For whatever reason pg just doesn't want to do that. Perhaps he doesn't want building HN itself to become a distraction (for other people) from building more useful stuff like startups. There may be a sort of contradiction where ninja rockstar programmers are too valuable to waste on improving HN, but HN is simultaneously too important to be put in the hands of mediocre programmers.

HN is kind of a double-edged sword because while it is educational it's also distracting and addictive. If making HN better also makes it more of a time-suck, the exercise could be a net negative for the community.