Invariant Classification Audit

Session: Session B of invariant-types governance cycle Date: 2026-04-17 Auditor session: session-b-2026-04-17-invariant-audit Total invariants audited: 72

Schema reference: docs/architecture/invariant_types.md Planning doc: docs/thoughts/2026-04-17_invariant_types_governance.md

---

CHECKPOINT REACHED

Per Q4 of the planning doc (§7 pending_substrate_reconfirmation threshold rule), the retroactive-flag pass exceeds both thresholds:

• Total emergent: 22
• Total flagged pending_substrate_reconfirmation: true: 19
• Fraction flagged: 86% (threshold: >20%)
• Absolute flag count: 19 (threshold: >10)

Substrates driving the flags (multi-flag substrates first):

• SD-017 (sleep infrastructure, provisional): 3 flags — INV-045, INV-046, INV-050
• SD-012 (homeostatic drive, candidate): 3 flags — INV-041, INV-052, INV-053
• SD-014 (hippocampal 4-component valence, candidate): 3 flags — INV-054, INV-056, INV-065
• ARC-019 (staged developmental curriculum, provisional): 3 flags — INV-043, INV-055, INV-056
• ARC-035 (vmPFC stored->active compiler, candidate): 2 flags — INV-037, INV-038
• ARC-041 (E1 cue-indexed retrieval, candidate): 2 flags — INV-040, INV-041
• Single-flag substrates: SD-011, SD-015, SD-026, ARC-016, ARC-030, ARC-038, ARC-039, ARC-040, ARC-042, ARC-046, ARC-049 (one flag each).

Auditor recommendation: proceed. The flag rate is high because most substrates that carry emergent invariants are still candidate or provisional — the substrate layer is under active construction (SD-011 through SD-028 registered in the last six weeks, ARC-030 through ARC-057 predominantly candidate). The flag is doing its job: it marks that the invariants’ subjects rest on machinery whose design is not yet governance-frozen. Reading 86% flagged as “rule misfire” would require a substrate layer that is already mostly active, which it is not.

If the rule did misfire, the fix would not be to relax the threshold — it would be to promote substrates. That is out of scope for Session B.

Session B stops before committing claims.yaml pending user confirmation. The registry (docs/governance/invariant_classification_audit.md) and the applied-but-unstaged edits to claims.yaml are ready for review.

---

Summary

• Clear universal: 41
• Clear emergent: 27
• Grey zone: 0
• Reclassified (out of claim_type: invariant): 4
• Sum: 72

Retroactive flag count (pending_substrate_reconfirmation: true): 32 Checkpoint threshold triggered: yes

Session C batch complete (2026-04-17):

• Group 1 (INV-047/048/061/062): reclassified to claim_type: derived_prediction after schema extension.
• Group 2 (INV-004/005/006): resolved universal (no re-wording); mechanism- agnosticism notes in docs/invariants.md already support the universal reading.
• Group 3 (INV-019/020/021): INV-019 and INV-020 resolved universal (information- flow principles stateable without REE substrate); INV-021 resolved emergent on [ARC-003, SD-026] paralleling INV-012’s emergent-on-SD-026 reasoning.
• Group 4 (INV-024/051/063/064): INV-024 resolved emergent on [ARC-020, SD-026] (conjunctive offline-write-boundary + commitment); INV-051 and INV-063 resolved emergent on SD-017 (sleep-cluster parallel); INV-064 resolved emergent on [ARC-001, ARC-002, ARC-003, ARC-019] (engines + curriculum).

All grey-zone entries from Session B are now resolved. Registry has 0 grey_zone entries; the audit gate condition for flipping scripts/validate_claims.py to --strict mode is satisfied (per REE_assembly/CLAUDE.md “Invariant Types” section). Flip is a governance decision, not automatic.

---

Clear Universal

These invariants can be stated and defended using information-theoretic, thermodynamic, decision-theoretic, or purely philosophical/logical vocabulary. They would survive a substantially different E1/E2/E3 substrate.

