On streaming sites such as twitch, audiences interact during a live broadcast through the chat – and for many, this communal viewing and the ensuing social interactions are an important reason to engage with live streams (Hamilton et al., 2014, Hilvert-Bruce et al., 2018, Wohn and Freeman, 2020). Twitch chats therefore constitute important third places that feature very fast-paced interactions full of new word formations, emotes, and banter that appear strongly community affirming. Drawing on stance theory, this talk discusses how chatters achieve joint attention and coherence in the chat and manage to construct community during live-gaming spectatorship in the chats of three different gaming streamers. I show how chatters engage in continuous explicit affective and evaluative stancetaking practices that function to focus attention on a shared object while also to synchronise chatters’ affective responses to the unfolding action. This is achieved through the development of in-group language with a complex, evolving system of interjections and emotes at its heart that allow chatters to take nuanced stances at great speed while positioning themselves as competent community members. Through complex practices of repetition and variation, these stances are constructed as shared across the community.