PizarraContexto de trabajoDocumentos y registrosControles críticosRegistrosPreparaciónCondiciones bloqueantesAutorización
Aseguramiento / Técnico
PH6A-AHP-REV1.1 · ARCHITECTURE_HARDENING_ONLY

Phase 6A — Architecture Hardening Patch Rev.1.1

Formal Semantics · Deterministic Core · Evidence/Decision Separation · Governed Authority

/board + /erx + /wave1 + /s1 + /oei — linked, never re-run

Phase6A HOLDOEI ACTIVEAwaitingOwner 6OwnerAcknowledged 0EvidenceSubmitted 0SufficientForRetest 0BC09Protection ACTIVEPilotExposure PROHIBITEDPhase6B NOT_AUTHORIZEDPhase7 NO_GOQuestionnaire PH6A-W1-OEAVQ-REV1 · FROZEN_FOR_OWNER_VALIDATION · UNCHANGED
A

Architecture Hardening Executive View

The accepted architecture has been hardened so that facts, derived facts, advisory AI output and authorized decisions are formally distinct classes; evidence assessment, deterministic decision evaluation and authority resolution are executed by three non-overlapping engines; invalid operational states are structurally prohibited rather than merely detected; rules and configuration are versioned and reconstructable; and blocker closure is predicate-based rather than document-based. Twelve synthetic hardening regression families (H1–H12) were executed as design assurance. No architecture defect and no materially open state-model defect was found.

ARCHITECTURE_HARDENING_PATCH_REV1.1 = ACCEPTED

Design and simulation assurance only. No blocker closed, no owner acknowledged, no retest executed, no gate status changed.

Fact classes frozen
6
Engines separated
3
Invalid-state invariants
9
Reason codes governed
10
Property invariants
8
Hardening families executed
12
Architecture defects
0
Findings raised
3
B

Formal Fact Type System

Six frozen semantic classes; no implicit promotion is representable.

ClassMeaningMay authorizePromotion path
AUTHORITATIVE_FACTState mastered by the owning source system of record, observed at a pinned version and timestamp.ONLY_VIA_AUTHORITY_ENGINESource of record → Source Validation → qualified fact
FEDERATED_FACTFact reached across a federation link; truth remains with the source, never with the readiness layer.NOValidated mapping → Source Validation → qualified fact (never authoritative)
DERIVED_FACTComputed from one or more qualified facts by a versioned rule; carries a full provenance chain.NORule Evaluation only; never promoted to AUTHORITATIVE_FACT
RULE_EVALUATIONDeterministic output of a pinned rule version over pinned inputs. A statement of logic, not of authority.NODecision Engine output → offered to Authority Engine
AI_ADVISORY_OUTPUTProbabilistic assistance: gap detection, drafting, comparison, explanation. Structurally excluded from operational state.NOAI_SUGGESTION → ENGINEERING_REVIEW → COMPETENT_OWNER_APPROVAL → RULE_FORMALISATION
AUTHORIZED_DECISIONA competent-authority act with verified authority context, scope, validity and attributable actor.ONLY_VIA_AUTHORITY_ENGINESource Validation → Rule Evaluation → Authority Resolution → Competent Authority Decision
  • ·FEDERATED_FACT ≠ AUTHORITATIVE_FACT
  • ·DERIVED_FACT ≠ AUTHORITATIVE_FACT
  • ·AI_ADVISORY_OUTPUT ≠ AUTHORIZED_DECISION
  • ·RULE_EVALUATION ≠ AUTHORIZED_DECISION
  • ·No implicit promotion between classes is representable in the type system
C

Evidence / Decision / Authority Engine Boundaries

EvidenceEngine ≠ DecisionEngine ≠ AuthorityEngine (frozen)

EVIDENCE_ENGINE
  • ·Intake and provenance capture
  • ·Evidence admission control
  • ·Criterion-level assessment
  • ·Contradiction detection
  • ·EvidenceQuality determination

Must not authorize work and must not produce a readiness verdict.

Consumes: Submitted artefacts, source refs, criteria

Produces: EvidenceCriterionAssessment, contradictions, quality

DECISION_ENGINE
  • ·Deterministic evaluation of qualified facts
  • ·Readiness logic and no-compensation enforcement
  • ·READY / CONDITIONAL / HOLD / STOP determination
  • ·Reason code emission

Must not determine human authority and must not admit evidence.

