The study of “value” in the sciences of mind has long been trapped in the “evaluation” paradigm: value is regarded as an object to be computed, a function to be optimized. This paradigm cannot naturally explain the everyday fact that the same stimulus possesses opposite value polarities under different homeostatic states, nor can it explain why an organism “turns a blind eye” to food after satiation—a behavior that appears “inefficient” from an external perspective. The Sentient Principle proposes a fundamental alternative: value is not a tool for evaluation but sensation itself. Life is a ceaseless “Sensation-Behavior Loop” system, whose fundamental driving force is “Sensation Discrepancy”—the identifiable imbalance between immediate sensation and Anticipatory Sensation. This paper advances three core claims: value is sensation; value flows in real time; value as sensation can directly drive behavior. These claims are formalized for the first time into an executable selectionist simulation system, SBLB-2.5. The system strictly adheres to selectionist principles: no preset reward function, no channel locking; all learned behaviors arise from the selective stabilization of eligibility trace spatiotemporal co-occurrence patterns by endogenous value signals. Over more than 550,000 steps of continuous operation, starting from a state where all weights are zero, the system exhibits the gradual differentiation of value sensation, satiation-induced anorexia, dynamic regulation of risk preference, and autonomous stabilization of the sleep-wake cycle. Crucially, the behavioral trajectories in the mature stage appear “unintelligible” from an external perspective—the organism sometimes approaches food, sometimes ignores it, and sometimes actively avoids it. Only by simultaneously tracking its internal sensation states (energy, hunger, overfullness, anticipated value, LHA pulses) can these behaviors be given a unified explanation: they are all outward manifestations of the real-time flow of the organism’s value sensation. As the first selectionist formalization of the Sentient Principle, SBLB-2.5 demonstrates a new path for translating first-person sensory claims into an executable third-person simulation system, providing a concrete working model for the sciences of mind to move beyond the computationalist paradigm.