presidential campaign on the truck driver about to be displaced. The argument sounded solid: highway driving is simple, labour is trucking's biggest cost, and sensors and models were both improving fast.
What actually happened
2020 brought no million robotaxis. Every year since, "full self-driving next year" has been re-announced, and the rerun continues. Waymo does run real driverless taxis now — but on geofenced areas, high-definition maps and remote assistance, ground out over a decade, not rolled out overnight as prophesied. Long-haul autonomous trucking remains pilots and fixed routes; meanwhile the headlines are about truck driver shortages.
Why it didn't come true
The long tail lives in the physical world. The model handles 99% of road conditions; the remaining 1% is construction zones, storms, a traffic cop's hand signals and pedestrians who can die — each case paid for in time and accidents. Then unit economics: dispatch, cleaning, remote takeover and insurance for a driverless fleet never added up to crushing the human driver's cost. Finally regulation and liability: who answers when someone is killed is a question a decade has not finished answering.
The unexpected part
What self-driving actually changed first was not taxis but warehouses and mines — enclosed sites, low speeds, fixed routes. The spotlight stayed on the hardest scenario while the technology entered quietly through the easiest one. The misalignment repeated everywhere afterwards: predictions fixate on the most dramatic version; reality starts from the most boring one.