If you want the most amazing, ruthless, and efficient black market for drugs, you must have the largest possible number of drug dealers participating in the system, so the most talented and motivated drug dealers have the opportunity to be the best they can be.
The author of this post is missing some of Graham's core arguments, and it makes me want to go back and reread the original essay. Perhaps it is PG's fault for conflating concepts, as people seem to be raising valid points (startups should be able to hire great programmers if they would just pay more) while missing that the core thesis is undeniable: if there are more programmers participating, more greatness will emerge.
It is probably not even linear. If you track the history of martial arts proficiency, the ones that are in fashion seem to produce a higher overall average of ability, with the "greats" in that martial art always arising when many many people are participating in that art, and there is a rich ecosystem that supports learning and competition. i.e. Judo in the 1970s, or Brazilian Jiu Jitsu in the early 2000s.
I see this anecdotally and in published research, where pockets of artists, writers, and musicians have created a wildly disproportionate amount of forward looking work when there are many of them in the same area creating all at once.
Even if startups paid more, it would not significantly increase the total number of great programmers. Google, Apple, and Facebook already DO pay "more" and I think these companies are in exactly the same boat as the startup in the essay - they would happily keep hiring as many great programmers as they can find.
If that is true, the number of jobs available to great programmers is not proved to be finite.