The dynamics of even the simplest cognitive process under the synergistic influences of fog and friction approaches the very borderlands of contemporary mathematics. One implication is that there can never be a functional ‘Artificial General Intelligence’ (AGI) in the current, all-encompassing, Tech Bro sense: each and every successful autonomous cognitive system must be, first, embodied in order to correct itself—to learn from a particular embedding environment, and second, carefully tuned to—trained on—that particular environment. The realms of hallucination and panic that bookend the standard Yerkes-Dodson inverse-U of aroused cognition can only be confronted by a learning dynamic that necessitates real-time embodiment. Here, we expand consideration of these matters to multidimensional resource rates and to Weber-Fechner Laws in first and second orders of approximation. Dreyfus was more than right.