If every research lab ran a blog-type setup where they published their findings (along with any other updates), the whole field could be revolutionized. The general public would have much more direct access to what the researchers are working on. We could even add the ability to donate to the researchers directly. This would hopefully foster collaboration as well. Instead of waiting for someone to come out with a paper to find out what they've been up to (unless you have a very close relationship with the particular researcher), you can just read their tweet/blog post/etc a la "tried sample X, results negative, but this is weird..."
Combined with a research semantic markup (think OpenGraph but for scientific concepts), this could be linked into searchable databases. Peer review could be built in, maybe even via public key crypto: "this article is signed by 17 trusted researchers".
This saves everyone money: no more having to subscribe to expensive research publications. You could have for-hire scientific editors built in as a service too. Every time you publish a paper you run it by an editor, but the editor now works for you; or you don't, and you just publish the paper and let the world decide if it is any good.
The platform itself would probably have to be fairly extensive. We would want it to be distributed so that anyone can run their own system if they choose. We'd also want to have some more centralized hubs of this type, analogous to GitHub/Bitbucket. These might include easy access to for-hire editors, Tex/LaTex support, etc. There should also be a default license for the content. Perhaps an extension of Creative Commons but with specific provisions for the peculiarities of this field.
My premise here is that while there is a whole lot of institutions that attempt to enable collaboration, they do often just get in the way. However, from what I've seen of researchers, they have the same mentality as the software developers: they want to share their findings with the largest possible audience and don't really care about much else. I think if it was easy to do this type of setup many would go along with it. What do others think?
Edit: a nice side-effect of this could be that you don't have to be associated with an academic institution to publish. Currently, I could learn, say, all there is to know about quantum physics by reading all sorts of publications and material that is more or less freely available. I could then theoretically come up with something brand new, but couldn't get my voice heard since I am not a research professor at an institution. However, with this system graduates of the likes of the Khan Academy could have the same access to publishing and peer review.