In particular, I am interested in personal information management. I maintain PlannerMode, an organizer for Emacs. I use it to blog (http://sacha.free.net.ph). My Emacs setup allows me to hit a keystroke to remember anything from almost anywhere. It automatically captures context and allows me to jump back to that. The source pages are on my hard disk, so I regularly regexp-search through them. This lets me use my blog to remember things.
People are surprised to find my task list, schedule and notes online. They ask me why I don't mind the lack of privacy. I find that publishing my notes works to my advantage. Almost every other day, someone writes in with a bit of info or a tip that'll help me finish a task. Plus, publishing my task list shames me into procrastinating less.
I'd like to work on making it easier to retrieve information from my blog and my e-mail. The Remembrance Agent (http://rhodes.www.media.mit.edu/people/rhodes/RA/) is great for indexing mail and a few other resources. It brings up entries that are relevant to the words around your cursor. You've probably run across the software and related papers—if not, check it out!
I'm also interested in XML. I hacked in RSS export for my blog. It autocategorizes my posts based on a regular expression. A minor modification will let me define arbitrary rules.
Answering JP, someone else on the technical groupblog:
I like the idea of RSS syndication because I want to aggregate and search blogs without having to scrape HTML. With either a simple regexp search or search-engine-like relevance ranking, finding information in other people's knowledgebases becomes much easier. If I download feeds, I can even search offline. Push on creation is useful, but people still need a way to get to old posts.