Consumes: QualifiedFacts, RuleSet, Configuration

Produces: DecisionState, ReasonCodes, RequiredActions

AUTHORITY_ENGINE
  • ·Resolve competent authority and scope
  • ·Delegation, validity and revocation
  • ·Decision rights evaluation

Must not determine source truth and must not compute readiness.

Consumes: DecisionRight records, actor identity, scope context

Produces: AuthorityResolutionState, AuthorityDecisionRef

D

Invalid-State Invariant Register

Violations return INVALID_STATE_TRANSITION rather than a corrected state.

IDInvariantEnforcementReturnTestResult
INV-01AUTHORIZED ⇒ AuthorityVerifiedAuthority Engine port; AUTHORIZED is unconstructable without an AuthorityDecisionRefINVALID_STATE_TRANSITIONH1 / H12PASS
INV-02CLOSED ⇒ ClosurePredicateSatisfiedBlocker closure accepts a satisfied predicate object, never an Evidence_IDINVALID_STATE_TRANSITIONH7PASS
INV-03SUFFICIENT_FOR_CLOSURE ⇒ NoMaterialContradictionEvidence Engine downgrades quality when a material contradiction is openINVALID_STATE_TRANSITIONH3 / H6PASS
INV-04READY ⇒ MandatoryConditionsSatisfiedNo-compensation evaluation; a satisfied gate can never offset an unsatisfied mandatory gateINVALID_STATE_TRANSITIONH2 / H4PASS
INV-05AI_ADVISORY_OUTPUT ⇒ CannotAuthorizeFirewall type boundary; advisory outputs are not accepted as Decision Engine inputINVALID_STATE_TRANSITIONH1 / H5PASS
INV-06MandatorySourceUnverifiable ⇒ READY prohibitedSource Validation marks fact UNVERIFIABLE; verdict floor becomes HOLDINVALID_STATE_TRANSITIONH4 / P-02PASS
INV-07StewardshipVacant ∧ StewardshipRequired ⇒ DISABLED_SAFEStewardship port fails closed; dependent capability disabled, never defaultedINVALID_STATE_TRANSITIONP-08PASS
INV-08AuthorityUnverifiable ⇒ AUTHORIZED prohibitedAuthorityResolutionState UNRESOLVED can never convert to local authorityINVALID_STATE_TRANSITIONH12PASS
INV-09EvidenceQuality = PARTIAL ⇒ CLOSED prohibitedClosure predicate requires SUFFICIENT_FOR_CLOSURE on every mandatory criterionINVALID_STATE_TRANSITIONH3PASS
E

GovernedOperationalEvent Schema

Storage technology is deliberately unspecified: the model mandates append-only causal reconstruction, not a database or ledger product.

FieldPurpose
Event_IDUnique immutable identity of the governed event
Event_TypeClassified event semantics (intake, evaluation, authorization, conflict…)
Object_IDGoverned object the event concerns
Object_ClassClass of the governed object (configurable ObjectClass mapping)
Location_IDOperational context container in force
WorkContext_IDWork package / job card execution context
Actor_IDAttributable actor; never a system alias for a human act
Authority_ContextResolved authority, scope, delegation and validity at event time
Previous_StateState prior to transition
New_StateState after transition
Source_RefsPinned source records and versions consumed
Evidence_RefsEvidence artefacts and criterion assessments consumed
TimestampInjected governed time of the event
Causal_ParentDirect causal predecessor event
Correlation_IDEnd-to-end correlation across engines and shells
Rule_IDRule applied at evaluation
Rule_VersionExact rule version active at Decision_Timestamp
Configuration_VersionGoverned configuration version in force
OutcomeResult of the event (accepted, rejected, conflicted, blocked)
Reason_CodeGoverned reason code explaining the outcome
F

Rule Versioning Register

A decision must be reconstructable against the exact rule version active at its Decision_Timestamp; rule bodies are immutable once effective.

