Tesla's 45 Austin Robotaxis now have 14 crashes on the books since June 2025

That seems low. I would like to see the quantity of human interventions too.

Discussion: https://news.ycombinator.com/item?id=47051546

>“The new crashes include a collision with a fixed object at 17 mph while the vehicle was driving straight, a crash with a bus while the Tesla was stationary, a collision with a heavy truck at 4 mph, and two separate incidents where the Tesla backed into objects, one into a pole or tree at 1 mph and another into a fixed object at 2 mph.”

>Electrek analysis found that the vehicles have traveled roughly 800,000 paid miles in that time period, amounting to a crash every 57,000 miles. According to the NHTSA, US drivers crash once every 500,000 miles on average.

Jesus, 10x worse than the average US driver is crazy. There's serious variability in human driving demographics, with some subsets already being that much worse than other subsets, things like "What city do you live in" can bring that stat down significantly.

So, to all the people who insist that autopilot is "better than human", do you still believe it? Assume it drives as many unpaid miles, that's still 5x worse than average drivers, which includes things like alcoholics, people who consistently drive high, people who don't really know how to drive, and that person you see on the road eating a fucking meal and doing their makeup and reading a book, and drivers who are currently asleep.

I uh did not actually expect the data to be this bad. Statistically, we need more data, but this is concerning. Not slowly tapping into objects that cannot move is probably a good way to improve those stats. Did Tesla abandon the industry standard ultrasonic sensors everyone else studs their bumpers with? Many other cars in the industry can parallel park themselves through these sensors, and reliably do not hit parked objects.