INV-001

• Title: No explicit ethics module or moral scoring layer.
• Rationale: Decision-theoretic design stance (ethics-from-constraint rather than optimisation of a scalar moral objective); holds across any architecture that respects the constraint, not a claim about specific REE structures.

INV-002

• Title: Coherence includes temporal/phase binding, not just static metrics.
• Rationale: Definitional / dynamical-systems claim about what coherence must include; independent of substrate choice.

INV-003

• Title: Language emerges as functional self-representation, not a bolt-on.
• Rationale: Functional-role claim about language stateable in any cognitive architecture with representations and agents.

INV-004

• Title: Post-commit consequence traces are persistent, not resettable.
• Resolved: 2026-04-17 in Session C batch (Group 2).
• Outcome: universal (no re-wording).
• Rationale: Decision-theoretic assertion of action-effect irreversibility. Reference doc (docs/invariants.md#inv-004) already carries an explicit mechanism- agnosticism note; “post-commit consequence traces” is implementation-level vocabulary, not substrate-dependent content.

INV-005

• Title: Harm to others contributes via mirror modelling, not symbolic rules.
• Resolved: 2026-04-17 in Session C batch (Group 2).
• Outcome: universal (no re-wording).
• Rationale: The load-bearing claim is the simulation-vs-symbolic-rules distinction as the ethical mechanism for other-harm contribution – substrate-agnostic. “Mirror modelling” is used in the general cognitive-science sense; ARC-010 grounds the invariant at the REE-specific implementation level but does not underwrite its subject.

INV-006

• Title: Post-commit consequence traces cannot be erased, only integrated.
• Resolved: 2026-04-17 in Session C batch (Group 2).
• Outcome: universal (no re-wording).
• Rationale: Decision-theoretic non-erasability property of past committed-action effects in any agent-world with persistent state. Reference doc carries a mechanism- agnosticism note paralleling INV-004’s. The integration-vs-erasure dichotomy does not require any specific offline-consolidation substrate to be stateable.

INV-007

• Title: Language cannot override embodied harm sensing.
• Rationale: Safety-ordering claim: symbolic/deliberate channels must not override aversive affective signal. Holds in any architecture with those channel types, regardless of how harm sensing is physically realised.

INV-008

• Title: Precision is routed and depth-specific, not global.
• Rationale: Generic predictive-processing / attention-as-precision claim; substrate-agnostic.

INV-009

• Title: Attention is precision modulation, not symbolic control.
• Rationale: Predictive-processing / information-theoretic claim about what attention is; substrate-independent.

INV-010

• Title: Offline integration exists and is required.
• Rationale: General computational-necessity claim (further generalised by INV-049); independent of how offline phases are physically realised.

INV-011

• Title: Imagination must be possible without belief update.
• Rationale: Decision-theoretic separability of counterfactual simulation from posterior update; holds in any agent that both simulates and updates.

INV-013

• Title: Cognition is predictive, iterative, and multi-timescale.
• Rationale: General dynamical/predictive-processing claim about cognition; does not pick out any REE-specific structure.

INV-014

• Title: Representation and regulation are strictly separated.
• Rationale: Architectural-category claim (representation vs regulation channels); would hold under any substrate that distinguishes those channels.

INV-015

• Title: Ethics emerges from constraint, not optimisation.
• Rationale: Decision-theoretic stance about the origin of ethical behaviour; substrate-agnostic.

INV-016

• Title: Stability is prioritized over maximal performance.
• Rationale: Engineering/design priority claim; stateable for any optimising agent.

INV-017

• Title: Runaway behavior is a control failure, not representational.
• Rationale: Classification of failure-mode in dynamical/control-theoretic terms; substrate-independent.

INV-018

• Title: Agency is required; passive predictors are not REE.
• Rationale: Definitional boundary claim about what counts as agentive; holds across substrates.