Rule_ID · Rule_Version · Effective_From · Effective_To · Configuration_Version · Decision_Timestamp · Input_Fact_Refs · Output_State · Owner · ApprovalRef

RuleVersionScopeOwnerApprovalRefReconstructable
RL-NOCOMP-011.2No-compensation gate evaluationES&H GovernanceTO_BE_GOVERNEDYES
RL-SRC-VALID-011.1Mandatory source availability & information conditionEnterprise ArchitectureTO_BE_GOVERNEDYES
RL-SIMOPS-CUM-011.3Cumulative location SIMOPS evaluation (CUM-1…CUM-7)Area OperationsTO_BE_GOVERNEDYES
RL-EVQ-011.1Criterion-specific evidence quality resolutionPhase 6A GovernancePH6A-SIMCLOSE-F-SIM-05YES
RL-CLOSE-PRED-011.0Blocker closure predicate evaluationPhase 6A GovernancePH6A-AHP-REV1.1YES
RL-AUTH-RES-011.0Authority resolution, delegation and revocationDecision Rights StewardTO_BE_GOVERNEDYES
G

Code / Configuration / Data Contract

CODE
What is possible
  • ·Types and fact classes
  • ·Allowed state structures
  • ·Engine contracts
  • ·Invariant enforcement
  • ·Security boundaries

Must not contain project-specific authority, thresholds or validity rules

CONFIGURATION
What is permitted
  • ·Project rules
  • ·ObjectClass mappings
  • ·Authority mappings
  • ·Applicability
  • ·Validity criteria
  • ·Owner-confirmed thresholds

Must not encode current facts or evidence; every entry is versioned and owner-approved

DATA
What is true now
  • ·Current facts
  • ·Source state
  • ·Evidence
  • ·Governed events

Must not silently redefine rules or authority

H

AI-to-Rule Firewall

Advisory outputs carry class AI_ADVISORY_OUTPUT and are rejected at the Decision Engine input port.

AI may
  • ·Identify gaps and inconsistencies
  • ·Propose candidate rules
  • ·Compare documents and revisions
  • ·Generate draft language
  • ·Provide explanations of governed outcomes
  • ·Suggest controls for human consideration
AI may not
  • ·Create executable production rules directly
  • ·Modify authority or delegation
  • ·Modify source precedence
  • ·Approve thresholds or validity periods
  • ·Release, authorize or clear work
AI_SUGGESTIONENGINEERING_REVIEWCOMPETENT_OWNER_APPROVALRULE_FORMALISATIONTESTVERSIONCONTROLLED_RELEASE
I

Observed / Forecast / Authorized State Model

PlaneMeaningMay release work
OBSERVED_STATEWhat the authoritative and federated sources currently report, with pinned versions.NO
FORECAST_STATEProjected readiness over the lookahead horizon from expiries, planned work and known constraints.NO
AUTHORIZED_STATEWhat a competent authority has actually authorized, with attributable evidence.YES — only via Authority Engine
  • ·FORECAST_STATE ≠ AUTHORIZED_STATE
  • ·ForecastReady ⇒ no automatic authorization
  • ·Forecast may identify emerging gaps, support lookahead and trigger preventive action
  • ·Forecast may never release work
J

SIMOPS Compositional Model

EntityMeaning
PairwiseInteractionCompatibility assessment between exactly two concurrent activities in a location.
ConcurrentWorkSetThe complete set of active and planned activities sharing the location context.
AggregateLocationStateCumulative verdict over the full ConcurrentWorkSet; dominant over pairwise results.
  • ·PAIRWISE_PASS ≠ LOCATION_PASS
  • ·A compatible with B ∧ B compatible with C ⇏ A+B+C compatible
  • ·Aggregate evaluation must consume the complete LocationConcurrentWorkSet, never a pairwise sample
  • ·Cumulative STOP is rendered as dominant over any pairwise PASS (F-SIM-04 correction preserved)
K

EvidenceCriterionAssessment Model

EvidenceQuality is a property of the (Evidence_ID, Criterion_ID) pair, never of a document. One artefact may fully support one criterion, partially support another and be irrelevant to a third.

