AQ-DVC-001 — Demonstration version control
A demonstration is usable as evidence only if a reader years later can establish exactly what ran, on what build, measured by what instruments, under what claim, and witnessed by whom. This document defines the mandatory record for every Akasha-Q run — evidential or not — and the identifier scheme that binds a run to its artefacts. It defines a record format to which future runs will be held; no run has yet been performed and no element has reached laboratory validation. The format applies from DM-0 onward and to every architecture tier (CA, Y1, FR, LH).
THE NINE MANDATORY FIELDS
| Field | Content | The failure it prevents |
|---|---|---|
| F-1 Run identity | AQ-DEMO-ID, allocated before the run is armed and bound at allocation to a hash of the claim text | Post-hoc selection of which attempt counts. Without a pre-allocated identity the standing prohibition in AQ-DEMO-FAIL-001 cannot be enforced |
| F-2 Element and scope | Element in canonical AQ-NODE-n form (AKQ-NODE-n in Parts up to 12 denotes the same articles), the layer L0–L9 exercised, the trust boundary TB-1–TB-5 in scope, and the architecture tier | A result obtained on one article being read as a result for another, and a CA-tier sketch being cited as a Y1 prototype result |
| F-3 Claim under test | One falsifiable sentence, plus the evidence-ladder rung the run would support if it passes and the rung the element falls to if it fails | The claim being rewritten after the result is known, so that any outcome reads as a success |
| F-4 Configuration identity | Software commit, FPGA bitstream identifier, firmware version, parameter set and calibration set — each as a hash or immutable identifier, never a friendly name | “It worked last week” with no way to establish what “it” was, and silent drift between the demonstrated build and the described build |
| F-5 Instrumentation and timebase | Instrument serials, last calibration reference, timebase source and its distribution path, sample rates and record lengths | Unattributable numbers. Without this field no reported value can be tagged MEASURED under DC-1; it degrades to DERIVED at best |
| F-6 Environment and boundaries | As-recorded, never as-intended: physical setting, link medium and length, ambient conditions actually observed, and which functions were emulated versus physically present | A bench-only assumption travelling silently up the six-rung reach ladder into a claim about a longer link |
| F-7 People and roles | Demonstration Lead, Safety Authority, Evidence Custodian, Claim Owner, and whether an Independent Observer was present | An unwitnessed run being cited at DM-3 or above, where independent observation is a condition of the maturity level |
| F-8 Provenance pointer | Evidence Fabric record identifier for the raw capture, with every reported value individually tagged MEASURED, DERIVED or MODELLED | A conclusion that cannot be traced back to a measurement. This field is what makes DC-5 achievable: the run reconstructable from the Evidence Fabric alone |
| F-9 Disposition | EVIDENTIAL or NON-EVIDENTIAL; outcome pass / fail / correct-abstention / VOID; attempt index; and for a fail, the AQ-DEMO-FAIL-001 register entry reference | A rehearsal being cited as a validation, and a failed attempt disappearing from the sequence |
Correct-abstention. Where the element returned abstain, safe-hold or deny and that was the right outcome under the trust state in force, F-9 records correct-abstention. Under DC-3 abstention is a permitted output; it is recorded as a pass against a claim about refusal, or as an expectation failure of the script (class FC-C), and never as an element fault.
If any of F-1 to F-9 cannot be completed, the missing field is written as Undetermined with the reason, and the run's disposition defaults to NON-EVIDENTIAL. A missing field never defaults to benign and is never inferred from a neighbouring run. Completeness is a precondition of evidence, not a formatting preference.
AQ-DEMO-ID ALLOCATION
Figure 1 — Identifier anatomy
AQ-DEMO-L2-NODE1-20260830-007-a02 | | | | | | | | | | | +-- attempt index within this allocation (a01..) | | | | +------ allocation sequence within the date (001..) | | | +--------------- allocation date, UTC, YYYYMMDD | | +--------------------- element under demonstration (AQ-NODE-1) | +------------------------ layer in scope (L0..L9) +-------------------------------- fixed prefix
- An ID is allocated before the run is armed. An ID may never be created after data exists.
