Yeah. I think Tesla is making a mistake in going for cost optimization before their product actually works. You can see that it's technically possible to drive on sight alone; that's how humans drive cars. (But, it's not completely clear cut to me. Having the cameras in every Tesla means that they have a lot of data to use in their simulations, and that might help them more than depth data from lidar. I don't really know.)

Their current strategy does get people hyped. I wonder how many people you can program a machine to run over before it becomes an unprofitable activity.

I have a random comment I'll interject but it's totally off topic. My impression from watching people show off their Teslas is that they are constantly in awe of being able to get to work without controlling their car. But many of us have been commuting to work hands free for decades -- it's called public transportation. One person uses the power of electricity to take hundreds of people around the city, every 1.5 minutes during rush hour. It is pretty nice. But we didn't need self-driving cars to get there, we just needed to design our living and working spaces for hands-free commuting. Shrug.

> his Tesla often tried to run him into the same barrier.

In a way, consistency here is good. If it only sometimes tries to crash itself, it's harder to fix and it may indicate a bigger recognition problem.

Of course, that doesn't mean the product is any good, or that selling it is reasonable.