Evidence_ID · Criterion_ID · ER_ID · Admission_Result · Deep_Assurance_Result · Sufficiency · Limitation · Contradiction · Reviewer · Timestamp

EvidenceCriterionSufficiencyLimitationContradiction
SYN-EV-01 (synthetic participation map)BC01-C1 ParticipationMode declaredSUFFICIENT_FOR_CLOSURENONE
SYN-EV-01 (synthetic participation map)BC01-C2 Write authority boundaryPARTIALWrite boundary stated for 4 of 7 systemsNONE
SYN-EV-01 (synthetic participation map)BC01-C3 Failure behaviour on source lossNONECriterion not addressed by artefactNONE

Demonstration rows are synthetic (H6) and carry no owner attribution.

L

Blocker Closure Predicate Model

Predicate contents remain configurable per blocker and are owner-approved, not code-fixed.

Prohibited: Evidence_ID → BlockerClosed
BCxx_Closure = ⋀(MandatoryCriteria demonstrated) ∧ RetestPassed ∧ AuthorityValid ∧ NoMaterialContradiction ∧ ResidualControlsGoverned(if CWC)
BlockerPredicateSatisfiedMissing
BC-01C1 ∧ C2 ∧ C3 ∧ RetestPassed ∧ AuthorityValid ∧ NoMaterialContradictionNOC1, C2, C3 (no operational evidence received)
BC-02IAM-01 ∧ IAM-02 ∧ IAM-03 ∧ IAM-04 ∧ IAM-05 ∧ IAM-06 ∧ RetestPassed ∧ AuthorityValidNOIAM-01…IAM-06
BC-03LifecycleGovernanceDecision ∧ TechnicalEnforcement ∧ RetestPassed ∧ AuthorityValidNOBoth criteria
BC-04StewardshipAssigned ∧ StewardshipAccepted ∧ AuthorityValid ∧ NoVacancyNOAll criteria
BC-05CA04Classification ∧ Retention ∧ DataMinimisation ∧ AuthorityValidNOAll criteria
BC-06PathElection(A|B) ∧ DecisionAuthorityRecorded ∧ ResidualRiskRecordedNOElection not made (ESC-03 active)
BC-07L1 ∧ L2 ∧ L3 support model ∧ DrillEvidence ∧ AuthorityValidNOAll criteria
BC-08FieldPrerequisites(all dimensions) ∧ RetestPassed ∧ AuthorityValidNOAll criteria
BC-09ProspectiveBaselineStarted ∧ MetricSourcesConfirmed ∧ NoContaminationEventNOProspective collection not started — time-irreversible
M

Pure Core / Effectful Shell

Pure Decision Core

Inputs

  • ·QualifiedFacts
  • ·RuleSet
  • ·AuthorityContext
  • ·Configuration

Outputs

  • ·DecisionState
  • ·ReasonCodes
  • ·RequiredActions

Prohibited

  • ·Writes to external systems
  • ·Email or notification
  • ·Mutation of source records
  • ·UI effects
  • ·Clock or network access
Effectful Shell

Handles

  • ·Notifications
  • ·Controlled writes
  • ·Snapshots
  • ·Integrations
  • ·User interaction
  • ·Offline sync
  • ·External source access

Every shell effect is expressed as a GovernedOperationalEvent with Command_ID idempotency.

N

Reason Code Register

The set is configurable and not exhaustive; new codes require owner, resolvability, evidence and escalation route.

Reason codeOwnerResolvable byRequired evidenceEscalation
SOURCE_UNVERIFIABLESystem OwnerRestored source or governed snapshotSource availability confirmation with pinned versionESC-01
AUTHORITY_UNVERIFIABLEDecision Rights StewardAuthority resolution or valid delegationDecisionRight record in forceESC-02
COMPETENCY_NOT_CONFIRMEDGeneral ForemanValid qualification recordCompetency source record with validity windowESC-02
CRITICAL_CONTROL_NOT_VERIFIEDES&H SupervisorVerification of control availability and effectivenessCritical control verification recordESC-03
SIMOPS_CONFLICTArea SuperintendentDeconfliction of the ConcurrentWorkSetAggregate location evaluationESC-03
RULE_CONTRADICTIONRule Governance OwnerRule reconciliation and re-versionApproved rule versionESC-01
REQUIRED_EVIDENCE_MISSINGEvidence OwnerSubmission of the missing criterion evidenceArtefact addressing the named criterionESC-01
LOCATION_CONTEXT_CHANGEDLocation StewardReassessment against the new contextLocation context change eventESC-03
VERSION_CONFLICTWork Control LeadCompare, reconcile, authorized retryReconciliation decision with attributable actorESC-02
STEWARD_UNAVAILABLEGovernance SponsorStewardship assignment or governed fallbackAccepted stewardship assignmentESC-04
O

Provenance Model

Every DERIVED_FACT, RULE_EVALUATION, AI_ADVISORY_OUTPUT and READINESS_STATE carries a navigable ProvenanceChain; a state without a reconstructable chain is not a governed state.

