Thesis
Why Tesla’s approach wins autonomy
Geofenced lidar fleets prove robotaxis can work on a map. Tesla is building a general driver that works on any road the cameras can see — then manufacturing the cars that run it.
1. Vision generalizes
Humans drive with eyes. Tesla does too — with redundant cameras and neural nets that improve continuously. Lidar and HD maps work brilliantly inside carefully maintained operational design domains. Outside those bubbles, cost and coverage collapse. Vision is how you cover the planet.
2. The fleet is the dataset
Millions of Teslas already on the road stream the edge cases that matter: construction zones, emergency vehicles, weird weather, human chaos. That is not a research demo — it is a living training set competitors cannot buy. Thirteen billion-plus FSD kilometers is the moat.
3. Hardware + software + factories
Autonomy is not only software. Tesla designs the silicon, trains the models, ships OTA updates, and builds Cybercab on its own lines. Vertical integration turns robotaxi from a pilot into a product you can manufacture by the hundred thousand.
4. Honest SAE framing
We do not pretend Level 2 is Level 4. Consumer FSD is supervised. Robotaxi ops and Cybercab target Level 4 inside expanding domains. Clarity builds trust — and Tesla’s trajectory from supervised to unsupervised is the clearest large-scale roadmap in the industry.
Keep reading:FSD guide·Cybercab·SAE levels