AQ-SDD-001 · Akasha-Q Charter

The seventeen-section dossier that governs every Akasha-Q demonstration run: what each section must contain, the seal boundary between pre-registered sections 1-9 and post-run sections 10-17, and the acceptance test that lets a section pass review.

AQ-SDD-001 — Scientific Demonstration Dossier: Template and Completion Rules

1. PURPOSE AND STANDING

The Scientific Demonstration Dossier (SDD) is the record produced for every demonstration run in the Akasha-Q programme, from the first optical-bench measurement on rung 1 of the reach ladder through to any later run on a dynamic platform. One run produces exactly one dossier. A dossier records a question asked under declared conditions; it does not record an outcome the programme hoped for.

No element of Akasha-Q has been built and nothing has reached laboratory validation. Every dossier that exists today therefore stands at Concept or Theory, and no dossier may assert a higher Part 12 evidence level until a sealed run has been executed and independently reviewed under these rules. The dossier is the only route by which an article, a layer or a design commitment moves up the evidence ladder. A claim that cannot be traced to a sealed dossier is not a programme claim. Dossiers are held in the restricted-access record and carry a release decision in section 17; they are not published by default.

2. THE PRE-REGISTRATION BOUNDARY

SEAL RULE

Sections 1 to 9 are written, reviewed and sealed before the first data acquisition of the run. The seal is a hash over the text of sections 1 to 9 written into the Evidence Fabric with a timestamp. Sections 10 to 17 may only be authored after the run has ended. Any edit to sections 1 to 9 after the seal voids the dossier: it is marked VOID, retained, and replaced by a new dossier identifier carrying a supersedes link. Re-running after a voided seal is a new run with a new run identifier.

The boundary exists because the failure mode this programme most needs to prevent is not a bad measurement — it is a criterion adjusted after the data arrives. Section 9 fixes what would count as success and what would count as failure while both answers are still equally unwelcome.

Figure 1 - Dossier lifecycle and the seal boundary

   DRAFT  (sections 1-9 open, freely edited)
     |
     |  review R1 scientific + R2 safety (L9)
     v
   SEAL ---- hash(sections 1-9) + timestamp -> Evidence Fabric (L6)
     |                                  \
     |  run executes                     \  any edit to 1-9 after seal
     v                                    v
   RUN RECORD  (10, 11 append-only)     VOID -> new dossier id
     |                                          supersedes: AQ-SDD-nnn
     v
   ANALYSIS  (12, 13, 14, 15, 16)
     |
     +--> 13 Failure Observations: non-empty, OR a signed
     |    detection-coverage justification. No third option.
     v
   DISPOSITION (17) -> evidence level asserted, or withheld

3. SECTIONS 1 TO 9 — PRE-REGISTERED BEFORE THE RUN

Sec	Title	What must be written	Acceptance test
1	Dossier Identity and Registration	Dossier and run identifiers; scenario name; article under test using canonical `AQ-NODE-n` naming (noting the `AKQ-NODE-n` form used in Parts up to 12 for the same article); reach-ladder rung; architecture tier CA, Y1, FR or LH; authors; the two reviewers; evidence level currently held and evidence level sought; supersedes and superseded-by links.	Every identifier resolves in the record. The level sought is exactly one rung above the level held for that article, or the section states in one sentence why a skip is proposed and who approved it.
2	Objective	What the run is for in engineering terms, and the specific programme decision it informs.	An engineer outside the team can state from this section alone what would change in the programme if the run produced nothing usable.
3	Research Question	A single interrogative sentence, answerable by the metrics of section 9, scoped to one article on one rung.	Every noun and qualifier in the question maps to a numbered metric row. A qualifier with no metric behind it is a rejection.
4	Hypothesis and Falsification Condition	The expected relationship stated so that it can be wrong, plus the exact observation that would falsify it.	The falsifying observation is detectable with the instrumentation listed in section 7. If no possible observation would falsify the hypothesis, the section is rejected.
5	System Version and Configuration Record	Exact identity of everything that can change behaviour: article revision, FPGA bitstream hash, acquisition firmware, AQ-TSE-01 build and rule-set version, L5 policy identifier, Evidence Fabric schema version, calibration certificates and dates, clock discipline source, software commit identifiers, configuration file hashes.	Every field carries a value that could re-create the configuration. No field reads “current”, “latest” or “as built”. The seal hash covers this section verbatim.
6	Architecture Exercised	Each layer L0 to L9 marked exercised, stubbed or not present; each trust boundary TB-1 to TB-5 marked the same way; which decision boundaries B-1 to B-5 the run traverses; which conflicts CF-1 to CF-7 are deliberately induced, if any. L8 is marked advisory only and out of the decision path.	No layer or boundary is blank — DC-2 applies to the dossier itself. Every stub names its substitute and states in one sentence what the substitute does not represent.
7	Protocol	Numbered executable steps, each naming the instrument used, the artefact recorded, and the operator action. Includes warm-up, tail period, recording start and stop markers, and the abort criteria with abort authority.	A second operator can execute the protocol without asking a question. Each step names the artefact it produces; a step that records nothing states so explicitly.
8	Input and Fault Conditions	Controlled inputs with their ranges; environmental variables recorded but not controlled; injected faults referenced to F-1 to F-14 with injection method, intended detection layer and expected trust-state trajectory. A control run states “no injected faults” and enumerates the faults deliberately excluded.	Excluded faults are enumerated, never implied. Every injected fault has a method and an expected trajectory; a fault with no expected trajectory cannot be scored and is rejected.
9	Metrics, Success and Failure Criteria	Each metric with identifier, definition, unit, DC-1 tag class (MEASURED, DERIVED or MODELLED), instrument, recording method, success criterion and failure criterion. Where no target exists, the row states that the target is undefined and states the rule by which a target will later be established from the measured distribution.	Every criterion is decidable from recorded data alone, with no post-hoc judgement. No criterion of the form “as good as possible”. No metric may be added after the seal.