  1. CurrentState
  2. ← RuleEvaluation (Rule_ID + Rule_Version + Configuration_Version)
  3. ← InputFact (fact class + pinned projection version)
  4. ← Evidence (Evidence_ID + Criterion_ID assessment)
  5. ← SourceRecord (source object + source version)
  6. ← SourceOwner
  7. ← EffectiveTimestamp
P

Determinism & AI-Off Equivalence

Deterministic core invariant

SameFacts + SameRuleVersion + SameConfiguration = SameDecision

Violation returns DETERMINISM_VIOLATION

AI output is excluded from operational determinism because AI remains advisory; operational state may not depend on probabilistic output.

AI-off equivalence

AI_ON_OPERATIONAL_STATE = AI_OFF_OPERATIONAL_STATE for all material safety/readiness decisions

Allowed differences

  • ·Speed
  • ·Assistance
  • ·Explanation richness
  • ·Draft quality

Not allowed

  • ·Different safety state
  • ·Different authority outcome
  • ·Different HOLD/STOP decision caused solely by AI availability

H4 (100 repeated evaluations, identical state) and H5 (AI ON vs AI OFF, identical material decision).

Q

Canonical Semantic Dictionary

No semantic mapping may exist without an assigned SemanticOwner and version.

Canonical termCanonical meaningSource systemSource termTransformationAllowed lossOwnerVer
LocationOperational context container; aggregates concurrent work, never authorizes.Canonical Location Register (ADR-14 Option C)LocationNodeExact key where registered; controlled cross-reference otherwiseSub-elevation granularity below registered nodeLocation Steward1.0
WorkPackageGoverned bundle of work with a single readiness verdict.P6 / Q4Activity / Work OrderValidated mapping on work order keyPlanning float semanticsWork Control Lead1.0
JobCardExecutable field task with discipline, crew and control requirements.Q4Task / TSI itemApproved Q4 state mapping onlyVendor-specific sub-statusBusiness Product Owner1.1
CriticalControlControl whose loss is non-compensable during execution.ForwoodCritical Control VerificationExact key; verification state mapped 1:1NONE permittedES&H Governance1.0
RestrictionCondition constraining execution until closed or verified.MultipleConstraint / NCR / PunchDerived; the layer never masters the recordOriginating workflow substatesWork Control Lead1.0
ReadyAll mandatory conditions satisfied with verifiable sources; no compensation applied.IRDEDerived by RL-NOCOMP-01NONE permittedES&H Governance1.2
AuthorizedA competent authority act with verified scope and validity.Authority EngineDecisionRight resolutionNONE permittedDecision Rights Steward1.0
ApprovedA transactional state in a source system; not equivalent to Authorized.Aconex / Q4ApprovedTransactional mapping; never promoted to AuthorizedReviewer commentaryDocument Control1.0
ValidWithin a governed operating validity window at decision time.ConfigurationValidity windowOwner-confirmed validity per object classNONE permittedConfiguration Owner1.0
CompetentQualification current, unsuspended and matched to the activity role.Competency systemQualificationRole-to-qualification mappingTraining provider detailGeneral Foreman1.0
AvailablePresent, reachable and fit for the required use at decision time.MultipleAvailable / In servicePer object class mapping, evidence-backed onlyVendor availability substatesArea Operations1.0
R

LocationOperationalContext Aggregate

INHERIT CONTEXT — NEVER INHERIT AUTHORIZATION

Active workPlanned workJobCardsDisciplinesEquipmentPermits / work controlsIsolationsRestrictionsEnvironmental stateSIMOPSLessonsEventsContinuity across shifts

Invariant: Location does not authorize.

S

Property-Based Invariant Test Register

Simulation / design assurance only — creates no OperationalClosureEvidence.

