REE Claim Index

This index lists all claim IDs with one-line summaries and links. For full metadata, see claims.yaml. Dependency edges are conceptual prerequisites and must stay acyclic; runtime feedback loops are documented in architecture flow specs rather than depends_on.

Claim Type: implementation_note
Scope: Claim index and navigation
Depends On: None
Status: stable
Claim ID: IMPL-018

Invariants (INV)

  • INV-001 No explicit ethics module or moral scoring layer.
  • INV-002 Coherence includes temporal/phase binding, not just static metrics.
  • INV-003 Language emerges as functional self-representation, not a bolt-on.
  • INV-004 Post-commit consequence traces are persistent, not resettable.
  • INV-005 Harm to others contributes via mirror modelling, not symbolic rules.
  • INV-006 Post-commit consequence traces cannot be erased, only integrated.
  • INV-007 Language cannot override embodied harm sensing.
  • INV-008 Precision is routed and depth-specific, not global.
  • INV-009 Attention is precision modulation, not symbolic control.
  • INV-010 Offline integration exists and is required.
  • INV-011 Imagination must be possible without belief update.
  • INV-012 Responsibility arises through commitment, not prediction alone.
  • INV-013 Cognition is predictive, iterative, and multi-timescale.
  • INV-014 Representation and regulation are strictly separated.
  • INV-015 Ethics emerges from constraint, not optimisation.
  • INV-016 Stability is prioritized over maximal performance.
  • INV-017 Runaway behavior is a control failure, not representational.
  • INV-018 Agency is required; passive predictors are not REE.
  • INV-019 Rehearsal traversal and irreversible durable write must remain separated.
  • INV-020 Constraint stores are authority-stratified from direct observational/symbolic writes.
  • INV-021 Responsibility-bearing durable updates occur only at typed commit boundaries.
  • INV-022 Trust/precision allocation must remain heterogeneous, not a single scalar.
  • INV-023 Protected offline recalibration/integration regimes are structurally required.
  • INV-024 Offline consolidation and online commitment must remain isolated at responsibility-bearing write loci.

Architectural Commitments (ARC)

  • ARC-001 E1 is the persistent predictive substrate.
  • ARC-002 E2 is the fast forward predictor of affordances.
  • ARC-003 E3 selects and commits trajectories.
  • ARC-004 L-space is a multi-timescale latent stack.
  • ARC-005 Control plane routes precision and modes.
  • ARC-006 Entities are sparse, persistent, bindable structures.
  • ARC-007 Hippocampal systems store and replay paths.
  • ARC-008 Commitment eligibility is gated by tau, rho, and phi.
  • ARC-009 Language is a symbolic mediation and coordination layer.
  • ARC-010 Social cognition uses mirror modelling and coupling.
  • ARC-011 Offline integration (sleep) is required for stability.
  • ARC-012 E3 does not require an explicit ethical cost term.
  • ARC-013 Residue is persistent latent-space curvature; hippocampal paths form a cognitive map.
  • ARC-014 Default Mode enables safe imagination without commitment.
  • ARC-015 Self-impact attribution and responsibility flow are required.
  • ARC-016 Modes are control-plane regimes applied to shared predictive machinery.
  • ARC-017 Minimal stream tags with typed exteroception and explicit reality-coherence lane (world, homeostasis, harm, self-sensory, precision, temporal/reality coherence, valence, action, self-impact).
  • ARC-018 Hippocampus generates explicit rollouts and post-commitment viability mapping.
  • ARC-019 REE requires staged developmental training with explicit curriculum gates.
  • ARC-020 Offline consolidation is protected by typed authority/write boundaries and cannot bypass commit-path governance.
  • ARC-021 Three BG-like cortico-striatal loops require distinct learning channels.
  • ARC-022 ⚠ Hierarchical effect-object abstraction pipeline (E1→E2→DMN-regime→Goal/Avoid→Hippocampus). Conflict with ARC-007 and ARC-014 — pending adjudication (Q-020).

