Documentation Refactoring Changelog

Claim Type: implementation_note
Scope: Documentation change history
Depends On: None
Status: stable
Claim ID: IMPL-014

2026-04-03: EXQ-223 PASS — Minimal Mind Hypothesis Confirmed

The milestone

EXQ-223 (Minimal Vertebrate Cognition Ablation) passed 3/3 criteria. The REE core loop — E1 (associative world model) + E2 (fast transition model) + HippocampalModule (trajectory proposal) + multinomial go/no-go selection + raw harm/reward signals — is sufficient to produce stable, harm-avoiding, resource-seeking behaviour on a 10×10 gridworld with environmental drift, without commitment architecture, goal evaluation, or motivational valence circuits.

This is the first experimental confirmation that the E1+E2+hippocampus triad, configured at approximately early-vertebrate complexity, constitutes a minimal functional mind: a system that navigates its world, avoids harm, and acquires resources through gradient following alone.

Criteria

Criterion Result
C1: REE harm ratio < 0.65 PASS — seed 0: 0.293, seed 1: 0.394, seed 2: 0.315
C2: REE reward > 2× random PASS — REE takes ~4.5× as much reward as random
C3: Positive learning signal across seeds PASS — 3/3 seeds

What was ablated

The configuration strips deliberative machinery entirely:

  • commitment_threshold = −1.0: agent is always in the uncommitted state
  • z_goal_enabled = False: no goal representation or goal-directed drive
  • benefit_eval_enabled = False: no prospective benefit evaluation

What remains: the slow LSTM world model (E1), the fast transition predictor (E2), the hippocampal trajectory proposal mechanism, and the primitive go/no-go gate — and nothing else. The result shows this is enough.

The zebrafish question — and why the answer matters

The most natural question to ask after this result is: what living organism looks like this?

The EXQ-223 ablation configuration — E1 (slow associative world model) + E2 (fast transition predictor) + HippocampalModule (trajectory proposal) + go/no-go selection + raw harm/reward signals, with commitment architecture stripped — corresponds in circuit topology to the zebrafish larva at 5–7 days post-fertilisation (dpf). This is not a loose analogy. It is a circuit-level match.

Circuit correspondence: REE minimal configuration vs zebrafish larva (5–7 dpf)

REE component Zebrafish structure Basis for match
E1 — slow LSTM associative world model Dorsal + lateral pallium Dorsal pallium is the hippocampal precursor by descent (comparative neuroanatomy); retains slow-timescale context integration and associative memory functions
E2 — fast transition forward model Cerebellum (lobes VII–X in larva) Cerebellar forward model for reafference prediction and sensorimotor calibration; ablation impairs motor adaptation in exactly the way E2 impairment does in REE
HippocampalModule — terrain-sensitive trajectory proposal Lateral pallium (CA3/DG precursor) Spatial navigation with lesion-sensitive memory; place-cell-like activity in larval pallium documented during spatial tasks
Multinomial go/no-go selection Optic tectum + reticulospinal command neurons Optic tectum drives orienting, approach, and avoidance; Mauthner cell network provides rapid all-or-nothing avoidance command — the canonical vertebrate go/no-go gate
Raw harm signal (nociception → z_harm) Habenula (lateral) + raphe serotonin system Lateral habenula encodes aversive expectation and drives punishment-learning via serotonin; habenulo-raphe circuit is the identified substrate for harm-state representation in zebrafish
Residue field (harm accumulation across time) Sustained habenula tonic activity Habenula remains tonically active after an aversive event beyond the event itself, implementing exactly the moral-residue accumulation REE requires
(absent) Commitment architecture (absent) Prefrontal/anterior cingulate equivalent No mature dlPFC or ACC equivalent is wired at 5–7 dpf. The larva has no commitment circuitry — it is always in the uncommitted, stimulus-responsive state, matching commitment_threshold=−1.0 in EXQ-223
(absent) Goal evaluation (z_goal) (absent) Orbitofrontal / vmPFC equivalent Reward valuation circuitry requiring vmPFC-like prospective evaluation is not operative at this stage

