← Part 26

AQ-DEMO-FAIL-001

The binding seven-step sequence for a demonstration that does not produce its asserted result: stop or contain, preserve evidence, declare, investigate, reproduce, correct or revise the claim, decide next action - with named role ownership, four situation handlings, and the rule that an unreproducible failure is the more serious finding.

Charter artifact · gated by Part 26 · revision 0, draft for internal review

AQ-DEMO-FAIL-001 — Failed demonstration protocol

This document defines the mandatory sequence to be followed when an Akasha-Q demonstration does not produce the result its script asserted. It describes a process to be followed, not a run that has occurred. No Akasha-Q element has reached laboratory validation, so every demonstration referenced here is a demonstration yet to be performed. The protocol is binding at every demonstration maturity level from DM-0 internal walkthrough through DM-6, across every architecture tier (CA, Y1, FR, LH), and at every rung of the six-rung reach ladder. It applies identically whether the run was watched by one engineer or by an external party.

STANDING PROHIBITION

A failed attempt shall never be quietly re-run so that only the successful attempt is shown. Every attempt consumes an AQ-DEMO-ID attempt index at the moment it is armed (see AQ-DVC-001), and the run register is append-only. Where a demonstration is presented after re-running, the full attempt sequence — each failed attempt and its disposition — is presented with it. Showing attempt a04 without disclosing attempts a01 to a03 is a reportable integrity breach, not a presentation choice. This prohibition binds internal reviews, recruitment material, partnership discussions and funding conversations equally.

ROLES

  • Demonstration Lead — owns the run, confirms a stop call, and is the single voice that declares an observed failure to anyone present.
  • Safety Authority (L9 Systems Engineering & Safety Governance duty holder) — owns containment of optical emission, RF, stored energy and moving mass. Overrides the Demonstration Lead on any question of physical containment, without discussion.
  • Evidence Custodian (L6 Evidence & Provenance Fabric duty holder) — freezes capture, seals raw artefacts, and is the only role permitted to mark a record VOID or to reopen a sealed record.
  • Investigation Owner — appointed at step 4 by the Safety Authority. May not be the sole author of the element under investigation.
  • Claim Owner — holds the written claim the run was testing and signs either the correction or the revision at step 6.
  • Independent Observer — required at DM-3 and above. Records observations separately from the run team and countersigns the step 3 declaration.

Figure 1 — The seven-step sequence and its two terminating branches

   RUN IN PROGRESS
         |
         v
  [1] STOP OR CONTAIN ....... any participant may call stop;
         |                    Safety Authority owns physical containment
         v
  [2] PRESERVE EVIDENCE ..... freeze capture, seal raw record,
         |                    NO re-arm, NO recalibration, NO power cycle
         v
  [3] DECLARE OBSERVED ...... spoken to all present before they leave;
      FAILURE                 written in the same session
         |
         v
  [4] INVESTIGATE ........... Investigation Owner appointed by L9
         |
         v
  [5] REPRODUCE --- reproduced? --- NO -->  UNREPRODUCIBLE FINDING
         |                                  (escalates; see S-4)
        YES
         |
         v
  [6] CORRECT THE ELEMENT  OR  REVISE THE CLAIM   (exactly one, signed)
         |
         v
  [7] DECIDE NEXT ACTION
      re-run | re-scope | suspend element | lower evidence level
         |
         v
   RUN REGISTER UPDATED (append-only; failed attempt retained forever)

STEP OWNERSHIP AND GATES

StepOwnerMandatory outputGate that must close before the next step begins
1 Stop or containAny participant calls; Demonstration Lead confirms; Safety Authority for any physical hazardRun halted in a state that is contained and otherwise unchangedSafety Authority states containment achieved, aloud
2 Preserve evidenceEvidence CustodianRaw capture frozen and sealed under the run's AQ-DEMO-ID; configuration identity, instrument state and timebase reference recorded as-foundNo reconfiguration, recalibration, restart or power cycle until the seal is confirmed
3 Declare observed failureDemonstration Lead, countersigned by the Independent Observer where presentA written statement in observation language only, with cause explicitly separated and left blankDeclaration recorded before anyone present leaves the room
4 InvestigateInvestigation OwnerA candidate cause set, each entry tagged MEASURED / DERIVED / MODELLED per DC-1The set contains at least one candidate that explains the observation with no change to the element (harness, script, or expectation)
5 ReproduceInvestigation OwnerA reproduction attempt on the sealed configuration and a stated reproduction rate over a stated number of attemptsThe first reproduction attempt runs under the sealed configuration, before any correction is applied
6 Correct or revise claimClaim OwnerEither an element change carrying its own new demonstration, or a rewritten claim at a lower evidence levelExactly one of the two, signed. “No action required” is not an available outcome
7 Decide next actionSafety Authority, with Demonstration Lead and Claim OwnerRegister entry: re-run, re-scope, suspend element, or lower evidence level — with reasonEntry appended to the run register and cross-referenced from the failed AQ-DEMO-ID

