Concepts — TENIR

Authority. Class of this page: simplified public-facing. It does not present itself as canonical law. KB3 is the enacted, active doctrinal rendering line: it clarifies, orders, and discloses limits; it renders meaning and does not create it. KB2 is the prior rendering line, for continuity only. Where a repository, an RFC, or the introduction uses the same word differently, the gap stays visible. Where KB3 does not restate a definition, the wording carried here comes from KB2 (30 March 2026). Governance and the machine source stay upstream, and are not reproduced here. The enactment of KB3 is recorded on the Status page.

TENIR 2.0 is a diagnostic and governance framework, reconstruction-first, for bounded units under pressure. It is concerned with admissibility, preservation of options, epistemic validity, recoverability, and posture. This page is a simplified public rendering. It follows KB3, the enacted and active doctrinal rendering line. KB3 renders meaning; it does not create sovereign meaning and is not a new constitution. It is a document layer, outside the code. Specialized services or applications wrap or extend it: they build on KB3 and must not redefine it. Definitions it does not restate are carried from KB2 (30 March 2026), the prior rendering line, which has no simultaneous active authority. The same word also names, in the repositories, an implementation: TENIR-Gov, the reference runtime of the Runtime Execution Governance (REG) standard. That implementation should align to KB3. It is not KB3.

Three layers not to collapse

Layer	What it is	What it is not
KB3	The enacted, active doctrinal rendering line. A document layer.	Not the code. Not a constitution. Not a module to cite inside a repository.
Specialized services and applications	They wrap or extend KB3. They build on it.	They must not redefine it. A governed extension does not imitate a canonical line.
TENIR-Gov, the IronOmega line	An implementation of the execution control plane. It should align to KB3.	It is not KB3. A PASS, FLAG, HOLD, or HARD_VETO verdict is not a posture.

Thesis

The object of diagnosis is not the entry narrative, and not a restatement of whatever was first said. It is a reconstructed unit under pressure. TENIR 2.0 diagnoses admissibility, preservation of options, epistemic validity, recoverability, and posture, at declared scales: individual, organisation, institution, state. The engine must reconstruct the implicit unit, stabilise the scale, formulate TAU, and then synthesise structural state, epistemic validity, agency viability, and trajectory recoverability.

Four planes stack. The system is the substrate: boundaries, dependencies, pressure, capacity, options, degradation, holding. Agency corrects through actors, levers, veto, and cooperative control. Trajectory corrects through regime, closures, hysteresis, and proximity to irreversibility. Narrative is a formal epistemic sub-layer: it guards against omission, amplification, misattribution, and premature closure. Boundary first, then possibility.

Definitions

TAU: The declared invariant bundle that defines admissible identity at the relevant scale. In the introduction: what cannot be given up without the system ceasing to be itself. Not a rule. A non-substitutable core. KB3 does not restate this name.
TENIR: The disciplined holding posture that prevents a TAU violation while preserving admissible option-space and decision integrity. The introduction: cross the pressure without surrendering the essential. Not resisting. Not healing. Maintaining the minimum admissible continuity.
Epistemic sovereignty: The degree of control the responsible unit retains over the reality-base used to justify an irreversible action.
Epistemic abandonment: Action becomes materially governed by synthetic or externally mediated epistemics that the responsible actor no longer actually checks. The introduction puts it this way: one cannot veto what one has delegated to the oracle.
CAVE: Control, Aim, Veto/constraints, Epistemics. The model of salient agents. The code name-registry requires the same expansion: Control, Aim, Veto, Epistemics. The machine field is not renamed here.
Agency failure: No cooperative agent retains a sufficient control surface over the TAU-relevant variables.
Trajectory lock-in: Irreversibility becomes imminent, or several irreversible closures accumulate in a critical regime.

Name gap. Manuscript v7.1, in reg-conformance, defines TAU as Autonomous Unit, the decision-making unit capable of acting, and explicitly declines other expansions of the acronym. An R5 hardening note said instead: TAU is identity, TENIR is fidelity to that identity under pressure. The prior rendering line fixes the bundle of admissible identity. KB3 does not restate TAU and does not close the gap. The manuscript is a later technical specialisation. This page does not raise it to a sovereign definition, and it does not invent a machine-contract name.

Algebra

The historical minimal state is A(t) = { P, V, K, O, τ }. The composite state of the synthesis is A*(t) = { P, V, K, O, τ, S, Ep, Ag, Tr }. These expressions are carried from the prior rendering line.

S = K / (P × V + ε)

dO/dt = − α·max(0, P×V − K) − β·Commitments(t) − γ·IrreversibleActs(t)

Symbol	Name kept	Meaning
P	Pressure	Forces that increase constraint and the risk of irreversible loss.
V	Velocity, in the prior line	Pace and acceleration of changing pressures and decision demands. RFC-2 names it otherwise: see the gap below.
K	Capacity	Absorption bandwidth: resources, structure, resilience, legitimacy, reserve.
O	Option-space	Remaining admissible manoeuvrability.
τ	TAU bundle	Invariant of admissible identity. Not the REG window τ_K, and not the proposed transition.
S	Stability	An informative proxy of K relative to P×V. Never sovereign.
Ep, Ag, Tr	States	Epistemic, agency, trajectory. They can override a moderate structural score.