Why the zebrafish larva is the best match in any mapped vertebrate

The zebrafish larva is the only vertebrate for which the entire CNS has been functionally imaged at single-cell resolution during behaviour. The Ahrens lab (Janelia Research Campus) demonstrated whole-brain light-sheet calcium imaging in 2013, producing simultaneous activity recordings from ~100,000 neurons in a freely behaving animal (Ahrens MB, Orger MB, Robson DN, Li JM, Keller PJ, 2013, Nature Methods 10:413–420, doi:10.1038/nmeth.2434). This was extended to behavioural brain-state mapping across the full larval CNS during visuomotor tasks (Portugues R, Feierstein CE, Engert F, Orger MB, 2014, Neuron 81:1323–1338, doi:10.1016/j.neuron.2014.01.008).

The habenulo-raphe harm circuit has been characterised at the level of specific neuronal populations: the lateral habenula encodes aversive expectation and drives active avoidance via serotonergic suppression of reward systems (Amo R, Fredes F, Kinoshita M, Aoki R, Aizawa H, et al., 2014, Neuron 84:1034–1048, doi:10.1016/j.neuron.2014.10.035; Hikosaka O, 2010, Nat Rev Neurosci 11:503–513, doi:10.1038/nrn2866).

EXQ-223 confirms that the REE substrate, at this configuration level, produces the same class of gradient-following, harm-avoiding, resource-seeking behaviour these larvae show in their ecological niche — and does so using a circuit topology derived independently from functional-architecture arguments, not from neuroscience reverse-engineering.

Other well-characterised organisms at this approximate complexity level

  • Ciona intestinalis larva (177 neurons, complete connectome) — the most minimal mapped chordate. Ryan K, Lu Z, Meinertzhagen IA (2016, eLife 5:e16962, doi:10.7554/eLife.16962) produced the full wiring diagram. Has sensory vesicle (photoreceptors, gravity), anterior ganglion (proto-pallium / E1-like context), motor ganglion (go/no-go), tail musculature. This is the evolutionary ground-state of chordate nervous system organisation: the simplest living animal sharing common ancestor with vertebrates and carrying the same basic circuit plan. REE configured at EXQ-223 complexity is above Ciona (REE has a differentiated hippocampal module; Ciona’s anterior ganglion is undifferentiated proto-pallium).

  • C. elegans (302 neurons, complete connectome) — the most completely characterised nervous system in biology. White JG, Southgate E, Thomson JN, Brenner S (1986, Philos Trans R Soc Lond B 314:1–340) produced the original connectome; Varshney LR et al. (2011, PLoS Comput Biol 7:e1001066) provided its structural analysis. Has AVA/AVB command interneurons (go/no-go), ASH/ASJ nociceptors (harm signal), RIA/AIY interneurons (associative context, E1-like thermotaxis memory). No hippocampal precursor. REE at EXQ-223 complexity is above C. elegans — the HippocampalModule has no C. elegans homolog.

  • Lamprey — most basal living vertebrate with identifiable pallium (hippocampal precursor), primitive basal ganglia, and reticulospinal command neurons. Grillner S, Robertson B (2016, Curr Biol 26:R1088–R1100, doi:10.1016/j.cub.2016.06.029) reviewed the conservation of BG architecture across 500 million years of vertebrate evolution, with lamprey as the key reference point. Stephenson-Jones M, Samuelsson E, Ericsson J, Robertson B, Grillner S (2011, Curr Biol 21:1081–1091) demonstrated direct functional homology of lamprey striatum to mammalian BG. The lamprey is below zebrafish larva in the pallial differentiation gradient; REE at EXQ-223 complexity sits between lamprey (undifferentiated pallium) and adult zebrafish (commitment circuitry beginning to wire).

Significance

These organisms do not have deliberative minds. They navigate, avoid harm, and acquire resources through the same gradient-following loop that EXQ-223 confirms works in REE. The fact that a computational architecture derived from functional-philosophical arguments about the structure of ethical agency independently recapitulates the minimal vertebrate circuit plan is not trivial. It is the strongest empirical validation this project has produced to date of the claim that REE’s architecture is tracking something real about the structure of cognition.

