Now that MO has been around for 2.25 years or so, it has a large amount of content in the form of questions and answers. The search features work reasonably well, so if I have a question that's been asked before, it is usually easy to find.

However, when using the website to browse in general (as opposed to search for a very specific question), I personally find that there is too much focus on the most recent questions and answers and not enough on the very best questions and answers of all time. This has its plus points, but I see a big minus. It's very hard to get a general feel for a topic within mathematics simply by browsing through the list of questions asked within that topic. I think this problem could get worse with time because some of the very best and most fundamental questions within each topic were asked in the early days of Mathoverflow, and these become less and less prominent as more new content flows in.

I think this is a general feature of StackExchange-based sites. Other Q&A sites, such as Quora, seem (to me) to do a better job of surfacing good content from all time and have a better browsing experience; on the other hand, they don't do that good a job in terms of getting questions answered quickly and precisely.

I'm curious if others agree with my assessment and what ideas people have on making the site easier to browse.

At the moment nothing can easily be done except to move to the next version of the Stack Exchange platform. One of the advantages to moving is that the next version has an API, which allows programmers to access MO content and display it in a different way.

I think vipul's proposal is worthwhile to consider even if not feasible in the near future. The design of MO has clearly enabled more widespread participation than other participatory math websites, such as Tricki or the polymath projects ( http://gowers.wordpress.com/2010/09/24/is-the-tricki-dead/ ). However, the type of participation on MO, short questions and answers, is limited in scope. I think there is considerable potential to harnessing the content that is produced on MO and extending it into a better reference. At the moment the way in which MO acts as a reference is only via search, but there is the potential to organize the content in a way that it might be more accessible and valuable.

As an example, look at the links between posts on MO, there are many (to give credit it was suggested to me by other MO members). What this shows is that users are already trying to create an organization and structure by connecting MO posts, but it is being done ad hoc by individuals linking between questions. A large organization of the content could be done more explicitly with perhaps little additional work. Here is the largest connected graph:

http://homepage.psy.utexas.edu/homepage/students/Tausczik/Yla/images/MOLinksBetweenQuestions.pdf

Connections were drawn from the original question that was being linked to to the question that linked to it (either in the question statement or comments). The vertex is labelled with the question author and the post id. Grey vertices are soft, advice, or education questions, green vertices are not.

@Gerhard I posted the graph too fast. My labels were off by one in the graph. I fixed the graph posted at the above link. Your question 31337 was first mentioned by you and then cited Will Jagy in response to a question by Umbra, number 72052.

@Joseph O'Rourke Yes you are right in how the arrows are constructed. I think there is justification for doing it either way. Really a link represents at least two connections. Some links were constructed because the question asker was inspired by an earlier question. Some links represent related (or nearly identical) questions that others referenced once a question was posted. Conceptually I was more curious about links in which one question inspired another. For those I think it is justified to have to have the arrows represent time or influence (and be backwards). For the other links I think conceptually double arrow might make more sense. I'm not making too much of the graphs. I think more work (like separating distinct types of connections) would be needed before deep observations could be drawn from them. I think they are interesting, because they show that a lot more could be done to organize and represent MO content.