MxBv

Posted on Mar 10 • Originally published at petronus.eu

Structural Pressure: The Missing Primitive in Long-Horizon Adaptive Systems

#structuralpressure #adaptivesystems #nc25 #creep

Structural Pressure: The Missing Primitive in Long-Horizon Adaptive Systems

Abstract

The existing literature on adaptive systems provides formal accounts of dissipation (entropy), recovery (resilience), active resistance (homeostasis), and consumption of order (negentropy). Yet no standard formalization in adaptive systems theory appears to isolate a phenomenon that every engineer, biologist, and practicing clinician recognizes intuitively: the structural cost of merely continuing to exist under external pressure, without acting, without correcting, and without recovering from a discrete failure event.

This paper identifies and formalizes structural pressure as a missing primitive in adaptive systems theory. Structural pressure is the monotone contribution to accumulated structural burden that arises from passive continuation under external load — even when the system performs no corrective action, issues no control signal, and exhibits no observable change in behavior. The system simply persists, and persistence itself has a cost.

We show formally that structural pressure occupies a gap left by all existing formalizations: it is not entropy (which is isotropic and undirected), not robustness (which concerns performance preservation, not structural cost), not resilience (which requires a discrete recovery event), not homeostasis (which operates through causal feedback), and not negentropy (which is an energetic concept). We provide a formal definition within the NC2.5 framework, prove that sustained pressure implies finite viability horizons even for non-acting systems, prove that load-neutral passive continuation is architecturally impossible for bounded coupled systems, derive falsifiable predictions that distinguish it from all neighboring concepts, and establish a formal mapping to creep in materials science — the closest physical analogue that, to the best of the author's knowledge, has not been systematically transferred to adaptive systems theory.

Crucially, we demonstrate that structural pressure is not a theoretical abstraction awaiting future validation. We instantiate the complete framework in lithium-ion battery calendar aging — a domain with over two decades of empirical data — and show that every theorem in this paper is already confirmed by the electrochemistry literature. The data exists. The phenomenon is measured. What has been missing is the correct architectural interpretation: the electrochemistry community models calendar aging as a separate degradation mode but lacks the framework explaining why behavioral traces are insufficient to detect it, why better cycling protocols cannot eliminate it, and why it represents a fundamentally different class of viability cost than cycle-dependent degradation. This paper provides that framework. The formalization does not discover new physics — it provides the correct structural reading of known phenomena, enabling transfer across domains.

The claim is not that no one has observed this phenomenon. The claim is that, to the best of the author's knowledge, no prior framework has named it, separated it from action, and given it a monotone accumulation law independent of the agent's decision surface. That is what this paper does.

1. The Gap

Consider an adaptive system embedded in an environment that exerts continuous perturbation. The system has a repertoire of corrective actions, but at the present moment it does not act. It issues no control signal, performs no parameter update, makes no decision. It simply continues.

The standard question in adaptive systems theory is: what happens when the system acts? The entire apparatus of control theory, reinforcement learning, robust optimization, and safety engineering is built around this question. Gains are scheduled. Policies are learned. Constraints are enforced. Costs are assigned to actions.

But no standard question asks: what happens to the system when it does not act but the environment continues to press?

The implicit assumption in the existing literature is: nothing structural happens. If the system does not act, its internal state either remains constant (stability) or evolves according to its own unforced dynamics (drift). External pressure, in this framing, is relevant only insofar as it triggers a response. If no response is triggered, the pressure is invisible.

This assumption is false. And its falsity is not subtle — it is the kind of blindness that, once named, cannot be unseen.

A bridge under constant traffic load deforms over decades without any single vehicle exceeding the design limit. A human being in a hostile work environment degrades physiologically without any single confrontation. A satellite in orbit accumulates radiation damage without any single event exceeding its shielding threshold. A software system under sustained adversarial probing loses structural margin without any single probe succeeding.

In every case, the system did not act. The system did not fail. The system continued. And the continuation itself consumed structural capacity.

The literature has a word for what happens when things fall apart: entropy. It has a word for what happens when systems fight back: homeostasis. It has a word for what happens when systems bounce back: resilience. It has a word for what happens when systems consume external order: negentropy.

It does not have a word for what happens when systems simply endure.

This paper provides one. And the evidence that it names something real does not require future experiments. Lithium-ion batteries sitting on shelves at elevated temperature have been losing capacity for decades — measured, published, replicated across thousands of papers — without anyone recognizing that this is the same structural phenomenon as a bridge under constant traffic, a human in a hostile workplace, or a software system under sustained adversarial probing. The electrochemistry community calls it "calendar aging" and treats it as a domain-specific degradation mode. This paper shows it is a universal architectural primitive: the structural cost of passive continuation under load. The data is already there. The framework to read it correctly was not.

2. Existing Formalizations and Their Boundaries

2.1 Entropy

Shannon entropy H(X) = −Σ p(x) log p(x) and thermodynamic entropy S measure the dispersal of order. Entropy is isotropic: it does not care where the pressure comes from. It describes the general tendency of systems toward disorder, not the specific structural cost imposed by a particular external load on a particular system configuration.

