Although the strongest social relationships feature most prominently in our lives, we also maintain a multitude of much weaker connections: the distant colleagues that we share a coffee with in the afternoon; the waitress at a our regular sandwich bar; or the ?familiar stranger? we meet each morning on the way to work. These are all examples of weak relationships which have a strong spatial-temporal component but with few support systems available. This paper explores the idea of ?Co-presence Communities? - a probabilistic definition of groups that are regularly collocated together - and how they might be used to support weak social networks. An algorithm is presented for mining the Copresence Community definitions from data collected by Bluetooth-enabled mobile phones. Finally, an example application is introduced which utilises these communities for disseminating information.