I think people like Knuth, Torvalds, Carmack, etc. are the models for career development. Why would you assume that a strategy that didn't work for any of the biggest guys around would be better for you than theirs?