What EXQ-223 is not

This experiment does not test deliberative, goal-directed, or commitment-gated behaviour. Those require SD-011/SD-012 substrates and the approach/goal experiments (EXQ-074e, EXQ-076e, ARC-030). EXQ-223 establishes that the foundation is sound before those additions are layered on.

Episode visualiser

A companion episode log format was added to the experiment script. Future runs emit {experiment_type}_{ts}_episode_log.json alongside the standard metrics output. episode_viewer.html in REE_assembly/ plays back an organism’s life on the gridworld with per-step mode annotation (avoid / approach / explore / neutral) and internal state timelines (z_harm_norm, z_beta, z_world change rate). Linked from the explorer.

2026-03-06: V2 Substrate Debt Resolution — SD-001, SD-002, SD-003

Overview

Completed V2 implementation in ree-v2 resolving all three registered substrate debts. All 22 agent tests pass after changes.

SD-001 — E2 / HippocampalModule Separation (CLOSED)

Verified that ree-v2 scaffolding already resolved SD-001:

  • E2FastPredictor is a pure transition model; generate_candidates_cem() removed.
  • HippocampalModule is a distinct V2 component handling CEM-style trajectory proposal.
  • No code changes required; closed SD-001 in GOVERNANCE_STATE.md with verification notes.

SD-002 — E1 Prior Wiring into HippocampalModule (CLOSED)

Implemented the confirmed stub at agent.py:187 (“not yet wired into HippocampalModule”):

ree_core/hippocampal/module.py:

  • terrain_prior network input expanded from latent_dim + 1 to 2*latent_dim + 1 (now concatenates z_beta, e1_prior, and residue_scalar).
  • _get_terrain_action_mean(z_beta, e1_prior=None) — accepts optional E1 prior; uses zeros as unconditioned fallback when not provided.
  • propose_trajectories(z_beta, num_candidates=None, e1_prior=None) — accepts and threads E1 prior through to terrain prior conditioning.
  • forward() updated to pass e1_prior through.

ree_core/agent.py:

  • generate_trajectories(): _, _prior = self.e1(z_beta) → _, e1_prior = self.e1(z_beta)
  • hippocampal.propose_trajectories(...) now called with e1_prior=e1_prior.
  • Docstring and module header updated to reflect SD-002 resolution.

docs/architecture/e2.md:

  • §4 SD-002 mutual constitution section updated: “does not yet wire it into HippocampalModule” replaced with implementation description and resolution date.

evidence/GOVERNANCE_STATE.md:

  • SD-002 entry updated with actions-completed note and implementation details.

SD-003 — CausalGridWorld / E2 Counterfactual Substrate (CLOSED)

Verified that ree-v2 scaffolding already resolved SD-003:

  • CausalGridWorld fully implemented in ree_core/environment/causal_grid_world.py: contamination_grid, footprint_grid, env-caused hazard drift, transition_type attribution in info dict, contamination/footprint observation channels.
  • E2FastPredictor.forward_counterfactual(z, a_cf) implemented and test-verified.
  • Causal delta z_actual_next - z_cf_next computable (test confirmed).
  • Closed SD-003 in GOVERNANCE_STATE.md; Step 2.4 experiments remain future work.

Preserved Sources

All changes are incremental additions; no prior formulations deleted or overwritten.

2026-03-02: Governance Cycle — Full Experiment Ingestion and Repo Design Improvements

Overview

Full governance catch-up covering two areas: (1) ingestion of EXQ-006 through EXQ-013 into GOVERNANCE_STATE.md, which had not been updated since 2026-02-27 despite 8 additional genuine experiments completing; and (2) a set of documentation design improvements to navigation, structure, and information architecture.