4. SECTIONS 10 TO 17 — COMPLETED AFTER THE RUN

Sec	Title	What must be written	Acceptance test
10	Execution Record	Operators, wall-clock window, the sequence as actually executed, and every deviation from section 7 with its time and reason. Aborts and restarts with the identifier of any superseding run.	The deviation list is non-empty or carries the operator’s explicit “no deviation” against their name. Step count and duration reconcile with section 7.
11	Raw Data and Custody	Dataset identifiers, hashes, storage locations, capture-clock provenance, chain of custody, and anything discarded with the reason for discarding it.	Every metric in section 9 has at least one dataset capable of producing it. All hashes verify. No deletion without a recorded reason and an approver.
12	Measured Results	A value per metric with its DC-1 tag class, its uncertainty, and a reference into section 11. Derived values show the derivation; modelled values name the model, its version and its assumption set.	Every value is tagged and traceable. A metric that could not be evaluated is written “undetermined” with the reason (DC-2); omission is a rejection.
13	Failure Observations	Every anomaly, unexpected trust-state transition, abstention, out-of-range reading, instrument fault, operator error and unexplained gap — whether or not it affected the result.	Non-empty, or accompanied by the detection-coverage justification defined in section 5 of this document and signed by the reviewer, not the author.
14	Analysis and Interpretation	The answer to section 3, a verdict on section 4, and an explanation for each section 13 observation or an explicit statement that it remains unexplained.	The verdict is exactly one of supported, not supported, or undetermined. No metric introduced here. No interpretation resting on an assumption absent from section 16.
15	Evidence Level Determination and Claim Statement	The Part 12 evidence level this run supports for the named article and layers; the exact sentence the programme is permitted to use externally; and the sentences explicitly not permitted on the strength of this run.	The level does not exceed what the rung can carry — a bench run cannot yield Field validation. The permitted sentence names its assumption set and contains no unqualified descriptor. Signed by a reviewer who is not an author.
16	Threats to Validity and Limitations	Assumption set, instrument limits, sample size, single-article risk, boundaries and layers not exercised, and what the run does not test.	At least one limitation that would materially change the interpretation if it were removed. “None” is not an acceptable entry.
17	Disposition and Follow-on	Next action: repeat, redesign, escalate a rung, or stop. Open F-* and CF-* items raised by the run. The identifier of the next dossier if one is opened. The release decision under the restricted-access record.	Names one next action with a named owner, or records the decision not to proceed together with its reason. An empty disposition blocks the dossier from closing.

5. SECTION 13 IS MANDATORY

NO EMPTY SECTION 13

A dossier submitted with section 13 blank is rejected without review. “None observed” is permitted only with a written detection-coverage justification: which failure modes among F-1 to F-14 the instrumentation of this run could have detected, at what sensitivity, over what fraction of the run, and why their absence is therefore informative rather than merely unrecorded. The justification is signed by the reviewer, not the author. An article that never produces failure observations is more often under-instrumented than well behaved. DC-2 applies to the dossier as it applies to the system: absence is a value, and it never defaults to benign.

6. TAGGING AND CLAIM DISCIPLINE

DC-1 in the dossier. Every number in sections 12 and 14 carries MEASURED, DERIVED or MODELLED. A modelled value beside a measured one without its tag is a rejection, not a formatting defect.
DC-3 in scoring. An abstain outcome is a permitted system behaviour and is never scored as a run failure by itself. It is recorded in section 13 and must carry the input condition that produced it.
DC-4 in interpretation. A trust state that fell during the run cannot be described as recovered unless the full recovery sequence appears in the record. Section 14 may not narrate a return to NORMAL that the Evidence Fabric does not show.
DC-5 as an acceptance test, not an aspiration. Reconstruction is scored by a reviewer replaying the consequential transitions from the Evidence Fabric alone, without console logs and without asking the operator.
Prohibited in every section: unhackable, impossible to intercept, universally quantum-secure, information-theoretically secure at system level, satellite-ready, space-proven, operational. Also prohibited: the words secure, resilient, autonomous, high-precision, low-latency and quantum-secure used as bare adjectives. Name the measured quantity and its assumption set, or state that the metric is undefined.
No target invention. Where section 9 has no threshold, it says so and states the rule for deriving one later from measurement. A threshold that first appears in section 12 is a seal violation.

7. REVIEW ROLES, SEAL MECHANICS AND IDENTITY

Two reviewers are named in section 1 and both sign twice, once at seal and once at close. R1 (scientific) tests sections 3, 4, 9, 14 and 15 — that the question is answerable, the hypothesis falsifiable, the criteria decidable and the claim no larger than the data. R2 (safety and governance, L9) tests sections 6, 7, 8, 10 and 17 — abort authority, boundary markings, fault handling and disposition. Neither reviewer may be an author of the section they clear.

Dossier identifiers are of the form AQ-SDD-nnn, with a scenario suffix where a template carries worked scenarios, for example AQ-SDD-001-S1. A voided dossier keeps its identifier permanently; identifiers are never reused and never silently corrected. Where a later run supersedes an earlier one, both dossiers remain readable and the link is bidirectional, so that any claim can be walked back to the first run that supported it and forward to whatever replaced it.