A system under structural pressure does not merely disperse. It degrades directionally, along specific structural dimensions determined by the geometry of the interaction between the system and its environment. Entropy, as standardly formalized, does not furnish the architectural primitive needed here: it does not isolate passive load-specific structural capacity consumption at the agent-environment interface. There exist structured nonequilibrium situations where entropic formalism coexists with directional fields, but these do not separate the passive-load contribution to structural burden from the action-dependent contribution — which is precisely the separation that structural pressure requires.

2.2 Robustness

Robustness, in the control-theoretic sense, measures whether system performance is preserved under bounded perturbation. The H∞ norm, structured singular value μ, and input-to-state stability (ISS) all characterize robustness as a property of the input-output map.

But robustness is entirely about the output — whether the system still performs. It says nothing about the internal cost of performing. A robust system may maintain perfect external behavior while its internal structural margin collapses. Two systems with identical robustness certificates may have radically different structural futures because one is under structural pressure and the other is not.

Robustness asks: does the system still work? Structural pressure asks: at what cost does it continue to work?

2.3 Resilience

Resilience, in the Holling sense, is the capacity of a system to absorb disturbance and reorganize while undergoing change. In engineering contexts, resilience typically denotes recovery after a discrete disruption event.

Structural pressure is not a discrete event. There is no shock, no failure, no disruption to recover from. Structural pressure is the continuous background cost of existing under load. Resilience measures what happens after the storm. Structural pressure is what happens during the silence between storms — the slow accumulation of damage from the weight of ordinary continuation.

2.4 Homeostasis

Homeostasis (Cannon, 1929; Ashby, 1956) is the maintenance of internal variables within acceptable bounds through feedback-driven corrective action. The thermostat detects deviation, the effector corrects, and the variable returns.

Homeostasis is a causal action loop. It requires sensing, comparison, and correction. In the NC2.5 framework, homeostasis is explicitly what structural pressure is not. Structural pressure accumulates precisely when the system does not act — when U(t) = 0. Homeostasis describes the cost of correction. Structural pressure describes the cost of not needing to correct and yet still degrading.

2.5 Negentropy

Schrödinger (1944) introduced the concept that living systems maintain order by consuming negentropy — by importing low-entropy energy from their environment. This is an energetic concept: the system pays for its order with a thermodynamic currency.

Structural pressure is not an energetic concept. The structural burden Φ in NC2.5 is not reducible to energy, entropy, or any thermodynamic potential. A system can be energetically stable — adequately powered, thermally regulated, with no energy deficit — and still be under structural pressure. The cost is not paid in joules. It is paid in structural capacity, in the progressive consumption of the system's ability to sustain coherent operation.

2.6 Creep (Materials Science)

Creep is the slow, permanent deformation of a material under constant stress below its yield point. A steel beam under constant load will eventually sag — not because the load increased, not because the beam was struck, but because sustained stress causes progressive internal restructuring (dislocation migration, grain boundary sliding, void nucleation).

Creep is the closest physical analogue to structural pressure. But creep is formalized only for physical materials. To the best of the author's knowledge, no systematic transfer of the concept to adaptive systems as an architectural primitive has been made. The mapping is straightforward — and its absence from the adaptive systems literature is the gap this paper fills.

Property	Creep (Materials)	Structural Pressure (NC2.5)
External condition	Constant stress σ < σ_yield	Constant environmental load P > 0
System response	No macroscopic action	No corrective action U = 0
Observable behavior	Shape appears stable	Performance appears acceptable
Internal process	Dislocation migration	Structural capacity consumption
Accumulation	Monotone strain ε(t)	Monotone burden Φ(t)
Consequence	Eventually: fracture	Eventually: viability loss (τ → 0)
Detection	Only by precision measurement	Only by structural monitoring
Recovery	Irreversible (permanent set)	Irreversible (monotone Φ)

3. Formal Definition

3.1 Setup

Let S be an adaptive system operating in environment E. Let:

Φ(t) denote the accumulated structural burden at time t (monotone non-decreasing)
U(t) denote the nominal intervention at time t
G(t) ∈ (0, 1] denote the coupling efficiency
τ(t) = C − Φ(t) denote the remaining viability budget
P(t) ≥ 0 denote structural pressure from the environment

3.2 Definition

Definition (Structural Pressure). Structural pressure P(t) is a non-negative, environment-dependent quantity that contributes to structural burden accumulation independently of the agent's intervention:

dΦ/dt = f(U(t), G(t)) + P(t)

where f(U, G) ≥ 0 is the action-dependent structural cost (with f(0, G) = 0 for any G), and P(t) ≥ 0 is the structural pressure term that depends on the environment-system interaction but not on the agent's control signal.

Figure 2: The core architectural separation. Action cost f(U,G) and structural pressure P(t) feed into Φ(t) through independent channels. When U = 0 (passive continuation), only the pressure channel remains active — yet burden still accumulates and viability still decreases.

3.3 Key Properties

Property 1 (Action-Independence). P(t) contributes to dΦ/dt regardless of U(t). In particular, when U(t) = 0:

dΦ/dt = P(t)

The system accumulates structural burden from pressure alone, without any intervention.

Property 2 (Monotonicity). Since P(t) ≥ 0 and Φ is monotone non-decreasing:

P(t) > 0 ⟹ Φ is strictly increasing ⟹ τ is strictly decreasing

Even a non-acting system loses viability under positive structural pressure.