INV-019

• Title: Rehearsal traversal and irreversible durable write must remain separated.
• Resolved: 2026-04-17 in Session C batch (Group 3).
• Outcome: universal (no re-wording).
• Rationale: Direct consequence of INV-011 extended to any channel that both explores and writes durably; substrate-neutral information-flow constraint. ARC-014 (Default Mode) is the REE-specific implementation locus but does not underwrite the subject.

INV-020

• Title: Constraint stores are authority-stratified from direct observational/symbolic writes.
• Resolved: 2026-04-17 in Session C batch (Group 3).
• Outcome: universal (no re-wording).
• Rationale: Substrate-neutral information-flow constraint: any architecture with multiple write channels requires typed authority separation. ARC-020 is the REE-specific implementation in the offline-consolidation pipeline but does not underwrite the subject.

INV-022

• Title: Trust/precision allocation must remain heterogeneous, not a single scalar.
• Rationale: Information-theoretic / predictive-processing claim; applies to any hierarchical precision-weighted architecture.

INV-023

• Title: Protected offline recalibration/integration regimes are structurally required.
• Rationale: Subsumed by INV-049’s general computational-necessity argument for offline phases; substrate-independent.

INV-025

• Title: Irreducible uncertainty is an epistemological constraint that cannot be engineered away.
• Rationale: Information-theoretic / epistemological claim; stateable without any substrate commitment.

INV-026

• Title: A self is an operational prerequisite for responsible agency.
• Rationale: Philosophical/decision-theoretic claim; the “self” here is an operational role, not a specific REE representation.

INV-027

• Title: An external world is a structural prerequisite for grounded representation.
• Rationale: Grounded-cognition claim; substrate-independent.

INV-028

• Title: Shared-world ethics requires modelling others as co-inhabitants of the same consequence space.
• Rationale: Ethical/decision-theoretic claim about what shared-world ethics requires; holds for any agent architecture capable of modelling others.

INV-029

• Title: Love as long-horizon care-investment is a structurally coherent agent disposition.
• Rationale: Decision-theoretic/philosophical claim about the coherence of the disposition under any sufficiently expressive agent architecture.

INV-030

• Title: Viability is defined relative to the agent’s continuity in a shared world, not as a scalar reward.
• Rationale: Decision-theoretic design stance on what viability means; rejects the scalar-reward formulation independent of substrate.

INV-031

• Title: Functional persistence in a shared world is the operational goal of cognition.
• Rationale: Decision-theoretic goal-formulation claim; independent of substrate.

INV-032

• Title: Moral agency requires both approach and avoidance drives.
• Rationale: Decision-theoretic/ethical claim about necessary-condition structure of moral agency; universal even though REE instantiates it via ARC-030.

INV-042

• Title: The five axioms jointly derive ethical objectives.
• Rationale: Axiomatic/logical derivation claim from philosophical axioms; no substrate required to state.

INV-044

• Title: Approximate Bayesian contextual inference is architecturally impossible to co-compute with online encoding.
• Rationale: Information-theoretic impossibility argument; substrate-independent.

INV-049

• Title: Any agent that builds a model of the world and acts in it requires periodic offline phases.
• Rationale: Explicit universal: the claim asserts “this is not a biological contingency but a general computational necessity for model-building agents.”

INV-057

• Title: Cross-species signal legibility evidences functional specificity.
• Rationale: Evolutionary/communication-theoretic claim about signal structure; does not depend on REE substrate.

INV-058

• Title: Play is a structurally necessary behavioral mode.
• Rationale: Decision-theoretic / learning-theoretic claim about the role of bounded low-stakes contexts in skill acquisition.

INV-059

• Title: Mutual frame-maintenance signaling is necessary for play.
• Rationale: Game-theoretic claim about frame negotiation; does not require REE-specific machinery to state.

INV-067

• Title: Verisimilitude definition.
• Rationale: Definitional claim in predictive-processing vocabulary.

INV-068

• Title: Temporal depth D_V(t).
• Rationale: Definitional extension of INV-067; substrate-independent.