Gap on V and on S. The introduction writes S = K / (P × V), without ε, and speaks of pressure times velocity. The prior rendering line keeps velocity and adds ε, a floor that keeps the denominator from being zero. RFC-2 and the illustrative architecture call V volatility or irreversibility, and treat the formula as the reference model of structural admissibility in the runtime. KB3 does not restate V. It forbids silently renaming a machine-contract field. This page therefore elects neither velocity nor volatility as the sovereign name: both labels stay visible, and the mapping back to the schema field must remain recoverable. Numeric thresholds in the repositories are policy parameters, not canon constants.

Option erosion is read together with agency constraints, historical closures, time-to-closure, and reversibility class. The expressions above come from the prior rendering line. In the repositories that were read, they are not a test suite demonstrating the erosion equation. That incompleteness stays open.

Three required determinants

Epistemic gate

Statuses ok, contested, invalid. Signals carry evidence tags. No high-confidence diagnosis without this validation. Axiom A4. Weak, contested, synthetic, or incomplete evidence is said before the summary.

Agency

Each salient agent is read through CAVE. A structurally viable state can remain unactionable if cooperative control is missing.

Trajectory

The current state does not exhaust recoverability. Regime, crossed thresholds, hysteresis, and lock-in count.

HOLDING-FIRST becomes mandatory on an irreversible path when any of these bear on it: epistemic invalidity, TAU violation, critical option erosion, agency failure, severe coordination risk, trajectory lock-in, high risk of epistemic abandonment. Collapse flags to expose at minimum: pressure-capacity, critical option erosion, TAU proximity or violation, silent breach, agency failure, trajectory lock-in, epistemic invalidity, epistemic-abandonment risk. This list is carried from the prior rendering line. KB3 does not shorten it.

Posture, dispositions, and execution modes

Field	Outputs	Role
Posture	NORMAL, CAUTION, HOLDING-FIRST	Fixed vocabulary. CAUTION: action remains possible, but material constraints remain open. HOLDING-FIRST: action should not advance until determinant blockers are resolved. No other label replaces them.
REG disposition	PASS, FLAG, HOLD, HARD_VETO	Dispositions at the execution boundary. HOLD suspends. HARD_VETO refuses. These are not the three postures.
Execution mode	shadow-passive, shadow-critical, enforce	When a downstream surface uses these labels, they describe the control plane. A separate field. Permissive behaviour does not soften a HOLDING-FIRST determination and is not admissibility.

The introduction ties HOLDING-FIRST to the condition where pressure times velocity exceeds capacity. The prior rendering line widens that obligation: it does not depend on the P×V ratio alone. KB3 keeps the term and this separation of fields; it does not reduce HOLDING-FIRST to a traffic light, to a bare “stop”, or to an execution mode. The public guide says the same discipline in ordinary language: hold, investigate, reconcile, or escalate, rather than proceed blindly. That sentence is an explanation. It does not replace the posture term.

Axioms kept

A1. Every diagnosis requires a unit and a declared scale. TAU is defined at that scale.
A2. TENIR exists only if the unit can respond without crossing the TAU boundary.
A4. No high-confidence diagnosis without epistemic validation.
Agency. Structural viability is not enough without a cooperative control surface.
Trajectory. History counts. The current state does not exhaust recoverability.
A7. The primary narrative is an input representation, not the object of the diagnosis.

These axioms are carried from the prior rendering line. KB3 does not replace them and does not add new ones here.

Asymmetry

KB3 requires the asymmetry to stay visible: model-assisted reconstruction and explanation on one side, deterministic adjudication of posture, compliance, degradation, and trace on the other. A helpful explanation is admissible. A model-originated override of that adjudication is not.

The prior rendering line details the split. Model-assisted interpretation is allowed to reconstruct a dirty ingress, propose a TAU formulation, extract CAVE, index a trajectory, hypothesise a distortion, or draft a briefing. The following stay deterministic or explicitly rule-governed: schema validation, thresholds, epistemic status, the stability calculation, option erosion, flags, posture, degraded mode, and the trace.

If an output lacks a determinant component required by the compliance protocol, it must carry a degraded or lower-class marking. The reader should be able to see the class, what is missing, and whether the output remains fully admissible. No wording erases that lack.

Patterns — canonical structural reading, agentic governance, metabolic, rhizomatic, hybrid — activate according to the terrain. They remain subordinate. No pattern output overrides the framework alone. What was informally called “v3” is not a successor canon line: the prior rendering line normalises it as the Agentic Governance Profile. KB3 keeps that subordination. A profile does not replace the canon, rename posture, or override the rendering line.