Christopher Allen dives into the detail of my ecosystem of networks. His analysis suggests some of the more optimal sizes below 12 and 150 with some great anecdotes.
Its important people understand the scaling limits of groups at 12 and 150. I never ventured a guess at optimal group size, as its heavily context dependent. Instead I would suggest that our challenge as creators of social software is to enhance social capital so productive teams can be dynamic while extending their limits with cohesion.