INV-069

• Title: Self as dynamically sustained verisimilitude-maintaining process.
• Rationale: Process-philosophical definition of self in terms of INV-067/068; general.

INV-070

• Title: Epistemic responsibility.
• Rationale: Ethical/decision-theoretic claim; substrate-independent.

INV-071

• Title: Language as similarity repair.
• Rationale: Functional-role claim about language in inter-agent coordination; substrate-agnostic.

INV-072

• Title: Violence conditions corollary.
• Rationale: Ethical/logical claim derived from INV-028/029/070/071; no substrate required.

---

Clear Emergent

Each entry’s subject matter becomes ill-defined if the listed substrate is retracted. Every emergent_from entry is already present in the invariant’s depends_on.

INV-012

• Title: Responsibility arises through commitment, not prediction alone.
• emergent_from: [SD-026]
• Rationale: “Commitment” as load-bearing ethical concept requires a mechanism by which endogenous goal-derived templates can write precision biases before a decision (SD-026); without that, commitment collapses into “whatever was already conditioned.” (Already classified in Session A — unchanged.)
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes (already set in Session A)

INV-021

• Title: Responsibility-bearing durable updates occur only at typed commit boundaries.
• emergent_from: [ARC-003, SD-026]
• Resolved: 2026-04-17 in Session C batch (Group 3).
• Outcome: emergent on [ARC-003, SD-026].
• Rationale: Structural generalisation of INV-012 (responsibility via commitment). “Responsibility-bearing” imports INV-012’s SD-026 dependency; “typed commit boundary” requires ARC-003 (E3’s commit action) to be a real architectural feature. ARC-003 added to depends_on; SD-026 added to depends_on.
• Substrate status at audit: ARC-003 active; SD-026 candidate.
• pending_substrate_reconfirmation applied: yes (SD-026 candidate).

INV-033

• Title: REE agents require second-order epistemic access to their own model confidence, wired into commit gating.
• emergent_from: [ARC-016]
• Rationale: “Wired into commit gating” and “structurally represented model confidence” are concepts that only have well-formed referents under ARC-016 (modes as control-plane regimes with routed precision). Without a mode/precision regime substrate, there is no commit-gating locus for second-order access to plug into.
• Substrate status at audit: provisional
• pending_substrate_reconfirmation applied: yes

INV-034

• Title: Goal maintenance is a necessary co-condition for ethical agency.
• emergent_from: [ARC-030]
• Rationale: Pairs with INV-032 but asserts the need for a prospective goal-maintenance drive that is symmetrically Go-side (not merely NoGo/harm-avoidance). The subject — “prospective goal-directed motivation” as distinct from harm-avoidance — requires the symmetric Go/NoGo substrate (ARC-030). In a pure-avoidance architecture the invariant is un-stateable, not merely unevidenced.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-035

• Title: A REE state must not be defined purely by sensory appearance.
• emergent_from: [ARC-003, ARC-004, ARC-007]
• Rationale: “REE state” is explicit in the title — the claim is definitional for state representation given commitment (ARC-003), multi-timescale latent stack (ARC-004), and hippocampal paths through residue-field terrain (ARC-007). Retract any of these and the distinguishing dimensions named (temporal position, active commitment, goal/antigoal, operative constraints) have no structural home.
• Substrate status at audit: all three active
• pending_substrate_reconfirmation applied: no

INV-036

• Title: A REE state is valid only if it supports transition prediction, valence/antigoal tagging, and uncertainty representation.
• emergent_from: [ARC-003, ARC-007, ARC-018]
• Rationale: Three validity-conditions on REE state objects; the state object itself is the ARC-003/007 construct, and “transition prediction supports trajectory evaluation” presumes ARC-018 (explicit rollouts and post-commitment viability mapping).
• Substrate status at audit: all three active
• pending_substrate_reconfirmation applied: no

INV-037

• Title: Stored-vs-active distinction requires a preparation substrate (vmPFC-analog).
• emergent_from: [ARC-035]
• Rationale: The stored/active distinction as an architectural claim is defined by the existence of a compiler (ARC-035 vmPFC) that converts stored content into live state components. Remove ARC-035 and the claim has no referent.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-038