Mechanism Hypotheses (MECH)

  • MECH-001 Astrocyte-aware regulatory stack mediates control-plane precision routing.
  • MECH-002 Precision control analogues shape cognitive regimes.
  • MECH-003 Precision must be tau-scoped with lossy projections.
  • MECH-004 Signal-to-knob wiring map for control plane.
  • MECH-005 Path authority and interruptibility via norepinephrine-like control.
  • MECH-006 Serotonin-like modulation governs representational collapse.
  • MECH-007 Attention must be fragmented across control axes.
  • MECH-008 Legacy: superseded by AA/PCM + control-channel mode framing (MECH-046/047/048, MECH-039).
  • MECH-010 Language emergence and bootstrapping dynamics.
  • MECH-011 Language and learning dynamics.
  • MECH-012 Language and institutions interplay.
  • MECH-013 Language failure modes and pathologies.
  • MECH-014 Minimal signalling channel requirements.
  • MECH-015 Trust and deception dynamics.
  • MECH-016 Precision recalibration during sleep.
  • MECH-017 Reality consolidation during sleep.
  • MECH-018 Residue integration during sleep.
  • MECH-019 Control plane shapes modes of cognition, not discrete choices.
  • MECH-020 Legacy: superseded by control-plane mode framing (MECH-019, MECH-039).
  • MECH-021 Subjective “now” is a control surface across horizons.
  • MECH-022 Hippocampal systems inject hypotheses gated by control plane.
  • MECH-023 Responsibility is geometric and path-dependent.
  • MECH-024 Selfhood, personality, and ethics converge structurally.
  • MECH-025 Action mode prioritizes short-horizon, high-precision, responsibility-linked control.
  • MECH-026 Ready vigilance primes restraint under high sensitivity without action.
  • MECH-027 Pathological modes reflect mis-tuned control-plane regimes.
  • MECH-028 Ethical behavior depends on mode transitions and learning preservation.
  • MECH-029 Reflective/DMN mode supports moral simulation via gated replay.
  • MECH-030 Sleep modes consolidate learning and ethical residue across regimes.
  • MECH-031 Derived social tags and empathy coupling via control-plane knobs.
  • MECH-032 OTHER_SELFLIKE detection is biased toward high recall to avoid empathy false negatives.
  • MECH-033 E2 forward-prediction kernels seed hippocampal rollouts.
  • MECH-034 Viability mapping updates are distinct from residue curvature updates.
  • MECH-035 VALENCE is vector-valued and ranked without scalar collapse.
  • MECH-036 Other-harm triggers veto only under high-certainty catastrophic outcomes.
  • MECH-037 Papez-like loop provides provenance gating / reality filtering.
  • MECH-038 Arcuate-like sequence-to-motor channel nudges language emergence.
  • MECH-039 Modes are stable regions in control-channel space, not separate modules.
  • MECH-040 Safety baseline and volatility are distinct control channels for arousal/readiness.
  • MECH-041 Affective expression broadcasts control‑plane regime to reduce social prediction load.
  • MECH-042 Telemetry exposure channels report internal control‑plane state for diagnostics.
  • MECH-043 Dopamine‑like modulation of precision‑weighting for unsigned prediction errors.
  • MECH-044 Hippocampal systems participate in relational binding and comparison.
  • MECH-045 Object‑file‑like buffers provide minimal entity persistence across time.
  • MECH-046 Amygdala analogue updates control‑plane mode priors from fast salience classification.
  • MECH-047 Pre‑commitment mode manager commits with hysteresis and switching costs.
  • MECH-048 μ/κ stability overlays modulate mode entropy and switching pressure.
  • MECH-049 Temporal phase compartmentalisation preserves ethical constraint independence.
  • MECH-050 Functional locality supports attribution without requiring anatomical columns.
  • MECH-051 Oxytocin/vasopressin analogues modulate relational topology and mode priors.
  • MECH-052 Prolactin analogue stabilises care‑investment persistence.
  • MECH-053 Habenula‑like aversive PE gate suppresses commitment under negative spikes.
  • MECH-054 Signed harm/benefit prediction‑error precision channels remain distinct.
  • MECH-055 Affective channel separation keeps hedonic tone, valence, and signed PE distinct.
  • MECH-056 Residue should constrain trajectories before distorting core representations.
  • MECH-057a Action-loop completion gate: committed action sequences suppress new precision updates until execution completes. (V2-testable with multi-step substrate; V1/V2 FAILs are substrate-limited)
  • MECH-057b Thought-loop trajectory promotion gate: hippocampal sequence completion must be verified before candidates are eligible for E3 selection. (V3-scoped — requires HippocampalModule promotion policy not present in V2)
