There is an intuitive contrast between a situation in which several people each privately know a certain fact, perhaps without knowing that the others know, and a situation in which that fact is clearly out in the open among them, or commonly known. The dominant characterization of common knowledge takes it to involve an infinite hierarchy of increasingly complex representations of what is known by each individual. If Williamson is right about knowledge needing margins for error, then the epistemic power of the group is capped at the level of the lowest performer, and then diminishes with each iteration, ultimately blocking states of common knowledge so characterized from having substantive content. This result is widely thought to be problematic either for Williamson or for the existence of common knowledge, but perhaps the fault lies with the dominant characterization. I argue that Williamson’s basic framework can be used to construct a rival account of common knowledge, in which participants combine their epistemic powers to exceed what any of them can achieve individually, just as several people can together lift a weight that is too heavy for any of them to lift alone. This account is a better fit with empirical data on conversational interaction; it also helps to explain why states of common knowledge are so attractive to us.