One thing I always think about is in 100 years (or 1000 years), will we still be opening a web browser, typing a few words into the same plain Google text box, and hitting the "Search" button to get a page with 20 blue links and a bunch of ads? I doubt it. So there's definitely a better way to do search, we just haven't discovered it. My personal opinion is that some combination of social search (i.e. FriendFeed) plus human-powered search (i.e. Wikia/Mahalo) plus semantic search (PowerSet) will be involved in the next evolution. Of course, if I knew exactly what that looked like, I'd be on a beach right now instead of hanging out here on Hacker News. :)
Good discussion by the way.
Basically, you'd get results back that are from conversations posted by your friends, side-by-side with traditional results from a search engine (using something like Yahoo's BOSS). Then, if you still didn't get the answer you are looking for, you can broadcast the original query out to your online contact list to see if anyone within your social network knows the answer. You wouldn't get instant results, but it would likely be a very good, trusted result. This probably works better for things like opinions (what's the best Chinese restaurant in Seattle?) vs. a fact-based search like "C# string replace".
I would seriously avoid trying to take on Google or MS or Yahoo by trying to out-index them. Cuil has millions in the bank and some of the world's foremost experts in search at the helm, and things aren't looking all that bright so far. There are lots of problems to solve in the world and lots of approaches on how to solve them - solving search by out-indexing Google should probably be pretty darn low on your "problems to solve today" list.