- IDs are monotonic within (layer, element, date). They are never reused and never deleted.
- A change to the claim text (F-3) or to configuration identity (F-4) closes the allocation and requires a new sequence number. Only a repeated attempt under an identical claim and identical configuration increments the attempt index.
- An allocation that is not run closes as NOT RUN with a reason. Under DC-2 an unexplained gap in the sequence is itself a finding, investigated as such.
- The ID appears inside the artefacts, not merely beside them: burned into recorded video frames, embedded in raw capture filenames, printed on every slide, figure and report line that cites the run.
- The register is append-only. A correction is a new entry referencing the original; an entry is never edited in place.
Figure 2 — Register lifecycle and the one-directional marking
ALLOCATED --> ARMED --> RUN --> SEALED
| | | |
| | | +--> EVIDENTIAL citable at its rung
| | | +--> NON-EVIDENTIAL never citable as evidence
| | +-----------> VOID instrumentation invalid;
| | retained, never deleted
| +--------------------> NOT RUN closed with reason (DC-2)
+--------------------------------> SUPERSEDED claim or config changed;
new ID, cross-referenced
MARKING DIRECTION:
EVIDENTIAL --> VOID permitted
EVIDENTIAL --> NON-EVIDENTIAL PROHIBITED (would launder a bad result)
NON-EVIDENTIAL --> EVIDENTIAL PROHIBITED (no upgrade path exists)
NON-EVIDENTIAL RUN MARKING
A run whose purpose is communication rather than measurement is marked NON-EVIDENTIAL at allocation. It may use replayed capture, simulated inputs, scripted timing or a partial article, provided each such element is declared on the material itself. It carries a visible NON-EVIDENTIAL RUN marker on the artefact — on the video frame, on the slide, in the figure caption — not only in metadata, because metadata does not travel with a screenshot. It may not be cited in any evidence-level determination on the Part 12 ladder, in any funding, partnership or recruitment document as a technical result, or in any external statement about capability. Every number shown in such a run is tagged MODELLED unless F-5 was completed.
REHEARSAL VERSUS VALIDATION — AQ-RL-26.14
A rehearsal exercises the presentation of an element. A validation tests a claim about the element. The two shall never be the same run. A rehearsal can never raise an evidence level; a rehearsal that contradicts a standing claim still lowers one.
| Dimension | Rehearsal | Validation |
|---|---|---|
| 1 Purpose | Exercise delivery, timing, narration and recovery of a presentation | Test one falsifiable claim about the element |
| 2 Claim binding (F-3) | None. The field names the claim being illustrated and is marked ILLUSTRATIVE | One claim, fixed before allocation and hash-bound to the ID |
| 3 Data sources | Replay, simulation, partial article and scripted input permitted, each declared on the material | Live path only. No replayed or scripted value may enter a reported number |
| 4 Instrumentation (F-5) | Not required. Where absent, no number from the run may be quoted anywhere, in any context | Required and calibration-referenced. Absent or invalid instrumentation voids the run |
| 5 Marking (F-9) | NON-EVIDENTIAL, always, allocated as such | EVIDENTIAL at allocation. May be downgraded to VOID; may never be re-marked NON-EVIDENTIAL to dispose of a bad result |
| 6 Failure handling | AQ-DEMO-FAIL-001 steps 1 to 3 apply in full; steps 4 to 7 apply where the observation contradicts a standing claim | AQ-DEMO-FAIL-001 applies in full, every step, every time |
| 7 Effect on the ladder | Can lower a rung. Can never raise one | Can raise a rung on a pass with a complete record; lowers a rung on a fail |
Row 7 is the asymmetry that makes the rule enforceable rather than decorative, and it follows directly from DC-4: degradation is monotonic without validation. Evidence is gained only under evidential conditions with a complete F-1 to F-9 record; it can be lost anywhere, including in a rehearsal, an internal walkthrough or a conversation held over a bench. Nothing reaches the Part 12 evidence ladder except through a sealed evidential run record, and every rung asserted anywhere in the charter carries the AQ-DEMO-ID that supports it.