Searching can be used to help categorize pages. Any set of pages resulting from a search should automatically be made into a new category of pages and hence remembered for possible later reuse.

After any search, the system knows that these pages, although they may appear in lots of different places, are in the same category when viewed a certain way. By doing the search the user is implicitly saying that there is an important relation between the pages that the system isn't presently capturing (or is capturing but isn't displaying, or is displaying but not in any way that's easy for the user to see).

The system should also keep the results of every search and that set of pages should be made into a new category simply because the user felt it important enough to search for something belonging to that category (and maybe even named and placed if the user wishes).

Because every search ever made is remembered, if that search or similar searches are made again then those pages are still available in a cluster so new searches are faster than old ones. Also those past searches can be examined to derive information about the kinds of things likely to interest the user in future, independent of saving the actual results to perhaps be used another time. This information can also be used to better determine the pages the user is presently searching for, thereby completing the feedback loop.