History of search:
1. New company releases new search engine, better than the competition because it pays attention to a previously ignored signal.
2. Seos work out how to game the signal.
3. New search engine is now as crappy as the search engines it replaced.
Really, internet search is like macroeconomics: soon after you understand {how to prevent recessions, how to find the best webpages}, the problem vanishes and is replaced by something even more complex and incomprehensible. Furthermore, this has already happened.
It's almost like a negative network effect. The more people that use a search engine, the more people try to game it, and the less useful it becomes for finding what you want. This seems like it'd make things much easier for new competitors like duck-duck-go to bring a superior product... at least until they gained enough market share to be worth SEO time gaming.
Judging by the votes, apparently it is more valuable to point out why a certain idea might not work, even if all of the alternatives have the same problem, than it is to suggest something with exciting possibilities. Nothing against you, personally -- your comment is quite right and I agree gaming is a major concern.
If I recall correctly, Yahoo tested using this data a few years ago and found the signals not to be as useful as others they used at the time.
Of course, Bing and Google have been working to include more social signals in rankings: http://searchengineland.com/what-social-signals-do-google-bi...
The way I see it, search engines are entirely inclusive, while bookmarking sites are selectively inclusive. For certain ultra-specific queries, I'd much prefer the search engine approach -- I want as much breadth as possible. For other not-so-specific queries, I'd much prefer the result sites to be vouched for by people. Looking one step further, probably for a decent chunked size of my search queries, I'm OK if the number of sites it is searching over is only about 10,000,000 (my estimate for how many sites on Delicious have been bookmarked by more than 20 users).
What I find exciting is that the two can be effectively combined. Take all the normal results that a search engine would show, but give it a boost based on the log of how many times it's been bookmarked. It really is that simple.
I think Google results could vastly be improved if they tied together their normal rankings with a "how many people have bookmarked this URL (and to some extent, URLs from this domain)" metric. I've worked closely with Delicious' set of data, and it's nothing short of incredible. Billions of instances of people categorizing sites and vouching for their quality are going unused in the search space.