Property 3 (Directionality). Unlike entropy, structural pressure is directional. It depends on the specific geometry of the system-environment interaction. The same environment may exert different structural pressure on different systems, and the same system may experience different structural pressure under different environmental configurations.

Property 4 (Non-energetic). P(t) is not reducible to energy expenditure, power dissipation, or thermodynamic entropy production. A system may be energetically stable while Φ grows due to P.

Property 5 (Irreversibility). The contribution of P to Φ is irreversible. No subsequent action can decrease Φ. Once structural capacity is consumed by pressure, it is gone.

3.4 Separability Conditions

The core equation dΦ/dt = f(U, G) + P(t) assumes that the action-dependent cost and the pressure-dependent cost are additively separable. This is not a trivial assumption: in many real systems, environmental pressure degrades coupling efficiency G(t) or alters the cost landscape of action f(U, G). The additive form requires the following orthogonality condition:

Condition (Structural Separability). The action-dependent cost f(U, G) and the pressure term P(t) are separable if and only if:

∂P/∂U = 0 and ∂f/∂P = 0

That is: the agent's actions do not influence the environmental pressure (pressure is exogenous), and the environmental pressure does not alter the cost function of action (the cost of doing something does not change because of pressure alone).

When separability holds, the system's viability budget is consumed by two independent channels: one from acting, one from enduring. This is the clean case. It holds when the structural coupling between environment and system operates through a different channel than the control interface — for example, radiation damage to a satellite (pressure channel) is independent of the satellite's attitude control commands (action channel).

When separability is violated, the system enters a coupled regime where pressure amplifies intervention cost or where intervention modulates pressure. In this case, the interaction term must be made explicit:

dΦ/dt = f(U, G) + P(t) + h(U, P)

where h(U, P) captures the cross-term. The pure structural pressure analysis (Theorems 1–5) holds exactly under separability and provides a lower bound on viability consumption in the coupled case, since h ≥ 0 when pressure amplifies action cost. Separability is therefore not a limitation of the framework but a clean analytical base case from which coupled extensions can be derived.

3.5 Canonical Instantiations of f(U, G)

The advisory analysis identified a gap in the specification of f(U, G). While the framework deliberately avoids prescribing a single functional form — to maintain generality across domains — canonical instantiations clarify the concept:

Instantiation A (Linear scaling): f(U, G) = U / G. This is the Ueff form from the Structural Intervention Cost patent. Effective cost scales inversely with coupling efficiency. Simple, interpretable, sufficient for many applications.

Instantiation B (Quadratic effort): f(U, G) = U² / G. Penalizes large interventions superlinearly. Appropriate when the structural cost of large corrections grows faster than linearly — for example, emergency maneuvers that stress mechanical components.

Instantiation C (Threshold-gated): f(U, G) = 0 when U < U_threshold; f(U, G) = (U − U_threshold) / G otherwise. Models systems where small corrections are structurally free (absorbed by internal elasticity) but corrections beyond a threshold consume viability.

The choice among instantiations is domain-specific and empirically determined. The framework does not prescribe it; it requires only that f(0, G) = 0 (no action ⟹ no action cost) and f ≥ 0 (action cost is non-negative).

4. Theorems

4.1 Finite Horizon Under Pressure

Theorem 1 (Pressure-Induced Finite Horizon). Let S be a system with initial viability budget τ(0) = C − Φ(0) > 0. If P(t) ≥ P_min > 0 for all t ≥ 0, and U(t) = 0 for all t, then:

τ(t) = C − Φ(0) − ∫₀ᵗ P(s) ds ≤ τ(0) − P_min · t

Therefore τ(t*) = 0 for some t* ≤ τ(0) / P_min.

Proof. By definition, dΦ/dt = P(t) ≥ P_min > 0 when U = 0. Integrating: Φ(t) ≥ Φ(0) + P_min · t. Therefore τ(t) = C − Φ(t) ≤ C − Φ(0) − P_min · t = τ(0) − P_min · t. Setting τ(t*) = 0 gives t* ≤ τ(0) / P_min. ∎

Corollary. No adaptive system under sustained structural pressure can maintain viability indefinitely without intervention — even if it never makes a mistake, never receives a shock, and never fails at any task.

This is the core result. It says: you can die from standing still.

4.2 Pressure Distinguishes Structurally Non-Equivalent Histories

Theorem 2 (Structural Non-Equivalence). Let S₁ and S₂ be two copies of the same system, both applying identical interventions U(t) with identical coupling G(t), but operating under different structural pressures P₁(t) ≠ P₂(t). Then:

Φ₁(t) − Φ₂(t) = ∫₀ᵗ [P₁(s) − P₂(s)] ds

and the systems diverge in viability despite identical behavior.

Proof. Since U and G are identical, f(U(t), G(t)) is identical. The only difference in dΦ/dt is P₁ − P₂. Integration gives the result. ∎

Corollary. Two agents that look identical from the outside — same actions, same outcomes, same performance metrics — may have fundamentally different structural futures if they operate under different pressures. No behavioral metric can detect this. Only structural monitoring can.

