I think what could eventually be even more impactful than the videos themselves is the statistics tracking he showed. With that kind of data across thousands or millions of people, you would see patterns of people who had the same problems and questions in the same places, and you could redo the videos to be more clear in those areas and answer those questions preemptively. With a web-scale audience you could do A-B testing experiments and optimize the performance of your teaching material. I think a curriculum optimized in this way could eventually be dramatically better than even the best traditional education has to offer today.

Today teaching is an art; this could turn it into a science.