I'm in the midst of a pretty similar house-hunt to this guy here in Denver. I currently live Downtown and love it. I'm not willing to move outside of the city, so we're focusing on a proper house in one of the city-center neighborhoods.
Proximity to a grocery store is absolutely at the top of my list. I like to buy my food as fresh as possible and generally only buy for the meal I'm cooking that night. So having not only a grocery store, but a high-quality one, near me is incredibly important. The difference of even a block or two is huge when it comes to something you want to walk to on a very regular basis.
Sure I pay attention to things like the crime rate, but in most cities it's a nearly meaningless stat. The actual chances of being involved in a crime are very low in any neighborhood in my city (and this is true of almost every city in the U.S.). I think this is true of people predisposed to urban living. We're simply willing to tolerate a bit of crime risk in exchange for everything else that goes along with it.
From a technology point of view, I imagine it makes online real-estate very difficult to optimize for. There are so many different priorities that people have that you have to find ways to build general data-mining tools that are still actually usable by most folks. That's a tough challenge, one I don't think anyone has come close to cracking. For instance walk scores seem to be the closest to what I want and they tell you virtually nothing. Don't get me started on how we "rank" schools too:)