> Buying them is an easy way to feel good in the moment and it's easy to tell yourself that you'll do the courses later.

What's more discouraging is that completing them may be little more than that as well. Sure, you put in the self-discipline and work to ace all the quizzes and maybe even turn in the practical final assignment, and you have a piece of paper to show for it, but did you really learn all that much? How much sticks with you in a couple of years?

I took Odersky's Scala course on coursera, which was fairly tough (relative to the ~10 other similar courses I've taken as an adult). Certainly felt good completing it. Didn't feel so good about dropping out of the sequel course a couple of weeks in as I realized I just couldn't complete the assignments in time without sacrificing too much of family and social life, but no matter. What do I remember of Scala today? Certainly not enough to program in it... Maybe some vague things about covariance and contravariance and how mutability makes it painful.

I also did a bunch of Andrew Ng's courses. The first covered a bunch of non-NN machine learning methods which are a lot less relevant than they were. I remember their names at least, but I certainly couldn't explain them in a code interview. Then I learned to write vectorized hand-calculated backwards passes in Matlab (well, gnu octave), and some early Tensorflow. Also well out of code interview accessible memory by now. Ng's courses were great at giving you a sense of accomplishment and removing all time-consuming frustration not related to the actual focus of the course... But learning these things a few years before most people didn't make me rich, and these days I'd have to ask a chatbot like everyone else. Skills you don't use atrophy, and I couldn't convince anyone to pay me to implement ML models.

So I guess what I'm getting at is, even at their best these courses may not give most graduates what we really hope for.