We don't just want to web to be smart, we want to have the information on our desktops to be useful too.
The Haystack from MIT is an all RDF tour-de-force (clear down to it's programming language in RDF Adenine) that aims to integrate every kind of information and make everything right-clickable.
The Haystack client stalled because it was too big, slow, and fragile. Not surprising for a prototype. That led to a refactoring into Eclipse plug-ins.
Another reaction to Haystack's monolithic approach is SIMILE's Piggy Bank, which embeds the semantic engine into a Firefox extension.
An interesting aspect of Piggy Bank's architecture is that in order to make using Java for processing and HTML for UI, it uses an embedded web server (Jetty).
Another Firefox extension using the embedded web server design is this JavaDB demo.
Piggy Bank also has a companion, Solvent, which is another Firefox extension that makes some aspects of writing web page scrapers easier.
IRISStanford makes their own stab at Haystack: http://www.openiris.org/
Zoterohttp://www.zotero.org/ is a nifty Firefox extension and free service. Looks like the Piggy Bank + Firefox Scrapbook extension I've had in mind.
They also have a scaper-maker extension ala Solvent called Scaffold.