I've been alpha testing the Freebase Gridworks tool from Metaweb, but haven't been able to talk about it until now. Since they just announced it, I guess it's no longer a secret.
Research scientist David Huynh has been interested in collective data operations since his days at the MIT CSAIL Simile project. You can see collective editing in this 2007 Potluck screencast. Jon Udell called this "stunning." After David moved to Metaweb, his 2008 Parallax demo showed the power of collective operations for browsing Freebase data (and UCG's DERI group forked a SPARQL version called SParallax).
The Gridworks tool is another riff on that same collective operations theme, but this time focused on data cleanup and reconciliation rather than mashups or browsing. There's a lot more to it than what you see in the screencasts (and, naturally, some limitations which are glossed over as well), but while it's still in testing I'll reserve any detailed discussion of features. Suffice it to say though, that the anticipatory buzz in the Twitter-sphere is justified. What remains to be seen is how well they'll follow through on completing the tool, as well as integrating it with the various types of data sources & sinks which are of interest to users.
From a selfish point of view, I'd like to see people use tools like this to contribute to the availability of cleaned up public data sets rather than just using it to clean their private data silos. Of course, convincing people to do that is a much bigger problem -- one which the whole Linked Data / Semantic Web community has yet to come up with a compelling answer for.