  • MECH-058 Slow target-anchor dynamics stabilize JEPA-like-referenced E1/E2 representation via functional rate separation.
  • MECH-059 Confidence channel (uncertainty-derived precision) must remain distinct from residual error for precision routing.
  • MECH-060 Umbrella commit-boundary separation principle with explicit phase control-variation map: pre-commit simulation and post-commit realized-error channels stay responsibility-distinct via write-boundary enforcement (MECH-061/062/066/067).
  • MECH-061 Commit-boundary token reclassifies pre-commit vs post-commit error routing.
  • MECH-062 E3 should use coordinated tri-loop gating (motor, cognitive-set, motivational) as the pre-commit eligibility realization layer for MECH-060.
  • MECH-063 Control plane should retain orthogonal tonic/phasic axes rather than collapsing into one scalar regulation signal.
  • MECH-064 Typed authority/control-store separation blocks direct exteroceptive writes into policy, identity, and capability stores.
  • MECH-065 Reality-coherence conflict lane should modulate loop precision and commitment thresholds before execution lock-in.
  • MECH-066 Pre-commit and post-commit channels may share representations but must stay strictly separated at durable write boundaries.
  • MECH-067 A machine-checkable phase/store/actor permission matrix is required to enforce commit-boundary write rules.
  • MECH-068 Consolidation selectivity: E3 selectivity lives in the consolidation operator, not in E1 feature basis.
  • MECH-069 Sensory prediction error, motor-sensory error, and harm/goal error are incommensurable.
  • MECH-070 E2 is a conceptual-sensorium motor model with planning horizon exceeding E1.
  • MECH-071 E2 harm prediction is better calibrated for agent-caused vs environment-caused transitions.
  • MECH-072 Foreseeable-harm gating on residue accumulation reduces false attribution without degrading harm avoidance.
  • MECH-073 ⚠ Valence is intrinsic to hippocampal map geometry, not applied downstream. Conflict with ARC-007 — pending adjudication (Q-020).
  • MECH-074 Amygdala as read/write head for valenced hippocampal map: encoding modulation, retrieval bias, pre-cortical priming, triggered remapping.
  • MECH-075 Basal ganglia perform dopaminergic gain/threshold setting on hippocampal attractor dynamics, not external value comparison.
  • MECH-076 Residue is structural deformation of hippocampal map topology (CA3 dendritic retraction, DG neurogenesis, palimpsest reconsolidation).
  • MECH-077 Therapeutic change = hippocampal remapping; modalities operate at different levels (CBT/OFC, EMDR/reconsolidation, psychodynamic/relational, psychedelic/attractor-flattening).
  • MECH-078 Amygdala bootstraps novel valence for unmapped hippocampal territory; anxiety disorders = systematic over-valencing of novel regions.
  • MECH-079 Phenomenological continuous selfhood is an artefact of stable hippocampal map geometry; identity change = map phase transition.
  • MECH-080 Rollout truncation set-points as psychiatric individual differences substrate: ADHD = truncation-biased; anxiety = extension-biased; OCD = attractor lock-in.
  • MECH-081 E2 sufficiency constraint reduces E1 effective dimensionality target: E2’s prediction task partially specifies E1’s feature-discovery search from above.
  • MECH-082 Top-down perceptual bias: hippocampal map distortion (MECH-076) propagates through E2 to bias E1 attentional sampling without explicit E1 retraining.
  • MECH-083 Acetylcholine as meta-level plasticity gain: continuous gain on how strongly bottom-up input modifies the hippocampal map (vs. retrieval from existing attractors); governs durable-write vs. read-through; distinct from NA’s attentional snap.
  • MECH-084 Noradrenaline as attentional snap and E1/E2 sampling ratio modulator: fast moment-to-moment orienting to bottom-up surprise; determines whether system orients now (NA) vs. whether that orientation leaves a trace (ACh); extends MECH-005.
  • MECH-085 ⚠ Serotonin as hippocampal map geometry parameter (rollout depth, distal value weighting, aversive basin depth). Conflict with MECH-006 (temporal-depth prohibition) — pending Daniel’s adjudication.
  • MECH-086 Dopamine as trajectory selection gain plane: exploitation signal operating downstream of landscape (5-HT), state representation (NA), and encoding gate (ACh); cannot compensate upstream failures.
  • MECH-087 Hierarchical ordering of four neuromodulatory control planes: ACh → NA → 5-HT → DA; upstream planes constrain downstream plane effectiveness; intervention order matters.
  • MECH-088 Psychiatric conditions as four-plane control failures: PTSD (5-HT+NA), depression (5-HT+DA), ADHD (NA+DA), OCD (DA+ACh), psychosis (NA+DA); extends MECH-027 and MECH-080.

