Back in 2013, they targeted 35GWh of battery production in 2020 for 500K/y vehicles (source: https://www.tesla.com/sites/default/files/blog_attachments/g...)

They've already exceeded that run rate on production and delivery (both batteries and vehicles) while being profitable.

They do have ~a million cars on the road but the software is neither capable of self-driving nor car sharing. They year isn't over, though, and the comment about robotaxis was that the car would be feature-complete[0], not operational.

[0] https://en.wikipedia.org/wiki/Feature_complete

He wasn't off his rocker as much in 2013 as he is now. Sleep deprivation and possibly drugs really do wonders for mental health long term....

Yet, his companies are in a much better position compared to 2013. What should one be worried about? The name of his children and his Twitter rants/jokes?

Or how the companies we're talking about are actually performing?

Do you think you might be oversimplifying complex human behavior to a fault? Especially with no real insight into his behavior?

Rest of the companies doing that are in more or less the same boat. Tesla ahead of all of them in bringing some pieces of the autonomy to market.

Other companies have made over-optimistic predictions but not on the same scale and with the same frequency. Tesla is ahead of them on the "bringing to market" part but not on the autonomy part, where they're not really working on the same level of difficulty problems as the ones trying for L4.

Tesla are aiming for full self driving, aren't they? If they have managed to find a simpler problem set that will still allow them to reach that outcome and release incrementally, isn't that a smarter approach?

I think that releasing a product is still completely orthogonal to building an autonomous vehicle at this point, so it's maybe a smarter approach for their bottom line but doesn't get them to autonomy any faster.

I think to accept that delivering incrementally is helpful for making an L4 car you have to believe that having more data is the bottleneck, which probably isn't the case after a few million miles.