No Continuation

Before the first live contingency arrived, D-01 had two bounded control axes and four individual live probes arranged as two matched pairs.

The delayed motor feed still read HOLD. The rotor held base speed. The heat front had cleared the final sensor from the sealed diagnostic sweep, and the stator had returned to its neutral phase. Nothing forced an immediate choice.

Evan kept D-01 in HOLD and opened the audit protocol he had committed before the next external event.

He called it LCA-01, a local continuation audit. It did not estimate how many universes remained. He no longer had an instrument that could justify that sentence. It asked a narrower question: within every software-controlled direction D-01 authorized, how many independent safe perturbations left a distinguishable difference in the following consequential cycle?

The distinction had become necessary because the five-channel rank still looked healthy.

The cracked Hall probes disagreed with one another. Rotor angle did not reduce to their common motion. Heat-front position kept its own residual. The battery banks retained separate noise. Evan whitened the five residual channels against calibrated replay and computed the exponential entropy of their covariance eigenvalues. That effective rank could remain near five while safe changes lost the ability to matter in the next cycle.

Sensor effective rank asked how many independent ways the measured residuals varied inside a window. Authorized retained-response rank asked how many independent post-cycle differences remained after a bounded change had passed through the live feedback path.

Evan had built the audit around the complete software-controlled subspace D-01 allowed. The first axis shifted stator phase relative to the traveling heat front, once positive and once negative. The second shifted brake-command timing relative to that front, again in a matched positive and negative pair. The direct contactor remained a separate hard safety action. Inert register writes matched command encoding, timing, and controller load without actuating the plant. The independent trips limited every live probe.

For each pair, the audit measured ordinary current, motion, and heat, then waited one consequential cycle and recorded the retained state. It standardized each retained-response channel by its replay standard deviation, making the two-column response matrix dimensionless. The columns represented differences that still mattered after the transient. It counted singular values only above a dimensionless threshold fixed against calibrated replay.

The audit froze five evidence anchors.

P-04 preserved Evan's false report and Lena's manual stop. The DECOM-01 closeout and DR-01 preserved the reconstructable escrow and the later destruction of OLD-7. C-01 preserved Lena's disclosure, the continuation gate, the independent hard stop, and her alternative explanation. NLE-01 preserved her signed DO NOT START refusal. D-01 preserved NEW-1 as an ordinary bearing inside a live topology neither its material nor its manufacturer had supplied.

The bytes had not changed. Their signatures still verified. The audit did not treat a document as a physical law. It treated each committed event as a boundary condition. Any successor that required one of those events not to have happened was rejected.

Lena remained present only in those records. Her warning sat above the analysis in the wording she had insisted on before D-01 started.

MEASURED SENSOR RANK IS NOT REMAINING CAUSAL RESERVE.

Evan ran the first window.

All four live probes were locally lawful. Phase and brake timing produced two independent retained differences. Both singular values cleared the audit threshold.

AUTHORIZED RETAINED-RESPONSE RANK: 2.

Relief arrived before interpretation. Two was not many. It was more than one.

He allowed the precommitted audit probe to advance. The stator shifted within its narrow envelope. The rotor responded by the modeled amount. Brake temperature followed late. The optical gate counted another mechanical second without correcting the controller clock.

The physical responses remained distinct. Their retained differences were already closer together.

In the second window, positive and negative phase remained different commands. Brake-command timing still advanced or retarded the heat response. All four individual probes again passed their local limits.

Both control axes produced visible transients.

After the following consequential cycle, their retained differences were operationally indistinguishable from collinear. Changing stator phase and changing brake-command timing had become two ways to arrive at the same one-dimensional displacement at the audit threshold.

The smaller singular value fell below the preregistered repeatability threshold. Sensor noise could not promote it back into a direction.

One retained direction remained.

AUTHORIZED RETAINED-RESPONSE RANK: 1.

The audit entered its precommitted review boundary before the third window.

The boundary had a fixed duration and a matched thermal restart envelope. The isolated sequencer would resume automatically. No report, interpretation, file copy, or decision to continue could change its timing or select another carrier condition.

He copied the current state to two removable records and compared their hashes on separate readers. They matched. He printed the summary because paper could fail in ways he understood. The printer drew each line once.

He put his hand on the wall contactor.

Opening it would abort the only remaining precommitted live window. The result would then mean that the protocol had stopped at one, not that one had persisted. Every safety rule said that distinction was enough.

The audit display still showed authorized retained-response rank one.

He released the handle without opening it. One more bounded probe window could still tell him whether that direction persisted. The probes were live and consequential. Measuring the remaining freedom could consume it.

Evan read the audit code again.