• Title: EVR pattern: post-hoc ethical scoring without active constraint preparation produces unconstrained trajectory generation.
• emergent_from: [ARC-035]
• Rationale: The EVR pattern is defined as the failure mode of a system with post-hoc scoring but no ARC-035 preparation substrate. Without ARC-035 in the architectural vocabulary, “absence-of-active-constraint-preparation” is not a specifiable condition.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-039

• Title: Schema-primed rapid consolidation.
• emergent_from: [ARC-007, ARC-038]
• Rationale: “Stable prior map” and “mature coherent residue field” are ARC-007 constructs, and the map-stability-gated consolidation rate is a claim about ARC-038’s replay-switching regime. Without hippocampal terrain maps and switchable replay, neither side of the conditional is well-formed.
• Substrate status at audit: ARC-007 active; ARC-038 candidate
• pending_substrate_reconfirmation applied: yes

INV-040

• Title: A minimal z_world cue pattern is sufficient to activate the appropriate harm/goal terrain precision configuration.
• emergent_from: [SD-005, ARC-041]
• Rationale: Title explicitly names z_world (SD-005 self/world split) and the E1 cue-indexed retrieval circuit (ARC-041). Retract either and the claim’s subject (“minimal z_world cue pattern triggers terrain preparation”) has no object.
• Substrate status at audit: SD-005 implemented; ARC-041 candidate
• pending_substrate_reconfirmation applied: yes

INV-041

• Title: Childhood phase is architecturally prerequisite for E3 ethical selection.
• emergent_from: [ARC-042, SD-012, SD-016]
• Rationale: The claim is that a specific developmental regime populates ARC-041’s ContextMemory (SD-016) and exercises the homeostatic drive (SD-012) so that ARC-042’s pre-initialised ethical selection machinery operates on real signal. Without these substrates the notion of “architectural prerequisite” is un-stateable.
• Substrate status at audit: ARC-042 candidate; SD-012 candidate; SD-016 implemented
• pending_substrate_reconfirmation applied: yes

INV-043

• Title: REE architecture enables but does not guarantee ethical development; caregiver love is required.
• emergent_from: [ARC-019, ARC-040]
• Rationale: The claim picks out “developmental phase in which the agent experiences love from caregivers” — which presumes ARC-040 (caregiver substrate) and ARC-019 (staged developmental curriculum). Without a developmental-phase architecture, the conditional has no antecedent.
• Substrate status at audit: ARC-019 provisional; ARC-040 candidate
• pending_substrate_reconfirmation applied: yes

INV-045

• Title: NREM-then-REM sleep phase sequence is the correct computational order.
• emergent_from: [SD-017]
• Rationale: The claim is specifically about the ordering within the SD-017 sleep-phase infrastructure. Without SD-017 the NREM/REM labels and the phase-dependency chain have no substrate referent. (INV-049 states the universal version; INV-045 is the substrate-specific ordering claim.)
• Substrate status at audit: provisional
• pending_substrate_reconfirmation applied: yes

INV-046

• Title: Dementia is computationally a failure of the contextual attribution pipeline.
• emergent_from: [SD-017]
• Rationale: “Contextual attribution pipeline” in this invariant means the pipeline maintained by SD-017’s sleep phases. Without SD-017 the attribution-pipeline-failure reframing is not stateable in those specific terms.
• Substrate status at audit: provisional
• pending_substrate_reconfirmation applied: yes

INV-050

• Title: Sleep phase architecture is regulated by three drives (circadian, homeostatic, MEL).
• emergent_from: [SD-017]
• Rationale: The three-drive claim is a claim about the regulation of SD-017’s sleep phase architecture. The drives exist abstractly, but the invariant’s subject (sleep phase architecture regulation) is SD-017-resident.
• Substrate status at audit: provisional
• pending_substrate_reconfirmation applied: yes

