Like you, I am a fan of autonomous driving technology. I also agree that Waymo has made some impressive strides, especially given that they are limited by human software-writing capacity, which is probably close to topping out. Tesla has chosen an entirely different strategy, as you know. By building a system on deep learning neural nets and inexpensive sensor suites, they are enabling scale and flexibility, without the need for HD maps, geofencing, multiple sensors and so forth. In Tesla's case, the limitation is no longer human software, but rather the curated training data. Kaparthy argues that the performance ceiling of the neural net approach is much much higher, even with existing hardware. Time will tell, but if I were betting, I would bet on Tesla to get to the more general solution first.