4.3 Pressure as Forced Regime Exit

Theorem 3 (Pressure-Induced Regime Transition). If structural pressure accumulates to the point where τ(t) falls below the admissibility threshold τ_adm for the current inertial regime R_i, the system must transition to an actively regulated regime R_a — even though no action has failed, no error has occurred, and no policy has changed.

Proof. The admissibility predicate Adm(R_i) requires τ > τ_adm. Since τ is strictly decreasing under P > 0, there exists t' such that τ(t') = τ_adm. At t', the inertial regime becomes inadmissible. The system must either transition to R_a or cease operation. ∎

Corollary. Structural pressure can force regime transitions in the absence of any failure event. The system is not responding to a problem — it is running out of the capacity to continue without responding.

4.4 Behavioral Indistinguishability Under Unequal Pressure

Theorem 4 (Observational Non-Equivalence). Let S₁ and S₂ be two systems generating identical action traces {U(t)} and identical task outputs {y(t)} over interval [0, T]. If P₁(t) ≠ P₂(t) on a set of positive measure within [0, T], then Φ₁(T) ≠ Φ₂(T), and therefore τ₁(T) ≠ τ₂(T).

Proof. Since U₁ = U₂ and G₁ = G₂ (same coupling by construction), f(U₁, G₁) = f(U₂, G₂) at every t. Then Φ₁(T) − Φ₂(T) = ∫₀ᵀ [P₁(s) − P₂(s)] ds ≠ 0 when the integrand is nonzero on a set of positive measure. Since τ = C − Φ, the viability budgets diverge. ∎

Corollary. Structural pressure is not observable from the action-output trace alone. No behavioral metric, no performance test, no trajectory comparison can detect unequal pressure. Only structural monitoring — measurement of Φ or τ directly — can reveal the difference. This is not a limitation of current measurement technology; it is a structural property of the architecture. Behavioral traces are provably insufficient.

4.5 Structural Pressure vs Drift

This subsection addresses what is likely the strongest objection to the claim that structural pressure is a primitive: that it is merely another name for drift, or a decomposition term within existing burden dynamics.

The distinction is precise:

Drift is the accumulated irreversible structural memory of realized structural deformation. Drift arises from the residual asymmetry of corrective action: each intervention leaves a trace that does not fully reverse. Drift requires that something happened — an action was taken, a correction was attempted, a policy was executed. Drift is the scar tissue of having acted.

Structural pressure requires none of this. Pressure accumulates when U(t) = 0 — when the system has done nothing. No action was taken, no correction attempted, no residual asymmetry generated. The system simply continued to exist under load, and continuation consumed structural capacity.

The formal distinction is sharp:

Drift may be zero-increment under absence of directed adaptation (if U = 0 and no internal dynamics produce deformation, drift does not grow). Structural pressure remains positive under load regardless.
Drift is deformation-memory: it records what happened to the system as a consequence of its own actions. Pressure is load-presence cost: it records what happens to the system as a consequence of the environment's weight, independent of any action.
Drift does not require external load. A system may drift from its own internal dynamics. Pressure requires external load — it is the cost of the environment pressing on a system that does not press back.
Pressure is structurally prior to drift. Pressure may force the system out of an inertial regime into active regulation (Theorem 3). Once in active regulation, the corrective actions produce drift. Therefore: pressure precedes drift. Pressure creates the conditions under which drift-producing responses become necessary.

The relationship is sequential, not synonymous:

P(t) > 0 → τ(t) decreases → regime becomes inadmissible → forced intervention U(t) > 0 → drift accumulates

Pressure is the cause. Drift is the downstream consequence of the response to pressure. Collapsing them is like collapsing gravity with falling damage — they are related, but one is the field and the other is the impact.

4.6 Pressure Absorption Impossibility

Theorem 5 (Impossibility of Load-Neutral Passive Continuation). Let S be a long-horizon adaptive system with bounded viability budget τ = C − Φ, operating under sustained nonzero external load (P(t) > 0 on a set of positive measure). Suppose S claims that passive continuation under this load incurs no burden increment (dΦ/dt = 0 when U = 0). Then one of the following must hold:

(a) P(t) = 0 almost everywhere — the load is not actually sustained, contradicting the premise;

(b) The system is perfectly shielded — there exists no structural coupling between the environment and the system's viability-relevant internal state, meaning the system is structurally isolated from its environment;

(c) The system has unbounded viability budget (C = ∞) — the system does not belong to the class of bounded long-horizon adaptive systems;

(d) The system's viability model is incomplete — it does not account for load-presence cost, and the claim of zero passive burden is an artifact of model omission rather than a structural property.

Proof. By the definition of structural pressure, P(t) > 0 implies dΦ/dt > 0 when U = 0, provided that structural coupling between environment and system exists and the viability budget is finite. If dΦ/dt = 0 is asserted despite P > 0, then either the coupling does not exist (case b), the budget is unbounded (case c), the load is not sustained (case a), or the viability model fails to represent the load-presence contribution (case d). No fifth case is available: under finite budget, real coupling, and sustained load, zero passive burden is architecturally impossible. ∎

Corollary (Free Endurance Is Impossible). In any long-horizon adaptive architecture with bounded viability budget, exposed to sustained nonzero external load through a real structural coupling, passive continuation must consume structural viability even when no corrective action is executed. An architecture that claims otherwise is either structurally isolated from its environment, unbounded, or incomplete.

