The Derivation

Step 1 — Destabilisation is measurable

AP31 establishes that every action is classified as stabilising or destabilising with confidence p computed from accumulated records. The geometric cost C is the coupling capacity reduction the action produces.

For justice: given the total record set of a node’s actions, what is the net destabilisation produced, and what is the projected future destabilisation given the accumulated pattern?

Past is measured. Future is predicted. Prediction tightens as records accumulate (Axiom R). A first offence has a wide prediction range. A tenth offence has a narrow one.

Record bias. Historical record sets are contaminated by systemic prejudice. Centuries of racialised policing, biased sentencing, and discriminatory enforcement are records. They are permanent (Axiom R). But records of a biased system encode the bias.

The consequence geometry computed from biased records will reproduce the bias. The architecture requires auditing the record set for systemic bias before computing the geometry.

Cleaning the input is a structural requirement, not an optional safeguard.

KS-32.1: If destabilisation cannot be measured with convergent confidence from accumulated records. KS-32.7: If the architecture is turned against a group rather than a measured deed — or if the record set it depends on is not audited for systemic bias — the output is corrupted and the architecture is weaponised.

Step 2 — Correction scales with destabilisation

The organism’s response is correction: an action that restabilises. Optimal correction = maximum coherence restored per unit of correction cost. Minor destabilisation → minor correction. Catastrophic destabilisation → catastrophic correction.

This is the distinction between structural justice and retribution. Retribution asks: how much should the offender suffer? Structural justice asks: what correction most efficiently restabilises the organism?

A retributive system may under-correct (the offender suffers but remains destabilising) or over-correct (consuming resources that could stabilise elsewhere).

Step 3 — The correction hierarchy mapped to viability geometry

Each correction level corresponds to a position on the coupled viability corridor (AP01).

Level 1 — Restitution. The node remains within the corridor. Coupling restored directly by the destabiliser becoming the restabiliser. Maximum efficiency — no third-party resources consumed.

This includes restorative justice practices: victim-offender mediation, community reparation, transformative processes. The axioms do not merely permit these. They rank them as the most efficient correction where they work.

Level 2 — Restriction. The node’s corridor is narrowed but the node remains. Fines (reduced resource coupling), community service (directed stabilisation), probation (monitored coupling), restraining orders (specific pathways severed).

Level 3 — Separation. The node temporarily exits the corridor. Incarceration. The organism bears the maintenance cost. The node cannot destabilise because it cannot couple.

Level 4 — Permanent separation. The node is permanently outside the corridor. Indefinite maintenance cost. The window remains in the building but is sealed.

Level 5 — Removal. The node’s corridor is closed permanently. No maintenance cost. The window closes. The I does not die — the I is the building, not the window.

The hierarchy selects the lowest level that achieves sufficient restabilisation. Removal is never selected when separation suffices. Separation is never selected when restriction suffices. Minimum intervention for maximum stabilisation.

You have just watched five levels lay themselves out on one axis. The axis is efficiency. Not desert. Not emotion. Not revenge. Efficiency — measured in coherence restored per unit of cost.

KS-32.2: If a lower level can produce equal stabilisation to a higher level for the same case, and the architecture selects the higher level, the ranking is wrong.

Step 4 — The removal threshold

Level 5 requires two conditions simultaneously.

Condition A — The fixed floor. The destabilisation D must exceed a minimum absolute threshold F. Let C_sep = maintenance cost of permanent separation per unit time. Let B_total = organism’s total coupling budget.

The floor F is the destabilisation at which maintaining the node in permanent separation costs more than removal: F ≥ C_sep / B_total.

Below this floor, Level 4 is always more efficient than Level 5 regardless of organism state.

The floor must also incorporate the cost of wrongful removal. Removal is irreversible (Axiom R).

If subsequent records reveal the measurement was wrong — misattributed destabilisation, faulty prediction, corrupted records — the window is permanently closed based on error.

This permanent loss of the wrongly-removed node’s future stabilising potential must be factored into F. The irreversibility cost raises the floor, which makes removal rarer. This is structurally correct.

Condition B — Organism-relative scaling. The effective threshold adjusts with the organism’s current robustness R. A robust organism (high coherence, wide corridor, large reserves) has a higher effective threshold.

A fragile organism (low coherence, narrow corridor) has a lower one. Candidate functional form: Threshold_eff = F × (R/R₀)^α, where R₀ is baseline robustness and α is a scaling exponent.

The exact α is Debt 25. Linear (α=1): threshold scales directly. Sublinear (α<1): threshold increases more slowly with robustness.

The confidence requirement. Because removal is irreversible and wrongful removal is permanent, the confidence threshold for Level 5 must be extraordinarily high — structurally higher than for any other level.

If the required confidence cannot be achieved given measurement uncertainty and adversarial record-injection (AP31 KS-31.9), Level 5 is structurally unreachable.

KS-32.3: If permanent separation is always more stabilising than removal regardless of cost, Level 5 is never justified.

KS-32.8: If the confidence required for Level 5 cannot be achieved given measurement uncertainty and adversarial record-injection, Level 5 is structurally unreachable and should be removed from the hierarchy.

Step 5 — The structural genocide brake

This is the step that matters most. Read it slowly.

A fragile organism that begins removing nodes becomes more fragile. Each removal costs coupling capacity (a window lost) and record-generating diversity (one fewer perspective). This loss is itself a destabilisation that enters the measurement.

Each removal raises fragility, which lowers the threshold, which could enable more removals. This is the structural pathway to genocide.

The geometry prevents it. Each removal’s cost (lost coupling capacity + lost diversity) is a destabilisation that raises the organism’s fragility, which changes the geometry.

Beyond some number of removals, the cumulative cost of further removal exceeds the cost of the destabilisation it addresses. The architecture produces a natural saturation point: the removal curve is self-limiting.

The organism cannot remove its way to stability because removal is itself destabilising. The saturation point is the structural genocide brake. It exists in the geometry.

A civilisation that reaches the saturation point and continues removing is no longer following the derivation. It is overriding the geometry with ideology. KS-32.7 (dehumanisation) fires. The architecture has been weaponised.

You have just watched the geometry build its own brake. No external authority imposes the limit. The coupling cost imposes the limit.

The same way ε is the only stable break — not by decree but by geometry — the genocide brake is the only stable correction boundary.

The architecture cannot be driven past the saturation point without self-destructing. That is the safety.

KS-32.5: If the genocide brake does not engage — if the removal cost does not produce a saturation point — the self-limiting property fails and the hierarchy is structurally dangerous.