I suspect the mistake of having a human operator watching Netflix instead of monitoring the road was probably the "most glaring". Software shitting the bed is a daily occurrence after all, but I can't remember the last time I watched TV while I was supposed to be driving my car.
Imagine having to sit at a desk watching a Twitch stream of someone driving a car, and if you see that they're about to crash you have to hit a button. I think the average person would do well at that task for a few hours... can you imagine doing it for 2000 hours per year if that's your full time job?