The first (and probably most useful) treats the blog as something like a diary: commentary on current events at some scale, e.g. personal, professional, news, et cetera. In this case, older content really does have much less relevance because the events themselves now live in the past. Your friends /might/ care about what you drank last night, but they almost certainly don't care about 5 months 12 days ago. You might want to comment on Linus' latest mailing list screed, but it might not matter for kernel-dabbling four years down the road.
The second use case seems to be where Daniel Tenner really is chewing on something. If you're writing new stuff that doesn't necessarily live in the moment - musings on philosophy or cool data structures or life advice or Paul Graham's articles - then the blog model doesn't fit as well. Newer posts don't matter more than older posts, and in fact might matter less.
Thinking on this, we also have the issue of serendipity. The writer has probably created other stuff that you might like reading even if it's not directly about the exact same topic. Seems like the sort of thing attacked by collaborative filtering and other methods. Combine that with a CMS-style knowledge architecture (e.g. structure by something other than chronology if chronology doesn't make sense), and you're getting somewhere.