INV-052

• Title: Goal-directed behaviour in adversive environments requires a tonic regulatory system.
• emergent_from: [SD-012]
• Rationale: “Tonic regulatory system maintaining benefit orientation” is an SD-012 concept (homeostatic drive modulation for z_goal seeding). Without SD-012 the claim’s subject is un-stateable.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-053

• Title: Chronic harm-dominated environments produce a self-maintaining depressive attractor.
• emergent_from: [SD-011, SD-012]
• Rationale: “z_goal seeding failure”, “terrain collapse”, “HABIT/PLANNED behavioural equivalence” are all SD-012/SD-011-resident concepts. Remove either substrate and the attractor description has no object.
• Substrate status at audit: both candidate
• pending_substrate_reconfirmation applied: yes

INV-054

• Title: Benefit terrain collapse creates a closed negative feedback loop predicting chronicity.
• emergent_from: [SD-014]
• Rationale: “Benefit terrain” names the hippocampal 4-component valence vector (SD-014). The collapse-and-feedback-loop claim is a claim about that specific terrain representation.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-055

• Title: A pre-childhood infant stage is developmentally necessary.
• emergent_from: [ARC-019, ARC-046]
• Rationale: “Infant stage” in the ARC-019 curriculum sense, requiring ARC-046’s hazard protection for sensorially broad exploration. Without those substrates the phase the invariant picks out does not exist architecturally.
• Substrate status at audit: ARC-019 provisional; ARC-046 candidate
• pending_substrate_reconfirmation applied: yes

INV-056

• Title: Developmental hardening should be substrate-specific: selective neoteny.
• emergent_from: [ARC-010, ARC-019, SD-014]
• Rationale: Names specific subsystems (social cognition = ARC-010; developmental staging = ARC-019; valence/goal = SD-014) that retain elevated plasticity. The substrate-specific hardening claim has no partition to refer to without these.
• Substrate status at audit: ARC-010 active; ARC-019 provisional; SD-014 candidate
• pending_substrate_reconfirmation applied: yes

INV-060

• Title: Play dominates child developmental phase; play type progresses as subsystem competence develops.
• emergent_from: [ARC-049]
• Rationale: “Play” in the ARC-049 sense (co-maintained context tag held by agent and environment). The progression-with-competence claim is about play episodes as defined by that substrate.
• Substrate status at audit: candidate
• pending_substrate_reconfirmation applied: yes

INV-065

• Title: Proxy goal necessity for multi-step planning.
• emergent_from: [SD-014, SD-015]
• Rationale: “Proxy goal” presumes a goal/resource representation (SD-015 z_resource encoder) and a valence vector (SD-014). Without these the proxy-goal construct is un-stateable.
• Substrate status at audit: both candidate
• pending_substrate_reconfirmation applied: yes

INV-066

• Title: z_self subtraction must precede coherency-based control plane signals.
• emergent_from: [SD-005]
• Rationale: z_self as a separable quantity exists only under SD-005’s self/world split. Without SD-005 the subtraction operation is un-stateable (no z_self to subtract).
• Substrate status at audit: implemented (treated as active-equivalent)
• pending_substrate_reconfirmation applied: no

INV-024

• Title: Offline consolidation and online commitment must remain isolated at responsibility-bearing write loci.
• Resolved: 2026-04-17 in Session C batch (Group 4).
• emergent_from: [ARC-020, SD-026]
• Rationale: Conjunctive emergence. ARC-020 (typed offline write boundaries) underwrites the “offline consolidation write locus” as a typed object; SD-026 (prospective precision template write channel) underwrites the “responsibility-bearing” qualifier by giving INV-012/INV-021’s commitment concept its subject matter. Retract either and the invariant’s subject becomes ill-defined.
• Substrate status at audit: ARC-020 candidate, SD-026 candidate.
• pending_substrate_reconfirmation applied: yes.