Open Questions (Q)

  • Q-001 What mechanisms produce entity emergence and binding?
  • Q-002 Appropriate spatial resolution for R(x,t)?
  • Q-003 Should R(x,t) be scalar or vector?
  • Q-004 How to calibrate tau_R relative to E1/E2?
  • Q-005 Can sleep anneal or reset R(x,t)?
  • Q-006 Is ethics developmental rather than additive?
  • Q-007 Do universal expressions map to stable control‑channel regimes?
  • Q-008 Legacy: resolved in favor of valence/μκ separation with calibration follow-up.
  • Q-009 Legacy: resolved via bounded care-override policy with hard catastrophic other-harm veto.
  • Q-010 Legacy: separation question resolved into MECH‑055.
  • Q-011 Legacy: resolved by placing diversity-floor control in pre-commit/offline sampling while preserving residue invariants.
  • Q-012 Can latent predictive world models stay agentically stable without explicit REE-like control constraints?
  • Q-013 Can deterministic JEPA plus derived dispersion match explicit stochastic uncertainty heads for REE precision routing?
  • Q-014 Do JEPA invariances hide ethically relevant distinctions in REE attribution contexts?
  • Q-015 What is the smallest commit-boundary token that still supports reliable multi-timescale attribution?
  • Q-016 What arbitration policy best resolves tri-loop gate conflicts without coupling collapse?
  • Q-017 What is the minimal orthogonal control-axis subset that preserves observed regime separations?
  • Q-018 What RC-conflict threshold/hysteresis calibration blocks authority spoofing without chronic over-suppression?
  • Q-019 (registered 2026-02-27) Three-gate vs single-gate BG architecture: three anatomically distinct loops (sensorium/thought/action-commitment) vs one gating endpoint?
  • Q-020 ⚠ Does ARC-007’s “no value computation” constraint survive MECH-073 (valence intrinsic to hippocampal map)? Pivotal for NC-01 through NC-09 cluster integration. Pending Daniel’s adjudication.

Implementation Notes (IMPL)

  • IMPL-001 Glossary of canonical REE terms.
  • IMPL-002 Repository metadata and contribution process.
  • IMPL-003 Minimum instantiation specification.
  • IMPL-004 Legacy REE overview summary.
  • IMPL-005 Failure mode taxonomy.
  • IMPL-006 Legacy migration mapping.
  • IMPL-007 Legacy refactor final output summary.
  • IMPL-008 Program phases, repository roles, and phase-gate criteria.
  • IMPL-009 Wiring notes and cross-reference summary.
  • IMPL-010 Android world environment contract.
  • IMPL-011 Toy world environment contract.
  • IMPL-012 Toy world scoring functions.
  • IMPL-013 Documentation operating procedure and prompts.
  • IMPL-014 Documentation change history.
  • IMPL-015 Legacy architecture overview.
  • IMPL-016 Trajectory selection detail for E3.
  • IMPL-017 Conflict index and resolution entry point.
  • IMPL-018 Claim index and navigation.
  • IMPL-019 Self-first, social-later developmental testing order heuristic.
  • IMPL-020 Formal JEPA↔REE terminology alignment glossary for interoperability.
  • IMPL-021 Hybrid JEPA↔REE architecture diagram contract and rendering checklist.
  • IMPL-022 JEPA-like E1/E2 representation-reference contract (inputs, outputs, knobs, and failure gates).
  • IMPL-023 REE-v2 representation-interface-first spec and phase gate with hook bridges to commit-boundary, tri-loop gating, and orthogonal control-axis completion.
  • IMPL-024 v2 REE-first canonical wording policy with JEPA interface translation and convergence-owned source-method language.
  • IMPL-025 Cross-version hook surface contract and registry, including v2 bridge families for commit-boundary, tri-loop arbitration, and orthogonal-axis telemetry.

Open Questions

None noted in preserved sources.

  • IMPL-018

References / Source Fragments

  • docs/processed/legacy_tree/docs/claims/claim_index.md

REE is developed by Daniel Golden (Latent Fields). Apache 2.0.