Vitality as the Efficiency of Self-Paid Self-Modeling

Complexity is ubiquitous, the living rare: stars, hurricanes, crystals, and language models are richly structured, far from equilibrium, yet none pays, through its own dissipation, for its world-model. Closed self-payment separates the vital from the merely complex; here we propose its quantitative operationalization. Vitality requires two conditions: the predictive efficiency of self-modeling — the fraction of retained environmental memory still predicting the future — and self-payment, whereby holding obsolete memory carries a thermodynamic cost, so a corrupted model returns as the system's own decay. Requiring the model and its paying dissipation to share one physical boundary distinguishes the scale from established programs — assembly theory, integrated information, teleodynamics, the free energy principle — each taking one axis seriously, none demanding this identity of boundaries. A two-stage procedure — structural screening by a triad of invariants, then vitality verification — covers six paradigm cases from star to metropolis, with the biosphere as a control limit and a reproducible Escherichia coli chemotaxis computation. The picture is asymmetric: structural potential is nearly universal, closed self-payment loops rare. Consequences follow: extraterrestrial-life searches must detect such loops atop chemical biosignatures; artificial systems' ethical status turns on when one begins paying for its model; the complex–vital boundary falls at positive efficiency under a closed loop. Developed for the stationary regime, it generalizes to non-stationary dynamics in companion work (Andriishin 2026). It is open to refutation along several independent lines, including a predicted power-law dependence of evolutionary-adaptation rate on self-modeling efficiency.