We introduce a simple two-stage game of endogenous network formation and information sharing for reasoning about the optimal design of social networks like Facebook or Google+. We distinguish between unilateral and bilateral connections and between targeted and collective information sharing. Agents value being connected to other agents and sharing and receiving information. We consider multiple utility specifications. We show that the game always has an equilibrium in pure strategies and then we study how the network design and the utility specifications affect welfare. Surprisingly, we find that in general, targeted information sharing is not necessarily better than collective information sharing. However, if all agents are either "babblers" or "friends", irrespective of whether the network is unilateral or bilateral, in equilibrium, targeted information sharing yields higher welfare than collective information sharing.