While it is possible to define ANT in a series of abstract bullet points to do so is to miss most of the point. Instead it explores and theorises the world through rich case studies. This means that, like symbolic interactionism, for ANT words are never enough: you need to practice it. In this paper we work empirically, drawing on an ANT-inflected ethnography of Norwegian salmon farming, and also dialogically. We do this because we want to show that for ANT theory is created, recreated, explored and tinkered with in particular research practices. Indeed, ANT is probably best understood as a sensibility, a set of empirical interferences in the world, a worldly practice, or a lively craft that cherishes the slow processes of knowing rather than immediately seeking results or closure. In particular it is sensible to materiality, relationality, heterogeneity, and process. At its best it understands itself as working in the world to create analytical contexts; but also on the world, to articulate and press particular contexts and their politics. As a part of this it explores the contingencies of power, generating tools to undo the inevitability of that power, while working on the assumption that other and better worlds are possible.