f and p appear to be the same thing: the annual income they would get from coding full-time.

If they're doing some self-improvement activity that requires that they stop wealth-creation altogether, p = 0. If they're coding half-time, p = 0.5f. Etc. You're right about f.

Also, if you can grow your wealth-creation skills faster than your discount rate, then obviously you should put all your effort into growing those skills and none into earning money now. Obvious within the terms of the model, at least. In practice, you have to use the skills as you develop them -- that's part of learning the skills. Choose a project that you need some new skill to complete. A real project, creating a product that that you are going to sell.

OK, but what if you have one project that's a little more educational and one that's a little more profitable? The equation should help you decide between them.

BTW, you've done a good job of explaining why I think college for programmers is stupid. (But I'm going anyway, for the friends/prestige.)

I believe this applies generally to any form of self-improvement: doing stuff, and learning to do stuff better, cannot be separated.

Hm. I tend to think that someone can often improve their skill faster per minute by reading advice than doing. The things I learned from my last project (~1 month of work) would easily fit in a blog post.

Plus what about reading about a subject like probability, game theory, or history that could potentially transform the way you look at things?

One hole in my model is it has no way of taking in to account self-improvement effects that are temporary, such as spending the time necessary to think of a really good idea for a project (significantly helps your wealth creation skill, but only until you finish that project).