This essay brings up some interesting points about group size and group dynamics. This kinda sorta ties into an idea I'd love to see tried: build a loose group of small highly specialized businesses set up to cheaply collaborate on projects. Each small business would be about 3-8 people and 3 or 4 of them (say a programming shop, a graphical design shop and an electrical engineering shop) would work closely together on a project with a few more acting as support (say a fab shop, a legal office and a secretarial service).
The essay at the top lends a bit of empirical support to this idea based on group sizes effecting how well a group will work together. As for this actually working, these people seem to think something in the same ballpark is a good idea.