INV-051

• Title: Optimal range of daily Model Error Load (MEL).
• Resolved: 2026-04-17 in Session C batch (Group 4).
• emergent_from: [SD-017]
• Rationale: Parallels the sleep-cluster decision on INV-050. Both upper-bound failure (“overnight update capacity exceeded”) and lower-bound failure (“third learning drive unstimulated absent offline phases”) are stated in SD-017 vocabulary; the MEL-range concept is not stateable without a substrate that defines offline update capacity. Added SD-017 to depends_on.
• Substrate status at audit: SD-017 provisional.
• pending_substrate_reconfirmation applied: yes.

INV-063

• Title: Minimum entropy intake sleep dependency.
• Resolved: 2026-04-17 in Session C batch (Group 4).
• emergent_from: [SD-017]
• Rationale: Parallels the sleep-cluster decisions on INV-045/046/050/062. The “entropy intake” concept is defined relative to the four sleep functions (schema consolidation, contrastive replay, E1 world-model updating, E2 motor-sequence learning); without SD-017 those functions have no substrate and the entropy-dependency claim has no subject. Added SD-017 to depends_on.
• Substrate status at audit: SD-017 provisional.
• pending_substrate_reconfirmation applied: yes.

INV-064

• Title: E1/E2/E3 maturational sequence necessity.
• Resolved: 2026-04-17 in Session C batch (Group 4).
• emergent_from: [ARC-001, ARC-002, ARC-003, ARC-019]
• Rationale: The title explicitly names the three engines — retract any of ARC-001/002/003 and the corresponding term in the maturational sequence has no referent. ARC-019 (staged developmental curriculum) supplies the “sequence” concept as a curriculum constraint. All four chosen together because the claim conjoins both the ordered objects and the ordering itself. Added all four to depends_on.
• Substrate status at audit: ARC-001/002/003 active, ARC-019 provisional.
• pending_substrate_reconfirmation applied: yes (propagates from lowest-status substrate per governance rule).

---

Grey Zone

These entries could plausibly resolve as universal (by re-statement) or as emergent on a specific substrate, but the audit is not confident enough to force a verdict in one session. Each entry carries a self-contained follow-up prompt for a Session C_n auditor.

---

Reclassified (out of claim_type: invariant)

These entries were reclassified to claim_type: derived_prediction (schema extension, 2026-04-17). They are well-formed predictions that hold only conditional on REE correctly modelling their target domain (human clinical neurology, pharmacology, psychiatry, or phenomenology), and are therefore not architectural invariants of the REE framework itself. Per the new schema, derived_prediction entries carry no invariant_type or emergent_from field. The ‘INV-‘ ID prefix is retained for provenance only.

See docs/architecture/invariant_types.md for the derived_prediction schema definition.

INV-047

• Resolved: 2026-04-17 in Session C batch (Group 1).
• Outcome: reclassified to claim_type: derived_prediction, prediction_domain: clinical_neurology.
• Rationale: Clinical prediction about the staged cognitive decline profile of human Alzheimer’s disease; holds only conditional on REE correctly modelling human cognition (Daniel’s 2026-04-17 objection).

INV-048

• Resolved: 2026-04-17 in Session C batch (Group 1).
• Outcome: reclassified to claim_type: derived_prediction, prediction_domain: clinical_pharmacology.
• Rationale: Clinical pharmacological prediction about human sleep disruption equivalence; holds only conditional on REE correctly modelling human sleep and consolidation.

INV-061

• Resolved: 2026-04-17 in Session C batch (Group 1).
• Outcome: reclassified to claim_type: derived_prediction, prediction_domain: clinical_psychiatry.
• Rationale: Clinical psychiatric prediction about a unified developmental etiology for human frame-confusion disorders; holds only conditional on REE correctly modelling human developmental frame-distinction.

INV-062

• Resolved: 2026-04-17 in Session C batch (Group 1).
• Outcome: reclassified to claim_type: derived_prediction, prediction_domain: dream_phenomenology.
• Rationale: Phenomenological prediction about the structure of human dream experience; holds only conditional on REE correctly modelling human offline cognition.