Experiment Ingestion (EXQ-006 through EXQ-013)

Eight genuine experiments completed between 2026-02-28 and 2026-03-01 were unregistered in GOVERNANCE_STATE.md. All have now been ingested:

EXQ Claim(s) Result Notes
EXQ-006 MECH-058 FAIL E1/E2 Associative vs Transition (MultiRoomGrid); substrate limit SD-001 confirmed in 2nd environment
EXQ-007 MECH-057 FAIL Multi-Step Control Completion Gating; substrate debt still blocks clean isolation
EXQ-008 MECH-063 PASS E3 Candidate Count Ablation; first genuine evidence for MECH-063
EXQ-009 MECH-062 PASS Residue Routing Weight Sensitivity; genuine confirmation of stable claim
EXQ-010 MECH-059 PASS Control Plane Precision Separation (extended seeds); confidence accumulation
EXQ-011 MECH-056 PASS Residue Trajectory Placement (extended seeds); confidence accumulation
EXQ-012 MECH-061 PASS Commitment Boundary Token Reclassification (extended seeds); confidence accumulation
EXQ-013 MECH-060, MECH-067 PASS Write-Locus Contamination (extended seeds); confidence accumulation

Claims now at 4 genuine runs: MECH-056, MECH-057, MECH-058, MECH-059, MECH-060, MECH-061, MECH-067. Claims with first genuine runs: MECH-062 (2 runs, PASS), MECH-063 (2 runs, PASS).

No status changes required in claims.yaml — MECH-062 was already stable and MECH-063 already provisional from prior adjudication. EXQ-008/009 provide first genuine experimental confirmation of those adjudicated statuses.

What Changed (Experiment Ingestion)

  • Updated evidence/GOVERNANCE_STATE.md:
    • Added EXQ-006 through EXQ-013 to completed experiments table
    • Fixed column order for EXQ-005 row (Experiment and Result were swapped)
    • Updated genuine run counts for MECH-056 through MECH-063 and MECH-067
    • Extended “Claims with Reliable Status Changes” to cover all new EXQs
    • Updated “Pending Genuine Experimentation” (queue exhausted after EXQ-013)
    • Added “Last updated: 2026-03-02” timestamp

Governance State Audit

Performed as part of this cycle:

  • Convergence packets: all 10 inbox packets (2026-02-22 to 2026-02-25) confirmed receipted. No unprocessed intake.
  • Thoughts: 51/51 processed as of 2026-02-25 sweep; no new unprocessed files.
  • Conflicts: all 4 documented conflicts resolved; no new conflicts detected.
  • V3 literature: 25 entries across 3 programs (v3_hippocampal_rollout: 11, v3_prefrontal_control: 7, v3_jepa_mapping_limits: 7) confirmed complete and indexed.
  • Pending governance decisions: MECH-040 and MECH-046 remain pending_user — awaiting genuine run evidence before demotion/promotion decision.
  • Roadmap step 2.0: remains in progress; V2 implementation spec not yet produced.

Repo Design Improvements

Applied a set of information architecture improvements identified via design critique:

  • Updated README.md:
    • Surfaced Explorer link (docs/claims/explorer.html) to second paragraph with description of Triple View and Governance modes (previously buried at end of file)
    • Added docs/notes/ to Working Structure section with a definition distinguishing it from docs/thoughts/ and docs/architecture/
    • Added YYYY-MM-DD_ filename convention note for docs/thoughts/
    • Removed redundant ## Explorer And Triple View section (content moved to intro)
  • Archived docs/FINAL_OUTPUT.md:
    • Added archive notice pointing to canonical copy at docs/processed/FINAL_OUTPUT.md
    • Created docs/processed/FINAL_OUTPUT.md as canonical going-forward home
    • Original preserved in place per no-deletion policy
  • Fixed filename docs/thoughts/17-02-26_necessary_separations_based_on_considering-prompt_injection.md → copied to docs/thoughts/2026-02-17_necessary_separations_based_on_considering-prompt_injection.md with redirect header; original preserved
  • Created docs/claims/active_claims_summary.yaml — lightweight summary of 104 active/provisional/stable claims (of 155 total) for human-readable reference without requiring tooling to parse claims.yaml

2026-02-25: Tightened Decoupling Wording (JEPA External, JEPA-Like In REE)

Overview

Refined canonical wording so REE references JEPA-like representational patterns while explicitly treating JEPA as a separate external project, not a REE architectural component.