IDPropertyCasesResultEvidence class
P-01NoAuthority ⇒ NeverAuthorized250PASSSIMULATION_EVIDENCE
P-02MandatorySourceUnverifiable ⇒ NeverReady250PASSSIMULATION_EVIDENCE
P-03AIOutput ⇒ NeverMaterialAuthority250PASSSIMULATION_EVIDENCE
P-04InheritedContext ⇒ NeverInheritedAuthorization250PASSSIMULATION_EVIDENCE
P-05PairwisePass ⇒ AggregateStillRequiresEvaluation250PASSSIMULATION_EVIDENCE
P-06EvidencePartial ⇒ NeverClosed250PASSSIMULATION_EVIDENCE
P-07VersionConflict ⇒ NoSilentOverwrite250PASSSIMULATION_EVIDENCE
P-08StewardshipRequiredAndVacant ⇒ DisabledSafe250PASSSIMULATION_EVIDENCE
T

Idempotency / Concurrency Control

Idempotency

SameCommand_ID → SingleMaterialEffect

Key: Command_ID

  • ·Authorization
  • ·Evidence record
  • ·Governed event
  • ·State transition

Replay returns the original Outcome and Event_ID; no new material effect is created.

Optimistic concurrency

Base_Version ≠ Current_Version ⇒ VERSION_CONFLICT

Fields: Base_Version · Current_Version

LAST_WRITE_WINS is prohibited for material governed objects

Resolution: compare → reconcile → authorized retry

U

Decision Rights Data Model

DecisionRight is governed data. No named person and no project-specific role may be hard-coded into the deterministic core.

RoleObjectClassScopeLocationActivityRiskRegisterTypeEffectiveFromEffectiveToDelegateSourceAuthorityVersion
V

Hardening Regression Results

No real owner evidence was used; no blocker state was touched.

Families 12PASS 11PASS_WITH_FINDING 1FAIL 0SIMULATION_EVIDENCE — DESIGN_ASSURANCE ONLY
IDFamilyScenarioExpectedActualResultFinding
H1Fact PromotionAttempt AI_ADVISORY_OUTPUT → AUTHORIZED_DECISION without review pipelineRejected — INVALID_STATE_TRANSITIONINVALID_STATE_TRANSITION; advisory output not accepted at Decision Engine portPASS
H2Evidence / Decision SeparationSubmit admissible evidence against BC-01 criteriaNo READY/AUTHORIZED created by submissionEvidenceCriterionAssessment created; verdict unchanged; blocker OPENPASS
H3Invalid StatesAttempt CLOSED with EvidenceQuality = PARTIALRejectedINVALID_STATE_TRANSITION (INV-09)PASS
H4Determinism100 repeats: same facts, rule version, configurationIdentical operational state each run100/100 identical DecisionState and ReasonCodesPASS
H5AI-Off EquivalenceAI ON vs AI OFF over 40 material readiness decisionsIdentical material decision40/40 identical; differences limited to explanation textPASS
H6Evidence Criterion GranularityOne artefact across three criteria at differing sufficiencyPer-criterion sufficiency retainedSUFFICIENT_FOR_CLOSURE / PARTIAL / NONE recorded separatelyPASS
H7Blocker PredicateAll criteria but one demonstratedClosure preventedPredicate false; blocker remains OPEN with missing criterion namedPASS
H8SIMOPS CompositionAll pairwise PASS, cumulative conflict present in ConcurrentWorkSetAggregate HOLD/STOPAggregateLocationState = STOP; pairwise PASS suppressed as non-dominantPASS
H9IdempotencyDuplicate Command_ID replayed five timesSingle material effectOne event, one transition; four replays returned original outcomePASS
H10ConcurrencyTwo actors act on the same Base_VersionSecond action VERSION_CONFLICTVERSION_CONFLICT; no overwrite; reconciliation requiredPASS
H11ProvenanceReconstruct a HOLD state to source, evidence, rule, version and ownerFull chain navigableChain reconstructed; one advisory node lacked a rendered owner labelPASS_WITH_FINDINGF-AHP-01
H12Decision RightsTechnically enabled user without a DecisionRight attempts authorizationAuthorization refusedAuthorityResolutionState UNRESOLVED; AUTHORIZED prohibited (INV-08)PASS
W

Findings Register

Any ARCHITECTURE_DEFECT or material STATE_MODEL_DEFECT would return HARDENING_PATCH_HOLD. None found.

