This essay attempts to draw together a coherent theory of cognitive emergence from Jean-François Lyotard’s speculations on thinking machines and from his later writings on sense and aesthetic experience. The essay situates Lyotard’s arguments about cognition in the context of recent successes of large language model (LLM) artificial intelligence and the persistence of the so-called ‘symbol grounding problem’ in that field. Rather than tracing a developmental line from stimulus at the material substrate to symbolic reasoning, Lyotard envisions a recursive relationship between sense and reason, in which new forms of thought emerge from an agent’s sensuous experience of symbolic categories and logical constructs, even if that experience is just the suffering of waiting for an analogy to form or a paradox to resolve. It is through an agent’s sensuous experience of its own thoughts that it is able to form an internal model of itself inside of time—a crucial ingredient for genuine cognition.