This is the impossibility result that elevates structural pressure from a useful bookkeeping term to a necessary architectural primitive. It says: if you are a bounded system, and the environment really presses on you, and you are really coupled to it — then standing still has a cost, and no architectural choice can make that cost zero. You can reduce it. You cannot eliminate it.

4.7 Pressure as the Necessary Shadow of Bounded Continuation

Structural pressure is not merely one more term in the burden equation. It is the dual of admissible continuation under load.

Consider the claim: "the system can continue operating indefinitely under sustained load without acting." Theorem 5 shows this claim is structurally inconsistent for any bounded, coupled, long-horizon system. Therefore, admissible continuation under load necessarily implies passive viability expenditure. The expenditure is not an optional modeling choice — it is the unavoidable structural cost of maintaining orientation under sustained load.

This connects structural pressure to the deepest layer of NC2.5: continuation under load is never neutral. Any maintained structural coherence under sustained external pressure has a viability price. Zero viability expenditure under nonzero load implies either zero coupling (trivial system boundary) or unbounded resources (idealization outside the adaptive class). For real systems in real environments, structural pressure is not a feature of the model — it is a feature of bounded existence.

4.8 Pressure vs Exposure

Not every interaction with an environment constitutes structural pressure. The distinction must be precise:

Exposure is the general condition of being situated in an environment. A system that passively observes its environment without structural loading is exposed but not under pressure. Exposure without structural coupling to the viability-relevant internal state does not consume viability.

Structural pressure is exposure that consumes viability under passive continuation. It requires that the environment's load structurally couples to the system's viability-relevant state through a channel that does not require the agent to act. The coupling may be physical (mechanical stress, radiation, thermal load), informational (sustained adversarial probing, noise floor), organizational (regulatory burden, compliance overhead), or interactional (sustained misalignment between system orientation and environmental demand).

The test is simple: if the system is exposed to the environment and does nothing, does Φ increase? If yes, the system is under structural pressure. If no, the system is merely exposed.

This prevents the term from diluting into a synonym for "being in an environment". Structural pressure is specific: it is the viability-consuming component of exposure under passive continuation.

5. Why This Was Missed

The absence of structural pressure from the literature is not accidental. It follows from three deep assumptions embedded in the adaptive systems paradigm:

Convention 1: Dominant frameworks attach cost to action. Control theory, RL, and optimization all assign costs to what the system does. The standard formulations treat inaction as the zero-cost baseline. The idea that not doing has a structural cost is foreign to frameworks where the cost functional is defined over the agent's action trajectory.

Convention 2: Standard formulations treat constancy as implying structural stasis. If the system is at equilibrium and the environment is constant, the system is assumed to remain at equilibrium. Structural pressure violates this: the environment is constant (the pressure persists), the system does not act, and yet it degrades. Event-centered degradation frameworks fail to represent this because they require a triggering event.

Convention 3: Degradation is typically modeled as event-contingent. Resilience theory, fault tolerance, and safety engineering are built around events — shocks, failures, violations. Structural pressure is not an event. It is the absence of events combined with the presence of load. It is the nothing that costs something.

These three assumptions conspire to make structural pressure invisible. If cost requires action, if constancy implies stability, and if degradation requires events — then a system that does nothing, in a constant environment, without any event, cannot be degrading. And yet it is.

6. Falsifiable Predictions

6.1 The Idle Degradation Test

Prediction 1. An adaptive system under sustained environmental pressure, performing no corrective actions (U = 0), will exhibit strictly decreasing viability budget τ(t). Conventional theory predicts τ(t) = constant when U = 0.

Test. Deploy two identical agents. Agent A operates in a benign environment (P ≈ 0). Agent B operates in a hostile environment (P > 0) but is prohibited from acting. Measure structural indicators over time. Under the structural pressure model, Agent B degrades; under conventional models, Agent B is unchanged.

6.2 The Identical-Action Divergence Test

Prediction 2. Two systems performing identical actions under identical coupling but different pressures will diverge in structural state despite identical behavioral traces.

Test. Deploy two identical agents executing the same fixed policy. Place them in environments with different structural pressures but identical task demands. After N episodes, measure structural indicators. Under the structural pressure model, they diverge; under conventional models, they are identical.

6.3 The Forced Transition Test

Prediction 3. A system under sufficient structural pressure will be forced into regime transition without any failure event, error signal, or performance degradation.

Test. Deploy a system under gradually increasing structural pressure with a fixed policy that satisfies all performance criteria. Observe whether regime transition occurs before any performance metric degrades. Under the structural pressure model, it does; under conventional models, it should not.

7. Implications

7.1 For Adaptive Systems Design

If structural pressure is real, then no amount of control engineering, policy optimization, or safety constraint can guarantee indefinite viability. A system that does everything right — optimal actions, perfect performance, no failures — will still lose viability under sustained pressure. Design must account for the structural cost of existence, not only the structural cost of action.

7.2 For Long-Horizon Deployment

Systems designed for extended deployment — satellites, infrastructure controllers, long-running software agents, autonomous vehicles — must incorporate structural pressure into their viability models. The question is not only "how long can the system perform?" but "how long can the system exist under this pressure without losing the capacity to perform?"

