Every scientist needs to have a collection of papers (s)he reads and might use for reference. For the last research projects, I used a desktop program called JabRef that was decent. Now, I switched to something really promising.

Mendeley is a startup that wants to do with research papers what last.fm does with music: Make it easy and fun to find some that you don't know yet but would probably like. This goal is really, really mouth-watering, since the amount of papers out there seems to grow exponentially. Today, there is no other way of knowing if someone else already did what you want to do by following references from papers and maybe keyword search on Google Scholar. It's a growing concern for researchers.

They want scientists to have a profile on their site and fill a giant pool of data by citing their research interests, uploading and tagging a lot of papers and maybe build social connections. Then, recommendation becomes possible.

Of course, they'd have the hen-and-egg problem every social community project has. Value can only be created when people are there in big numbers, and people only come when value is already there. This is why the Mendeley people decided to start with solving some other problems scientists have, so that users have a reason to come:

• They built a really slick desktop client to organize papers. It runs on every OS, does full-text search and extracts meta data from PDF automatically if you want. I think it's already better than JabRef.
• The client also syncs automatically with your web portfolio. You can have your library with you wherever you want.
• You can open a research project and share papers only within that research group.
• You can always import and export your library (for instance to Bibtex).

They said recently that users have uploaded one million papers already. That's a lot of potential for data mining right there. And did I mention that they have a Last.fm investor on their board?