These micro-learnings reduce local uncertainty without disrupting your overall internal model. So free energy is reduced without destabilizing the system epistemic pleasure without cognitive cost