AQ-EVIDENCE-WALL — Claim register: format, rules and operation
STANDING AND SCOPE
The Evidence Wall is the register of every statement Akasha-Q is permitted to make about itself to a reader outside the immediate working group. It is not a summary of the programme and not a presentation asset. It is a controlled list of sentences, each bound to an evidence level and to a navigable path into the L6 Evidence & Provenance Fabric. Nothing appears on the wall except through AQ-PUBLIC-CLAIM-GATE; nothing stays on the wall that cannot be walked to its primary record by a person who did not write it.
The wall exists because the programme's honest position is asymmetric: the intended capability is large and the accumulated evidence is small. No element of Akasha-Q has reached laboratory validation. The highest evidence level currently occupied on the wall is Theory, with a small number of Concept entries; the rungs Lab validation, Field validation, Dynamic platform, Flight/space and Operational demonstration are defined, published on the wall, and deliberately empty. An empty rung is information (DC-2): it is displayed, not hidden.
THE SEVEN-FIELD ENTRY
Every entry carries exactly seven fields. A field may not be left blank; where nothing is known the field carries the explicit token UNDETERMINED with a one-line statement of what would determine it.
| Field | Required form | Failure rule |
|---|---|---|
| 1 Claim id | AQ-CL-nnnn, allocated by the register custodian, never by the author; a revision carries the suffix R. | Author-allocated ids are rejected at admissibility. |
| 2 Claim | One sentence, as approved. The approved text is the used text; it is not a summary of a longer statement held elsewhere. | Multi-sentence or conditional-chained claims are split or rejected. |
| 3 Evidence level | Exactly one rung of the Part 12 ladder, shown as a chip. No intermediate or hyphenated levels. | A level with no artefact of that class in the trace is a strike condition, not a correction. |
| 4 Supporting experiments | Named artefacts with type and date: derivation, simulation run, bench record, review record. Each with the assumption set id it was executed under. | "Internal analysis" without an artefact id is not a supporting experiment. |
| 5 Replication status | One of: none / internal re-run / independent internal / external. Names the replicator and whether the replicator is an author of the original. | none is a valid and common value; it may never be omitted to imply more. |
| 6 Limitations | What the claim does not say, the assumption set on which it depends, the quantities that remain undetermined, and what observation would falsify it. | An entry whose limitations field restates the claim positively is returned. |
| 7 Trace reference | One evidence bundle id, AQ-EB-nnnn-x, resolving inside L6. | A trace that resolves only to a document folder, a person, or a meeting is not a trace. |
EVIDENCE LEVEL MOVEMENT
Level movement follows DC-4. A level falls on a single adverse observation — a failed re-run, a discovered model error, a withdrawn artefact — and the fall is recorded the same working day without a meeting, by the custodian, on the report of anyone in the programme. A level rises only through the full sequence: artefact produced, independent navigation audit passed, technical review at the gate, disposition recorded. There is no provisional or pending upgrade state; between the two, the entry sits at its lower level.
THE THREE-STEP NAVIGATION RULE
A claim may remain on the wall only while a non-author can reach its primary record in three steps or fewer, unaided, from the entry itself.
Figure 1 - the permitted navigation path, and the removal condition
CLAIM ON THE WALL
AQ-CL-0031 [ Theory ]
|
| step 1 : trace reference
v
EVIDENCE BUNDLE AQ-EB-0031-a
manifest: artefacts, authors, dates, assumption set ids
|
| step 2 : one named artefact from the manifest
v
ARTEFACT DER-L3-014 (derivation) | SIM-L2-007 (run) | REV-009 (review)
|
| step 3 : primary record
v
L6 EVIDENCE FABRIC ENTRY
raw output + assumption set + operator + immutable id (DC-5)
more than 3 steps, OR any step blocked for a non-author
-> REMOVE FROM WALL (not reword, not footnote)
Removal is not falsification. A removed claim may be true; the finding is only that the programme cannot currently show why it should be believed. Removed entries return to the gate at stage S1 and may be re-posted unchanged once the path exists. Removal is executed by the custodian on the navigator's finding and is not subject to panel override — a blocked path is cured by fixing the trace, never by argument.
WORKED ENTRY — STANDING
| Claim id | AQ-CL-0031 |
| Claim | Under assumption set AS-4, the L3 key-agreement construction specified for AQ-NODE-01 admits a finite-key analysis whose parameters are to be fixed by measurement on built hardware. |
| Evidence level | Theory — architecture tier CA |
| Supporting experiments | DER-L3-014 (derivation, under AS-4); REV-009 (internal technical review record of DER-L3-014). No bench artefact exists. |
| Replication status | none. DER-L3-014 has not been re-derived by a non-author and has had no external review. |
| Limitations | AS-4 assumes idealised source and detector behaviour that no built component has been shown to satisfy; detector-side manipulation is outside the assumption set; every parameter in the analysis is MODELLED (DC-1) and no numeric target is asserted. Falsified by any implementation whose measured source or detector characterisation lies outside AS-4. |
| Trace reference | AQ-EB-0031-a |
WORKED ENTRY — STRUCK AFTER CORRECTION
AQ-CL-0018 — Struck on 2026-08-14 by the custodian on the report of the simulation author.
| Claim (as posted) | In simulation of the AQ-NODE-01 acquisition path under channel model CM-2, the L2 timing residual stays inside the budget stated in AS-7. |
| Evidence level (as posted) | Simulation — tier CA |
| Strike reason | The harness that produced SIM-L2-007 instantiated a disciplining reference that the CA architecture does not contain. The run is valid as a run; it is a model of a node other than the one the claim names. The claim was therefore never entailed by its own trace. |
| Disposition | Entry retained on the wall, struck through, with the strike record navigable at step 1 (AQ-EB-0018-strike). Not deleted, not silently edited. |
| Reissued as | AQ-CL-0018R — Undetermined at Concept: "The CA architecture defines an L2 timing residual budget whose bound is undetermined pending a disciplining-source decision; no simulated or measured residual is claimed." |
Struck-claim handling. A strike is permanent and public within the register. The struck sentence remains legible so that a reader who saw it can find out what happened to it; the strike record names the reason, the date, the artefact at fault and the reissued claim id if any. A struck sentence may not be resubmitted verbatim. Any paraphrase of a struck sentence is treated as a resubmission of it and is refused at admissibility.
MAINTENANCE BETWEEN SHOWCASES
The wall is a continuously maintained register, not a document prepared for an audience. Showcases read the wall; they do not shape it.
| Activity | Cadence | Owner | Recorded output |
|---|---|---|---|
| Admission of new entries | Continuous, gate-only | Register custodian | Gate disposition id on the entry |
| Navigability audit — all entries changed since last review, plus a rotating sample of unchanged entries | Every programme review | Non-author navigator | "Last navigated" date per entry; removals |
| Level fall on adverse observation (DC-4) | Same working day, no meeting | Custodian, on any member's report | Level-fall record in L6 |
| Wall freeze before any DM-n demonstration | 10 working days prior | Custodian | Frozen manifest, hashed into L6. Additions barred; removals, strikes and level falls remain permitted |
| Question capture after a demonstration | Within 5 working days | Presenter and custodian | Each unanswerable question logged as a candidate claim or as a named undetermined metric |
| Full re-navigation of every entry | Annually and on any architecture revision | Navigator pool | Re-navigation report; entries not re-navigated are removed |
An entry that has not been navigated within one full review cycle is not assumed still good. It is marked UNCONFIRMED and displayed as such until a navigator clears it. Absence of a recent audit is a value, not a default to benign (DC-2).