An anecdote: I finished a decent draft of a paper two weeks ago and posted it to my website an hour before submitting it to a journal. The code to replicate it is on github available to anyone who wants to look at it, and the statistical methods I use I've put in a GPL R package. I've already discussed aspects of the paper and some implementation issues with other researchers over email, presented the paper at conferences, and I think people are already writing papers that use the statistic I propose (again, despite the paper having been submitted to a journal for the first time two weeks ago). In the process of discussing the paper (in person, over email, etc) I've already essentially done two rounds of peer review without journal submission.

Maybe you could give details about what you mean by "antiquated idea behind publishing" because I'm pretty sure what I described is not 1970s speed. Now, if your point is that what I've described is not common practice, I agree completely. But that means that people need to be persuaded to use existing technical tools -- I can't see a specific step that's going to be a significant pain point (I'll grant that not everyone can/will use github, but uploading a zip archive of the paper's directory to a personal webpage is easy and is almost as good for code and data dissemination).

Other anecdotes are blogs that announce and discuss new research:

http://johnhcochrane.blogspot.com/2012/10/nber-asset-pricing...

http://johnhcochrane.blogspot.com/2012/09/woodford-at-jackso...

That is legitimate academic discourse, put out by an extremely established and credentialed economist, almost instantly. On blogger.