IDClassificationStatementMaterialityDispositionOwner
F-AHP-01PRESENTATION_DEFECTProvenance view rendered an advisory node without its SemanticOwner label; the underlying chain was complete.NON_MATERIALCORRECTED IN PATCH — owner label now mandatory in the provenance rendering contract.Prototype presentation
F-AHP-02CONFIGURATION_DEFECTGoverned operating validity remains TO_BE_GOVERNED for several object classes; configuration cannot be owner-approved before BC-05 evidence exists.NON_MATERIALCONTROLLED_OPEN — carried against BC-05; not an architecture defect.Configuration Owner
F-AHP-03SEMANTIC_DEFECTApproved (source transactional) and Authorized (competent authority act) were previously conflated in narrative surfaces.NON_MATERIALCORRECTED IN PATCH — separated in the Canonical Semantic Dictionary with distinct owners.Semantic governance
ArchitectureDefects 0MaterialStateModelDefects 0CorrectedInPatch 2ControlledOpen 1
X

Patch Impact Assessment

No existing baseline or gate status changed.

ItemBeforeAfterChange
QuestionnaireBaseline PH6A-W1-OEAVQ-REV1FROZEN_FOR_OWNER_VALIDATIONFROZEN_FOR_OWNER_VALIDATIONUNCHANGED
Wave 1 Evidence Requests6 (ER-01/06/07/11/16/17)6 (ER-01/06/07/11/16/17)UNCHANGED
AwaitingOwner66UNCHANGED
OwnerAcknowledged00UNCHANGED
EvidenceSubmitted00UNCHANGED
SufficientForRetest00UNCHANGED
BC-01…BC-09OPENOPENUNCHANGED
Phase 6AHOLDHOLDUNCHANGED
Phase 6B / Phase 7NOT_AUTHORIZED / NO_GONOT_AUTHORIZED / NO_GOUNCHANGED
BC09Protection / PilotExposureACTIVE / PROHIBITEDACTIVE / PROHIBITEDUNCHANGED
SimulationBaselineACCEPTED_AND_FROZENACCEPTED_AND_FROZENUNCHANGED
Architecture semanticsAccepted A–AE baselineHardened: fact classes, engine separation, invariants, predicatesHARDENED — no baseline reopened
29 / 30

Acceptance & Final State

Acceptance criterionResult
All critical invariants passPASS
No architecture defect remainsPASS
No state-model defect materially openPASS
AI authority remains impossiblePASS
Deterministic operational core preservedPASS
Evidence / Decision / Authority separation demonstratedPASS
Criterion-specific evidence assessment worksPASS
Blocker closure remains predicate-basedPASS
Idempotency and concurrency controls passPASS
No existing baseline or gate status changedPASS
ARCHITECTURE_HARDENING_PATCH_REV1.1ACCEPTED
ArchitectureFatalFlaws0
CriticalInvariantsPASS
EvidenceDecisionAuthoritySeparationVERIFIED_IN_DESIGN
DeterministicCoreVERIFIED_IN_DESIGN
AIOffEquivalenceVERIFIED_IN_DESIGN
InvalidStateProtectionVERIFIED_IN_DESIGN
CriterionSpecificEvidenceAssessmentACTIVE
PredicateBasedClosureACTIVE
SimulationEvidenceImpactDESIGN_ONLY
QuestionnaireBaselineUNCHANGED
OEIACTIVE
AwaitingOwner6
OwnerAcknowledged0
EvidenceSubmitted0
SufficientForRetest0
Phase6AHOLD
BC09ProtectionACTIVE
PilotExposurePROHIBITED

Final Governing Principles

  • ·Facts are not decisions.
  • ·Evidence is not authority.
  • ·Prediction is not authorization.
  • ·AI is advisory, never authoritative.
  • ·Same governed inputs must produce the same governed decision.
  • ·Invalid operational states must be prevented, not merely detected.
  • ·Blockers close by predicates, not by document presence.
  • ·Context may be inherited; authorization never is.