He is an absolute treasure, I have watched all his videos more than 4 times and I don't think I would've been able to have a good mental model about deep learning without them, regardless of the amount of Bengio, Goodfellow etc lectures I have seen, none of them come even close.

He is singlehandedly enabling millions of people to understand what is going on, what + and * do, actually demystifying the "wires".

I just wish he start thinking of himself as more than 'collapsing weights', regardless if it turns out to be true.

I agree, I think I learned the most on this topic from his videos. And before that (a while ago), it was Andrew Ng coursera's class. The latter had hands-on project, which is much better than just listening in term of retention.. I don't know if Andrej Karpathy has more structured classes somewhere.