@skua @aires
The results are weighed for accuracy, and of the top N of those, they are reweighed by content, and presented in that order. It isn't the most accurate thing, not yet, but the web as a whole is weighted for recency, not content, at the moment.

Crawling is just a time factor, really. I'm not spending much on the project, so I only have a single multi-threaded crawler

I haven't implemented keywords like "not", because I'm experimenting with a few ways of filtering. They should help