Parent poster has a point that the difficulty of actually identifying the exceptional programmers poses a major problem here. P.G. has actually written about this himself, arguing that great programmers are the only ones who can identify other great programmers, and that they can only do this if they have collaborated with them on actual work.

See the 'Recognition' section in "Great Hackers" essay, http://www.paulgraham.com/gh.html: "So who are the great hackers? How do you know when you meet one? That turns out to be very hard. Even hackers can't tell."