7.3 For the Creep Analogy

The mapping between material creep and structural pressure opens a rich vein of formal transfer. Creep has three stages (primary, secondary, tertiary); structural pressure may exhibit analogous phases. Creep has a Larson-Miller parameter relating temperature and time to failure; structural pressure may admit analogous parametric collapse. The materials science literature on creep rupture, with nearly a century of experimental data, becomes a formal source of hypotheses for adaptive systems.

7.4 Pressure Is Not Solved By Better Policy

This is perhaps the most consequential implication.

Optimization can reduce action cost f(U, G). A better policy can achieve the same task with less intervention. A more efficient coupling can reduce Ueff. All standard engineering responses — better control, better learning, better planning — operate on the f(U, G) term.

None of them touch P(t).

Structural pressure is not a property of the agent's decisions. It is a property of the environment's weight on the agent's structure. No improvement to the policy, no refinement of the reward function, no reduction in control effort, and no increase in coupling efficiency can eliminate the pressure term. The only responses to structural pressure are architectural: increase the initial budget C, reduce exposure to pressure (change the environment or the agent's structural interface with it), or accept a finite horizon.

This means that structural pressure requires architectural accounting, not merely better control. It is a fundamentally different engineering concern — one that cannot be delegated to the optimizer, the planner, or the learning algorithm. It must be recognized as a separate budget line in the viability model, or it will silently consume the system from beneath an otherwise perfect behavioral surface.

Figure 3: The irreducible cost of endurance. Even when no action is taken, passive load P(t) feeds into Φ(t), drives τ(t) downward, and eventually forces regime inadmissibility. Resistance is not free.

8. Canonical Domain Instantiation: Lithium-Ion Battery Degradation

To demonstrate that structural pressure is not merely an abstract concept but a measurable physical phenomenon already producing data that existing frameworks misinterpret, we instantiate the full framework in a concrete engineering domain: lithium-ion battery calendar aging.

8.1 The Domain

A lithium-ion battery cell sitting on a shelf, fully charged, at elevated temperature, not being cycled, degrades. Its capacity fades, its internal resistance increases, and its ability to deliver rated performance diminishes — all without a single charge-discharge cycle. This is calendar aging, and it is one of the most extensively studied phenomena in electrochemistry.

The standard electrochemical explanation involves solid-electrolyte interphase (SEI) growth, lithium inventory loss, and electrode passivation. These are internal structural processes driven by the sustained chemical potential gradient between electrode and electrolyte — a constant environmental load.

8.2 The Mapping

NC2.5 Variable	Battery Instantiation
System S	Battery cell
Environment E	Temperature + state of charge + chemical environment
Structural burden Φ(t)	Capacity fade + resistance growth (cumulative)
Viability budget τ(t)	Remaining useful life (RUL) = rated capacity − Φ(t)
Intervention U(t)	Charge-discharge cycles (cycling stress)
Coupling efficiency G(t)	Coulombic efficiency × thermal management quality
Structural pressure P(t)	Calendar aging rate at current temperature and SOC
f(U, G)	Cycle-dependent degradation per unit of charge throughput

8.3 Verification Against Theorems

Theorem 1 (Finite Horizon). A fully charged battery at 45°C with U = 0 (no cycling) loses approximately 5–8% capacity per year from calendar aging alone. Starting from rated capacity C, the battery reaches end-of-life (τ = 0, typically defined as 80% of rated capacity) in approximately 2.5–4 years of pure shelf storage. This is exactly the prediction: P > 0, U = 0, τ → 0 in finite time.

Theorem 2 (Structural Non-Equivalence). Two identical batteries executing identical charge-discharge profiles, but stored at different temperatures between cycles, will diverge in remaining useful life. The battery stored at 45°C will age faster than the one stored at 25°C, despite identical cycling histories. This divergence is entirely due to the pressure term P(T, SOC), not to the cycling term f(U, G).

Theorem 4 (Behavioral Indistinguishability). During active cycling, both batteries deliver identical voltage profiles, identical charge throughput, and identical apparent performance. The performance metrics — charge capacity, discharge rate capability, voltage under load — are identical or nearly so in early life. The divergence in Φ is invisible from the behavioral trace. Only direct measurement of internal resistance (EIS) or capacity fade (reference cycles) can detect the difference.

Theorem 5 (Absorption Impossibility). No battery chemistry, no thermal management system, and no charge protocol can reduce calendar aging to zero while maintaining a charged, room-temperature cell with electrolyte contact. Calendar aging rate can be reduced (lower SOC, lower temperature), but elimination requires removing the chemical potential gradient — which means discharging to 0% or removing the electrolyte, both of which are trivializations (the battery is no longer a functioning battery).

8.4 What This Demonstrates

The battery domain demonstrates something stronger than a useful analogy. It demonstrates retrospective empirical confirmation of a theoretical framework.

The electrochemistry community has been measuring structural pressure for over two decades — they just call it "calendar aging" and model it as a domain-specific degradation mode alongside "cycle aging". They have both terms of our equation: calendar aging = P(t), cycle aging = f(U, G). They measure them separately. They publish them separately. They predict remaining useful life using models that combine both terms additively — exactly as dΦ/dt = f(U, G) + P(t) prescribes.

