The spontaneous developments of the Web 2.0 taught us how unexpected, rich and widespread new practices and forms of social coordination may be. Applications like the Wikipedia and Facebook illustrate how significant is the role of a computational and social “backdrop” to enable that coordination and, in fact, the “social intelligence” that emerges from it. I would like to argue for the advantage of developing a general-enough social coordination framework to support social intelligence, that it may be built using agreement technologies and that achieving such generality entails a tension between the sometimes antagonistic drives of theory and applications.