International audience Beginning with a survey of the shortcoming of theories of organology/media-as-externalization of mind/body—a philosophical-anthropological tradition that stretches from Plato through Ernst Kapp and finds its contemporary proponent in Bernard Stiegler—I propose that the phenomenological treatment of media as an outpouching and extension of mind qua intentionality is not sufficient to counter the ‘black-box’ mystification of today’s deep learning’s algorithms. Focusing on a close study of Simondon’s On the Existence of Technical Objects and Individuation, I argue that the process-philosophical work of Gilbert Simondon, with its critique of Norbert Wiener’s first-order cybernetics, offers a precursor to the conception of second-order cybernetics (as endorsed by Francisco Varela, Humberto Maturana, and Ricardo B. Uribe) and, specifically, its autopoietic treatment of information. It has been argued by those such as Frank Pasquale that neuro-inferential deep learning systems premised on predictive patterning, such as AlphaGo Zero, have a veiled logic and, thus, are ‘black boxes’. In detailing a philosophical-historical approach to demystify predictive patterning/processing and the logic of such deep learning algorithms, this paper attempts to shine a light on such systems and their inner workings à la Simondon.