In the language of economics, one could say that the market for programming talent in many non-profits is undeveloped, and so donating can maximize utility, despite other jobs paying better.

As another example, the market for the environment is hardly developed at all. And the market for programming talent within companies is quite underdeveloped too; that's the crux of PG's argument anyway, working at a startup gets you closer to market price.