tuned via the feedback of users looking for and finding similar results
this is the part that kind of appealed to me -- the idea that it might have a way to look at what you are searching for and suggest new community-built searches which could get you closer to what you were really looking for.
most of us have developed a kind of searching "skill" where we know lots of little tricks to get a search engine to return what we want and bypass many irrelevant results. but there are always cases where you can't find what you want and you could benefit from some domain expert knowledge about how to refine your search to get what you want.
This is why I used to love AOL's search. It had a feature like that, where it gave a list in a sidebar of similar and related searches that you could click if you didn't easily find what you were looking for.
AOL's results were all from Google, but it had a few extra features that really enhanced the standard way of searching.
In addition to that one, they had a 'clipper' that allowed you to mark particular results as you went looking down the page and when you moved to the next page, the clipped results would follow you, being placed in a top section of the page, above the 2nd page's results. (you could also 'unclip' them if you later changed your mind)
You could do this for page after page till you had one giant page of what you found to be the most relevant links for what you were looking for.
Then you could just save the page to your hard drive and/or send the page to someone else that might have needed it. It made sharing and saving large batches of good results a snap.
Then they changed it all and became almost identical to Google in every way they could, doing away with all the great power search features that made it unique and I haven't had a use for the regular AOL search, since. (I am pretty much a Google user 100% now)
If there was a way to add some sort of feature like this to searching, in addition to the saved and refined searches, I'd jump all over it.