CLASSIFICATION AT STEP 3

The declaration names what was observed and assigns one class. The class may be revised at step 6; it may not be left unassigned.

  • FC-A capability — the element did not do what the claim asserted.
  • FC-B evidence — the element may have done it, but the record cannot show that it did.
  • FC-C expectation — the element returned abstain, safe-hold or deny and the demonstration script assumed execute. Under DC-3 abstention is a permitted output; this is a failure of the script, recorded as such, and never rewritten as an element fault.
  • FC-D instrumentation — the harness, capture chain or timebase failed. The run is voided, not passed.
  • FC-E containment — a safety or containment event. L9 leads from step 1 and the element is suspended at step 7 by default.

THE FOUR SITUATION HANDLINGS

S-1 — INTERNAL RUN, NO EXTERNAL WITNESS

The full sequence applies. Nothing is skipped because nobody outside the team saw it. The absence of an Independent Observer removes the countersignature, not the record: step 3 is still written in the same session and step 7 is still appended. A team that runs the protocol only when watched does not have the protocol.

S-2 — FAILURE IN FRONT OF PEOPLE OUTSIDE THE RUN TEAM

The Demonstration Lead declares the observed failure in the room, in plain observation language, before the audience disperses, and offers no cause on the spot. The element is not re-run for that audience in that session. Those present are given the AQ-DEMO-ID, told that an investigation is opening under this protocol, and told what will be issued when it closes. Characterising a failure to a witness as a glitch, a gremlin or “just the network” is prohibited: those are cause claims stated before step 4 has begun.

S-3 — FAILURE FOUND AFTER A RUN WAS RECORDED AS A PASS

The Evidence Custodian reopens the sealed record and changes its disposition with a written reason; the original entry is retained unedited. Every downstream citation is traced and corrected, including material already sent outside the programme, and a correction notice is issued to every party that received the original claim. Under DC-4 the element's evidence level falls immediately on this single observation and is not held pending investigation. Correction is issued whether or not anyone outside the programme would otherwise have noticed.

S-4 — FAILURE THAT CANNOT BE REPRODUCED

An unreproducible failure is a more serious finding than a reproducible one, and is handled as an escalation rather than a closure. A reproducible failure is a bounded defect with a known trigger. A failure that will not reproduce means the element has a state the programme cannot enumerate: its behaviour is not a function of the quantities currently being recorded. Consequently: (i) the element's evidence level is lowered, never held; (ii) the element may not advance a rung on the reach ladder while an unreproduced observation is open; (iii) the unrecorded state is named as an instrumentation gap and closed by adding a recorded quantity, not by waiting for the observation to recur; (iv) under DC-2 the outcome is written as an explicit undetermined — never as “no fault found”, never as a pass, never as an absence.

EFFECT ON THE EVIDENCE LADDER

A failed demonstration resets the element's asserted rung to the highest level still supported by evidence the failure does not touch. Where that cannot be determined at step 3, the rung is set to Undetermined until step 6 closes; it is not held at its prior value while the investigation runs. An element asserted at Lab validation whose validating run fails returns at best to Theory, and to Concept where the failure touches the operating principle itself. Restoration is by fresh demonstration under AQ-DVC-001 evidential conditions with a new AQ-DEMO-ID, mirroring DC-4: the level falls on one observation and rises only through the complete recovery sequence.

Under DC-5 every step above must be reconstructable from the Evidence Fabric alone, with no reliance on anyone's recollection of the room. If the sequence cannot be reconstructed from the sealed record, the protocol was not followed, regardless of what was done at the time.