He checked the singular-value threshold against synthetic cases. He removed the evidence anchors one at a time. Without P-04, the solver invented a continuation in which his report had triggered the stop. Without the DECOM-01/DR-01 bundle, it invented OLD-7 intact in the chamber. Without C-01, it removed the disclosure, continuation gate, and hard-stop history. Without NLE-01, it treated D-01 as authorized. Without D-01's current commit, it solved a machine that had never started.

Those were not continuations. They were substitutions for the state he had reached.

He restored every anchor.

The five-channel covariance effective rank was 4.49.

The regulator clock ticked. The Hall probes disagreed. The rotor turned. The brake warmed. The remote motor record remained signed. Beyond the utility door, Ruth crossed the kitchen and opened a cupboard, then a drawer. The two sounds were different and stayed different.

Unlike terms did not cancel.

They also did not guarantee a common future.

The fixed HOLD expired. The isolated sequencer enabled the final precommitted window with the contactor closed and every hard trip armed.

All four individual probes in the two matched pairs produced their expected local transients. Stator current changed. The rotor answered. Brake heat arrived by the calibrated lag. The clocks advanced independently. The matched register writes produced no physical response.

One cycle later, every retained state lay inside the repeatability band of every other one.

The response matrix had singular values of 0.03 and 0.01 against a threshold of 0.08. Both control axes still moved the plant. Neither left a distinguishable direction that mattered next cycle.

The solver was not reporting dead sensors. It was not reporting frozen time. It was not claiming that the universe had a measurable branch counter. Rank zero in the complete authorized control subspace could still leave one isolated successor outside any steering choice. LCA-01 had found no authorized retained difference, not an empty set.

Evan opened the wall contactor.

The blades separated. Stator current fell. The eddy brake accepted the rotor's stored energy below its temperature limit. The touchdown bearing remained clear while speed declined. Every safety action worked.

The singleton shutdown trajectory remained locally valid. The rotor crossed its next speed mark when the brake model predicted. The regulator clock completed another second. Beyond the utility door, Ruth closed the cupboard.

The delayed motor feed changed from HOLD to APPLY. The battery controller authenticated and latched the signed event without consuming it. LCA-01's precommitted gate had priority, and the event carried no expiry deadline. D-01 had already committed to giving it a nontrivial APPLY or HOLD consequence. The contactor could stop actuation. It could not make that retained requirement unhappen.

The feasibility layer tried D-01's two bounded carrier candidates for the pending step. APPLY required software to restore actuation after the direct contactor had physically opened. HOLD preserved the safe shutdown but made the signed contingency causally inert, violating D-01's committed live-edge requirement.

The local shutdown and controller predictions remained lawful. Neither candidate satisfied the complete committed transition. The result was a model finding made before the controller attempted it, not a measurement taken after an empty boundary.

The probe result had committed before the event arrived. The feasibility addendum committed after authentication and before any attempted transition, while the current state and its one shutdown trajectory still existed.

LCA-01 / LOCAL CONTINUATION AUDIT

scope: D-01 authorized control subspace
claim type: OPERATIONAL / LOCAL

frozen evidence anchors:
  P-04 termination record                  VERIFIED
  DECOM-01 closeout + DR-01 destruction    VERIFIED
  C-01 packet, gate, hard stop, explanation VERIFIED
  NLE-01 signed refusal                    VERIFIED
  D-01 current-state commit                VERIFIED

authorized control-space dimension:        2 / COMPLETE FOR D-01
matched live-probe pairs per window:        2
individual live probes per window:          4
response channels:                         STANDARDIZED TO REPLAY STD DEV
retained-response threshold:               0.08 / DIMENSIONLESS
sensor effective-rank convention:          EXP ENTROPY / WHITENED COVARIANCE
independent safety trips:                   ARMED
local subsystem models:                     PASS

window   sensor effective rank   response singular values   response rank
W-2      4.47 / 5                0.31 / 0.12                2
W-1      4.52 / 5                0.19 / 0.04                1
W-0      4.49 / 5                0.03 / 0.01                0

instantaneous actuator response:            PRESENT
nominal singleton shutdown successor:       PRESENT
pending carrier candidates:                 2
complete feasible candidates:               0 / MODEL
audit protocol committed:                   BEFORE EVENT ARRIVAL
probe result committed:                     BEFORE EVENT ARRIVAL
feasibility addendum committed:             AFTER AUTH / BEFORE TRANSITION
empty successor set observed:                NO
global branch count inferred:                NO
factorization observed:                      NO

result: LOCAL RANK ZERO / NO CONTINGENT SUCCESSOR FOUND

The current state remained physical. The audit found no address for the transition D-01 had promised.