But they lack three things that the architectural framework provides:

First, the explanation of why behavioral traces fail. Battery management systems (BMS) estimate state-of-health from voltage curves, charge throughput, and impedance measurements during cycling — behavioral traces. Calendar aging is invisible in these traces during active cycling. Only dedicated EIS measurements during idle periods can isolate the calendar component. This is Theorem 4 instantiated: behavioral indistinguishability under unequal pressure. The battery community knows this empirically but has no architectural theorem explaining why it must be so.

Second, the explanation of why better cycling cannot eliminate calendar aging. No charge protocol, no temperature management during cycling, and no optimization of the cycling profile can reduce SEI growth during storage. Better policy touches f(U, G); it cannot touch P(T, SOC). The battery community knows this empirically — they measure calendar and cycle aging as independent modes — but has no architectural framework explaining that this independence is a structural necessity, not merely an empirical observation.

Third, the cross-domain transfer. The battery community treats calendar aging as an electrochemistry problem. The structural engineering community treats creep as a materials science problem. The software engineering community treats state-space exhaustion under sustained load as a systems problem. None of them recognize that these are the same architectural phenomenon: the structural cost of passive continuation under load. The formalization in this paper makes the transfer explicit and provable.

This is the value of the formalization: it does not discover new physics. It provides the correct architectural interpretation of known phenomena, reveals why domain-specific explanations are necessarily incomplete, and enables transfer across domains that would otherwise remain siloed.

9. Toward Operationalization

The advisory analysis convergently identified operationalization as the primary gap. This section addresses the three highest-priority implementation requirements.

9.1 Structural Monitoring Interface

Theorem 4 proves that behavioral traces cannot detect structural pressure. Therefore, any implementation requires dedicated structural monitoring — measurement of Φ(t) or τ(t) through channels independent of the action-output loop.

Interface contract:

StructuralMonitor {
    getburden(t) → Φ(t)     // accumulated structural burden
    getpressure(t) → P(t)    // current pressure estimate  
    getviability(t) → τ(t)   // remaining viability budget
    getregime(t) → R          // current operational regime
    isadmissible(R, τ) → bool // admissibility predicate
}

Physical domains: Structural monitoring maps to existing measurement infrastructure — battery EIS for internal resistance, strain gauges for mechanical creep, radiation dosimeters for cumulative exposure, thermal cycling counters for solder joints. The monitoring channel is independent of the control channel.

Digital domains: This is the harder case. The advisory correctly identified that software systems lack obvious "structural sensors". Candidate proxies for Φ(t) in software systems include:

Memory fragmentation index (monotone growth under sustained allocation pressure)
State-space coverage exhaustion (fraction of reachable states already visited, under adversarial probing)
Model confidence degradation under distribution shift (sustained exposure to out-of-distribution inputs)
Floating-point drift accumulation in long-running numerical processes
Connection pool exhaustion rate under sustained request load without corrective scaling

The critical requirement is that the proxy must be monotone under passive load (P > 0 ⟹ proxy increases even when U = 0) and must not be observable from the action-output trace. If the proxy is correlated with behavioral output, it violates Theorem 4's separation and is not a valid structural sensor.

9.2 Discrete-Time Estimation

The continuous-time formulation dΦ/dt = f(U, G) + P(t) must be discretized for implementation:

Φ(t+Δt) = Φ(t) + f(U(t), G(t)) · Δt + P(t) · Δt

When U = 0 (idle periods):

ΔΦ_idle = Φ(t+Δt) − Φ(t) = P(t) · Δt

This directly estimates P(t) from observed burden increments during known idle periods. The estimator is:

P̂(t) = ΔΦ_idle / Δt

When U > 0 (active periods):

ΔΦ_total = f(U, G) · Δt + P(t) · Δt

If f(U, G) can be estimated from the known U and G (using the chosen canonical instantiation), then:

P̂(t) = (ΔΦ_total − f̂(U, G) · Δt) / Δt

This is a residual estimator: pressure is what remains after subtracting the expected action cost. Under noisy measurements, a Kalman filter or exponential moving average over multiple Δt windows provides smoothing.

9.3 The Larson-Miller Transfer

The advisory recommended formal transfer of the Larson-Miller parametrization from creep science. The Larson-Miller parameter (LMP) collapses creep rupture data across multiple temperatures and stress levels onto a single master curve:

LMP = T · (C_LM + log₁₀(t_rupture))

where T is temperature (Kelvin), t_rupture is time to failure, and C_LM is a material constant (typically ≈ 20 for metals).

Transfer to structural pressure: Replace temperature with pressure intensity, and time to rupture with viability horizon:

LMP_adaptive = P_eff · (C_SP + log₁₀(t*))

where P_eff is effective (time-averaged) structural pressure, t* is time to viability loss (τ → 0), and C_SP is a system constant calibrated from empirical data.

If this parameterization holds — and the battery calendar aging data suggest it should, since Arrhenius-type temperature dependence maps directly to pressure-intensity dependence — then the entire century of creep rupture methodology becomes available: master curves, accelerated testing protocols, safety factors, remaining life estimation from partial degradation data.

This is not guaranteed to transfer exactly. But the structural homology between creep and structural pressure (Table in Section 2.6) provides strong reason to expect that the functional form transfers, even if the constants do not. Empirical validation is required.

