Data, Models and Earth History in Deep Convolution: Paleoclimate Simulations and their Epistemological Unrest. Climate and Earth system models are not only used to project future conditions but also to re-construct past climatic events. This is the second in a series of papers¹ that argue that paleoclimatology, the study of climate prior to the period of direct, instrumental measurements, is an epistemically radical field, one which directly abolishes the distinction between data and model, and reconfigures the notion of experiment. In doing so our notion of deep (or geological) and shallow (or historical) time has become convoluted. The present paper shows, first, how the introduction of general circulation models not only shifted the analysis of the complex causes and processes of ancient climatic change toward numerical simulation techniques, but also how this very introduction was accompanied, if not also fundamentally impacted upon, by paleoclimatic questions in the first place. Second, these computer experiments, and the temporal processes and scales they make operative, turn out to be a potent catalyst in bringing about a new sense of the temporality in which our present transition into the Anthropocene unfolds. By discussing the historical development of paleoclimate modeling, and through examining the productive heuristic qualities of its practices, this paper introduces the unconventional and pragmatic episteme by which paleoclimate simulation challenges our fixation with the category of uncertainty.