Not a designer but it's my opinion that low contrast text has it's benefits beside looking good. Sure, low contrast is very bad for walls of text such as the o

Not a designer but it's my opinion that low contrast text has it's benefits beside looking good. Sure, low contrast is very bad for walls of text such as the ones on this very site (when you "Ask HN" or "Show HN" something, but DDG's implementation directs my eyes toward the page title. If you are going to have a lot of info show on any one place, you'd better direct the viewer's attention. I would keep the different contrast levels, but would do it differently. Right now my there are several levels. My eyes are directed to page title and the words I searched for. Page title doesn't really help that much for most pages (unless I'm looking for something quite unspecific) and I already know the keywords I looked for. Right now I go to the keyword in a three line summary and have to find where does it fit; I begin reading from the middle, it's more exhausting and potentially slower Page title is less important than the text surrounding the term I looked for and the domain name of the site. Bold that, not the keyword I already know; direct my eyes toward info I want. FavIcon being the only colored element besides Dax stands out too much; put it left of the result like this: [title] [summary] [FavIcon] [summary] [summary] [url, mouseover complete url, link from this page] On a next version: I'd use my indexer to capture keywords for the result document. I would try to cluster the results based on that, so "man top" returns clusters of linux help and of clothing and the user selects what he meant, filtering out the noise. I don't know how taxing that would be, but not every page has to be indexed this way. Even 10% of the index having these keywords would be statistically significant to build the clusters. The keywords from this index would be highlighted in the results. Search could be machine tailored in case the user decide to go deeper than the 10%, so as to return documents likely to belong to the cluster, but not processed into it.