] I'm not gonna be satisfied until the computer knows what i
] want to look at based on correlating its semantic map with
] whatever i'm working on. If I happen to be reading a document
] about global warming affecting coastal sea bird populations, i
] want the computer to know that and be looking for related
] information before i ask it, so when i do ask for more, it's
] right there, sorted by subtopics at multiple levels, coded for
] relevance and preferably with a prefetched version of the
] resources already stored locally. -k]

Timball talk about something like this with me once. Some Linux/Gnome project or something.

Basically each X app had its own kind of RSS feed, about what was currently being "done" on that app. A central program simply compiled this info, and would look for relevant info, be it on your computer (meta info is your friend!), or do a google query. Tim said this thing got demoed at some Linux conference (don't know which one) and that they ran into some fundamental barrier with the number of searches 1 IP can send to google in a certain time span.

Who knows how vaporware this thing is, but its sounded cool.

Jesse Ruderman: Experience Google's new look