On Hacker News today, someone posted a Page-Rank-resorted listing of all of Paul Graham's essays. The content itself I'm pretty familiar with since we are acolytes of the Y Combinator school.
The #1 thing I am struck by here is that even though I am familiar with most of the content and have read most of them already -- it serves as an amazing reminder of what is important and notable. In a glut of information, the most important and scarce resource is attention.
Luckily computers are quite good at processing large amounts of information. Algorithms like Page Rank let smart programmers make sense of the world. Facebook is one of the few other sites that have realized this and are actively incorporating it into their product in the form of edge rank.
If chronological sort is dead, why is the default sort of almost everything else on the web pretty much chronological? It is certainly easy to do. ORDER BY created_at is probably the most used SQL query there is ont he web. Clearly there is a problem here -- and people are rushing in to solve it.
Some friends of mine from Anomaly Innovations created The Cadmus, a twitter client that shows your twitter feed via relevance instead of chronological. They found in AB testing their relevance algorithm that they could increase engagement by 40%! This seems so compelling I'm rather blown away Twitter hasn't done anything in this area yet. Other groundbreakers to watch out for in the space include Chris Dixon's incredible team at Hunch, YC-backed Directed Edge, and Louis Gray's my6sense.
While there are leaders in the pack, the relevance space is littered with failures and false starts -- so much so that there is a bit of a stigma with being another one of those 'relevance and machine learning' startups.
With good reason. Many fail.
This form of social-oriented content relevance is a tough nut to crack. It's fundamentally different from traditional Netflix-style machine learning since the number of watchable movies in the world is very much a finite set. In contrast, the amount of interesting content on the web is essentially infinite. The field is still a bit wide open because few people have both the dataset to work and test on, AND the financial backing to see the project all the way through.
I know this: The future of the web does not include chronological sort as the default view. In the next few years, the frontier of this will be pushed by smart and winning teams that will figure out a way to productize interestingness and provide it on tap. We're only in year 20 of building Vannevar Bush's memex -- a world of human-computer-symbiosis. Conquering the chronological problem is the next logical step.