The submitted article (blog post) has an interesting factual claim with which I disagree, but that leads to a technical suggestion that I find intriguing.
"I'll start: for me, the main problem with HN is that it's extremely hard for new links to reach the front page, or even get more than one or two upvotes."
I disagree that it's extremely hard for new links to reach the front page. Most newly submitted links don't deserve to get to the front page, and too many of the links that reach the front page are still not the best links that were submitted on the same day. (For the record, each time I visit the front page, which is several times a day if I am interspersing breaks into working in my home office, as I am right now, I then go to the new page. So I am scanning the new page often to look for good new links to upvote. I upvote them if I see them. Some of them make it to the front page, but many do not. Many worse links receive no upvotes at all, not from me and not from anyone else, but some worse links, for example links with linkbait titles from low-quality sources, nonetheless make it to the front page.) Every link has some shot at making it to the front page. The front page will best reflect community consensus about what belongs on the front page if all experienced users who have reviewed the site guidelines
http://ycombinator.com/newsguidelines.html
(which were recently updated) take turns scanning the new page from time to time to look for the good stuff.
While I disagree with what the main problem identified above is, the suggestion in the submitted blog post about what to do about it is not half bad:
"One way to solve this might be to reward people who "discover" promising new links, i.e. give them their first couple upvotes. For example, if you're one of the first 5 people to upvote a post that eventually goes on to earn more than 20 upvotes, you get extra karma."
What this helps deal with is the first-past-the-post problem in submitting articles, which is that some people are all too aware that they get NIL karma from a submission unless they are the very first to submit that link, so they use RSS feeds and scan titles (without fully reading articles) to decide what to submit. That often results in submission of inferior articles and especially it results in submission of noncanonical URLs that mess up the operation of HN's duplicate detector software. So I would not be against implementing this idea, which would put the community on notice that everyone can take time to READ articles before deciding what to submit, and indeed everyone has something to gain from promoting a good article from the new page to the front page.
On the other hand, if I were directly answering the question "How would you improve Hacker News?" with a focus on software and interface tweaks, I would put a prominent link to the site guidelines, or perhaps even a snippet of key guidelines in every submission form and every comment form, and I would implement pg's idea and my follow-up to that
http://news.ycombinator.com/item?id=4397542
of looking at how users use their comment upvotes as a signal of which users are most determined to uphold the community guidelines here, adjusting their voting influence accordingly. In general, I think that most good stuff submitted to HN, either as article submissions or as comments posted to other people's submissions, gets too few upvotes. Look for good stuff and upvote it early and often is my approach to improving the community.