Data, today, constitutes the ground for contemporary work in artificial intelligence, and for what is known as Web 2.0 (Social Media, Netflix and other streaming services, Uber and other platform-based apps). Much of this data is generated in the interaction between the user and the interface, at the edge of various screens. While the Web 2.0 apps draw on data to produce objectified knowledge of the user of the app to generate actionable insights that optimize some business parameter, AI models that have captured public imagination draw on data to extrapolate outputs analogous to what the models ‘learn’ from the data input to them during the training phase. Neither web-apps nor AI systems in the past, however, had much to do with data. Today’s computational systems are structured by a new design paradigm, with data at the heart of it all. This chapter is an anthropological critique of computational data systems and argues that these systems operate by the extraction of what I call the surplus value of sociality. I illustrate this process, taking as a case, the Walnut money manager, an app that became popular following the demonetization of high-value currency notes in India. I also argue that an anthropological critique of big data must pass through a critique of the production of anthropological knowledge. Finally, I show the distinction between anthropological modes of knowledge production, and that of big data systems and AI models. My epistemological position in this chapter is marked by a critical view of theory from the south.