In the context of dark times defined by acceleration and compulsive optimization, this article problematizes Artificial Intelligence (AI) as a dogmatic temporal regime. It posits a structural conflict between two logics: back-propagation, formalized by Werbos (1974) and Rumelhart et al. (1986), and front-propagation, characteristic of the human psyche. While the algorithm operates via a reinforcement learning loop that adjusts the future to minimize error relative to an accumulated past (Skinner, 1938; Goodfellow et al., 2016), human experience requires the irreversibility of time and durée for meaning construction (Bergson, 1910; Valsiner, 2014).

The discussion unpacks this antagonism by analyzing the microgenesis of the pause. It is argued that algorithmic urgency fuses the domains of As-Is and As-If, eroding the boundary tensegrity of the Self necessary for ethical imagination (Tateo, 2016; Vaihinger, 1924). Unlike the stochastic variation of generative networks, which achieve similarity without transcending the original model (Goodfellow et al., 2014), human persistent imitation generates genuine novelty through future-oriented affective hypergeneralization (Baldwin, 1892; Valsiner, 2007). The paper concludes by proposing theorizing as temporal resistance: recovering the thickness of the present through zones of semiotic buffering is the ontological condition to sustain political agency.