This point goes against all of his other points. A skill that takes six months to get extremely good at is going to be commoditized in short order. So, it seems like a mistake to bank on this kind of knowledge being relevant in 40 years.
Honestly though, I am starting to feel like a fool for completing a CS degree. While I have a great career, most of the what I learned was useless. Back in 2012, I couldn't get a job doing machine learning because nobody wanted people with those skills, and the few companies that did looked for Ph.Ds. So even though I could build do cool stuff with ML in plain Octave, I couldn't get a job in the field.
Today, all you need to do is to break into the field is read a few books and learn a tool. The skill has been completely commoditized, rendering all that I work I did moot. Then again, at least now I sort of have a job in the field.