9.4 Admissibility Threshold τ_adm

Theorem 3 requires an admissibility threshold τ_adm below which the inertial regime becomes inadmissible. The advisory asked: how is this threshold determined?

Option A (Fixed percentage): τ_adm = α · C, where α ∈ (0, 1) is a design-time parameter (e.g., α = 0.2 means the system must transition when 80% of viability is consumed). Simple, predictable, appropriate for systems with well-characterized initial budgets.

Option B (Pressure-adaptive): τ_adm(t) = β · ∫ P(s) ds / t, where the threshold adapts to the average pressure experienced. Under high pressure, the threshold rises (earlier transition); under low pressure, it remains low (longer inertial operation). Appropriate for systems in variable environments.

Option C (Derivative-triggered): τ_adm is not a level but a rate condition: transition when dτ/dt < −r_critical. This triggers on the rate of viability loss, not the absolute level. Appropriate for systems where sudden pressure increases are the primary risk.

The framework does not prescribe a single method. It requires that τ_adm be explicit, documented, and deterministic for a given system configuration.

The literature on adaptive systems has formalized what happens when things fall apart (entropy), what happens when systems fight back (homeostasis), what happens when systems bounce back (resilience), and what happens when systems consume order to survive (negentropy).

It has not formalized what happens when systems simply endure.

This paper names the gap: structural pressure — the monotone structural cost of inertial continuation under external load. We show that this concept is irreducible to entropy, robustness, resilience, homeostasis, negentropy, or drift. We provide a formal definition, prove that sustained pressure implies finite viability horizons even for non-acting systems, establish that pressure is provably undetectable from behavioral traces alone, show that pressure forces regime transitions without failure events, prove that load-neutral passive continuation is architecturally impossible for bounded coupled systems, and derive falsifiable predictions that distinguish structural pressure from all neighboring formalizations. We further show that pressure is structurally prior to drift — it is the field, not the impact — and that no improvement to the agent's policy, control law, or optimization can eliminate the pressure term from the viability equation.

The paradigm shift is not in the observation — every practicing engineer knows that standing still under load has a cost. The shift is in the formalization: giving this cost a name, a monotone accumulation law, a connection to viability, a separation from the agent's decision surface, and a proof that better policy cannot remove it.

You can die from standing still. Now there is a theorem for it.

10. Conclusion

It has not formalized what happens when systems simply endure.

This paper names the gap: structural pressure — the monotone structural cost of inertial continuation under external load. We show that this concept is irreducible to entropy, robustness, resilience, homeostasis, negentropy, or drift. We provide a formal definition with explicit separability conditions and canonical instantiations, prove that sustained pressure implies finite viability horizons even for non-acting systems, establish that pressure is provably undetectable from behavioral traces alone, show that pressure forces regime transitions without failure events, prove that load-neutral passive continuation is architecturally impossible for bounded coupled systems, derive falsifiable predictions that distinguish structural pressure from all neighboring formalizations, and instantiate the complete framework in a concrete engineering domain (lithium-ion battery calendar aging) where decades of empirical data confirm every theorem.

We further show that pressure is structurally prior to drift — it is the field, not the impact — and that no improvement to the agent's policy, control law, or optimization can eliminate the pressure term from the viability equation. The only responses to structural pressure are architectural: increase the budget, reduce the coupling, or accept a finite horizon.

The closest physical analogue — creep in materials science — has been studied for over a century in metals, ceramics, and polymers. Its systematic transfer to adaptive systems has not, to the best of the author's knowledge, been made. The mapping is structurally exact at the level of monotone passive burden accumulation, though the substrate mechanics differ. The Larson-Miller parameterization, if validated for adaptive systems, would make the entire century of creep rupture methodology available for viability estimation.

You can die from standing still. Now there is a theorem for it.

References

Ashby, W.R. (1956). An Introduction to Cybernetics. Chapman & Hall.
Cannon, W.B. (1929). Organization for physiological homeostasis. Physiological Reviews, 9(3), 399–431.
Holling, C.S. (1973). Resilience and stability of ecological systems. Annual Review of Ecology and Systematics, 4(1), 1–23.
Schrödinger, E. (1944). What Is Life? Cambridge University Press.
Shannon, C.E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423.
Norton, F.H. (1929). The Creep of Steel at High Temperatures. McGraw-Hill.
Larson, F.R. & Miller, J. (1952). A time-temperature relationship for rupture and creep stresses. Transactions of the ASME, 74, 765–771.
Barziankou, M. (2025–2026). Navigational Cybernetics 2.5: Canon v2.0. PETRONUS corpus.
Vetter, J. et al. (2005). Ageing mechanisms in lithium-ion batteries. Journal of Power Sources, 147(1-2), 269–281.
Barré, A. et al. (2013). A review on lithium-ion battery ageing mechanisms and estimations. Journal of Power Sources, 241, 680–689.
Birkl, C.R. et al. (2017). Degradation diagnostics for lithium-ion cells. Journal of Power Sources, 341, 373–386.
Keil, P. et al. (2016). Calendar aging of lithium-ion batteries. Journal of the Electrochemical Society, 163(9), A1872–A1880.