Thermodynamic Realism: A Glossary

by Andraž Đurič

Thermodynamic Realism is a deductive framework that grounds truth, ethics, and the dissolution of the is-ought problem in the physics of persistence, entropy, and information. This glossary defines its core terms. Each entry ends with a pointer to where the term is grounded in the apex paper: axioms are labelled A0 to A3, background empirical premises BP0 to BP3, numbered results T1 to T23 and Corollary C1, and section numbers refer to that paper.

---

Antifragility. In the framework's sense, the capacity of a system to improve its predictive accuracy from exposure to environmental variance, rather than merely withstanding that variance. The framework holds that antifragility, so defined, is a necessary condition of long-horizon persistence in a high-variance environment: variance that is not converted into improved modelling accumulates as divergence and is selected against. The claim is qualitative and does not depend on any quantitative "antifragility index." (T17.)

Axiom. A premise the framework treats as bedrock and does not derive. There are four: the Persistence Tautology (A0), Physicalism (A1), the Second Law of Thermodynamics (A2), and the physicality of information (A3).

Background empirical premise. A fact about our universe that the derivations rely on but that is not derivable from the axioms alone, stated explicitly so that nothing is smuggled. There are four: the low-entropy initial condition (BP0), non-equilibrium existence (BP1), finite accessible resources (BP2), and evolutionary dynamics with multi-scale organization (BP3).

Consistency Tax. The thermodynamic overhead imposed by any mismatch between a system's internal model and the territory it models, whether the mismatch arises from honest error or from intentional deception. The tax is paid even when the system is unaware of the mismatch, and it spikes when the mismatch is actively corrected. (T9.)

Coupling density (K). A quantity, introduced as a modelling construct, that measures the information-theoretic dependence between a higher-level system and the lower-level systems nested within it. Tight coupling (K approaching 1) means the higher-level system can override its lower-level parts, as an organism does in apoptosis or an institution in turnover, to preserve itself. K can be operationalized via transfer entropy. (T16.)

Dark-room limit. The scope boundary of persistence selection. In an environment with no variance and unlimited resources, a false model is never disconfirmed and pays no Consistency Tax, so accuracy is not selected for. The framework applies only where variance is non-zero and resources are finite, which covers the actual universe but not this idealized edge case. (Scope note to T5.)

Domain-bound operator. An operator whose meaning is defined only within a certain domain and is type-errored, neither true nor false but malformed, when applied outside it. The framework's central meta-ethical claim is that "ought" is domain-bound: it has content only relative to an agent with persistence conditions. The illustrative comparison is "what is north of the North Pole?" (T11, §5.1.)

Epistemic overshoot. The accumulation of unmeasured model-territory divergence in a collective system whose telemetry channels have been suppressed, until an exogenous perturbation forces the hidden divergence into the open and the system collapses non-linearly. This is the framework's hypothesis about censorship-driven civilizational collapse, offered as an analogical extension rather than a proven theorem. (T21.)

Epistemic Profit. The free energy a system recovers when it reduces its model-territory divergence and so stops paying the Consistency Tax on that divergence. (T10.)

Epistemic selection. The framework's name for the process by which entropy selects for accurate models. Because inaccurate models waste free energy on prediction error, systems with cheaper and more accurate maps tend, statistically and over many perturbation cycles, to outlast systems with expensive and inaccurate ones. Entropy selects for truth. (See Persistence selection; T5.)

Landauer ceiling. The in-principle upper bound on a system's rate of adaptation, set by its available power divided by the Landauer cost of a bit operation. The bound is very loose, since real systems operate far below it; its role is conceptual, establishing that adaptation rate is bounded by energy budget at all. (T18.)

Map-territory divergence. The mismatch between a system's internal model (the map) and the actual structure of its environment (the territory). Also called model-territory divergence. Divergence generates prediction errors, and prediction errors dissipate free energy. (T4.)

Markov blanket. A statistical boundary that separates a system's internal states from its external environment, marking where the system ends and the world begins. The term is imported from the active-inference literature; the framework adopts it as the vocabulary for the boundary that persistence already requires, and notes that such blankets nest across scales, cells within organisms within institutions. The maximal scale, the universe, has no Markov blanket, because it has no outside. (T15.)

Past Hypothesis. The premise that the accessible universe began in a macrostate of extraordinarily low entropy. Every free-energy gradient available to any system is a portion of that initial condition still discharging. For self-gravitating matter the low-entropy state is the smooth one, which is why the early universe was a tightly wound spring rather than a featureless equilibrium. (BP0.)

Persistence selection. The framework's engine, also called the Persistence Selection Principle. In an ecology of systems competing for finite free energy, systems with lower model-territory divergence waste less energy on error correction and, on average and over many perturbation cycles, outlast systems with higher divergence. It is a statistical tendency, not a deterministic law. (T5.)

Persistence Tautology. The framework's first axiom: a system that fails to maintain the conditions of its own persistence ceases to exist as an observable. It asserts nothing about value. It states only that existence has prerequisites and that what fails to meet them is no longer there to be observed. (A0.)

Perturbationally robust compression fidelity. The framework's definition of truth: the minimal-loss compression of environmental structure that remains adequate for a system's persistence across an expanding range of perturbations. This is a definitional choice, motivated by the framework's selection results but not deduced from them; alternative definitions of truth exist. (T7.)

Predictive Calm. The conjectured felt correlate of Epistemic Profit: the subjective reduction in cognitive load when a system's models track the territory smoothly. The framework marks this explicitly as an interpretive bridge to felt experience, not a derived result, and nothing downstream depends on it. (T10.)

Schema and content. A distinction the framework draws to block a misreading. The ought-schema is universal and identical for every agent: do what sustains your persistence under perturbation over your horizon. The ought-content is the specific action that schema resolves to, and it is agent-relative, fixed by the particular agent's persistence conditions, scale, and coupling. A universal schema with agent-relative content yields no universal prescription and no global optimization target. (§5.1.)

Semantic premise. The single premise the is-ought dissolution rests on: that "ought" is an agent-relative operator whose content is fixed by the persistence conditions of the agent it is indexed to, and that, deployed without such an agent, "ought" claims are type-errored rather than false. The framework defends this premise with a positive argument from the practical function of "ought" rather than assuming it, and identifies it as the framework's principal philosophical exposure. (§5.1, §7.)

Sustained negentropy capacity. A system's capacity to maintain its low-entropy internal structure against dissolution over a horizon. Within the domain of agents, the framework's "ought-fact" for a given agent is the configuration that maximizes this capacity over the agent's actual horizon, a fact that is physically determinate even when it is computationally inaccessible. (T12.)

Thermodynamic Gap. The finite interval between the universe's low-entropy beginning, fixed by the Past Hypothesis, and its dark-energy-driven approach to equilibrium at heat death. The framework applies only inside this interval: before it there is no gradient for selection to act on, and after it none remains. The universe's total organized complexity rises and falls across this interval as a descriptive trajectory, not as anything any system optimizes. (§7.)

---

This glossary accompanies the Thermodynamic Realism apex paper. Licensed CC BY 4.0.