I think the "release often" philosophy is a good one, but too often and you just open yourself up to "this doesn't work, that's wrong, this is broken etc"
I think knuckle down, code a bunch of good stuff then show it off. Of course you'll still get the complaints, but it'll be a lot clearer to a lot more people how it works and the potential it has.
All that said, all they've shown us here is a very simple demo, we have no understanding of the underlying message passing protocol and its robustness. It does look promising, but looks of course...
And I guess early means that you don't need to wait until your code-base is feature complete or at 1.0 to release the code.
By the same token I think you're allowed to do what the Diaspora guys are doing and lay the foundations first, set the tone of the project and the general direction and style.
I hope these guys do well but if it ain't them it'll be someone else cuz peer to peer federated social networking is in the near future of social networking and I'd much much much prefer if it was open source than not.
The number one rule of software engineering. Your code sucks.
Of course, there's still the issue of actually countering Facebook with a more organic social structure. If you will excuse my carting out of this fine dead horse once again: http://interuserface.net/2010/05/how-diaspora-can-succeed-mo...