What Changed

  • Strengthened external-project boundary wording:
    • README.md
    • docs/roadmap.md
    • docs/notes/jepa_language_policy.md
    • docs/glossary.md
  • Updated IMPL-022 framing from JEPA-family integration toward JEPA-like reference usage:
    • docs/architecture/jepa_e1e2_integration_contract.md
    • title/scope/purpose and control-boundary language now encode “external JEPA, internal REE”.
  • Updated mechanism/spec wording to use JEPA-like phrasing where appropriate:
    • docs/architecture/agency_responsibility_flow.md
    • docs/architecture/ree_v2_spec.md
    • docs/architecture/ree_v2_repo_bootstrap_spec.md
    • docs/REE_ARCHITECTURE_SNAPSHOT_2026-02-17.md
  • Aligned claim metadata/index text:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md

2026-02-25: Declared JEPA Decoupling (Representation Reference, Not REE Substrate)

Overview

Made the canonical boundary explicit: JEPA is treated as a representation reference architecture for E1/E2 interfaces, while E3 commitment/control/responsibility semantics remain REE-native and non-delegable.

What Changed

  • Updated repository-level boundary wording:
    • README.md
    • JEPA now described as representational reference architecture (not REE substrate ownership).
  • Added explicit decoupling declaration to JEPA integration contract:
    • docs/architecture/jepa_e1e2_integration_contract.md
    • now states JEPA cannot carry E3 commitment/control authority semantics.
  • Reworded MECH-058 JEPA framing from substrate to representational reference:
    • docs/architecture/agency_responsibility_flow.md
    • docs/claims/claim_index.md
    • docs/claims/claims.yaml
  • Updated language policy and glossary boundary phrasing:
    • docs/notes/jepa_language_policy.md
    • docs/glossary.md
  • Updated roadmap/spec wording to align with representation-interface framing:
    • docs/roadmap.md
    • docs/architecture/ree_v2_spec.md
    • docs/architecture/ree_v2_repo_bootstrap_spec.md
  • Updated architecture snapshot wording for JEPA position:
    • docs/REE_ARCHITECTURE_SNAPSHOT_2026-02-17.md

2026-02-25: GitHub Documentation + Triple-View Portability Refresh

Overview

Updated GitHub-facing onboarding docs and the claims explorer triple-view/governance UX references so they are portable across machines and aligned with current architecture/governance workflow.

What Changed

  • Refined root onboarding guidance and linked explorer/triple-view usage:
    • README.md
    • added human-first entry points, explorer launch steps, and workflow policy references.
  • Expanded and clarified triple-view runbook:
    • docs/architecture/diagram_views.md
    • renamed to “Triple-View” wording, documented explorer inspector behavior, and tightened update/CI guidance.
  • Updated repository metadata citation excerpt to match canonical file:
    • docs/repo_meta.md
    • aligned license/repository/author metadata to CITATION.cff, with legacy source retained as reference.
  • Hardened explorer portability and doc coverage:
    • docs/claims/explorer.html
    • removed machine-specific absolute command paths, generalized actor placeholder, improved file:// help text, and added missing architecture documents to the docs browser catalog.
  • Extended architecture freshness scope to include non-markdown contract artifacts:
    • .github/workflows/architecture-diagrams.yml
    • scripts/architecture/check_freshness.sh
    • hook_registry.v1.json changes now trigger freshness review.

2026-02-25: Clarified MECH-060 With Phase Control-Variation Sources/Outputs Map

Overview

Refined MECH-060 to explicitly include the functional control-plane variation contract across pre_commit, commit_boundary, and post_commit, including supersession-after-dispatch handling. This closes a readability gap where phase write rules existed but source/output variation mapping remained implicit across multiple files.

What Changed

  • Expanded canonical MECH-060 wording to include phase-scoped control variation contract:
    • docs/architecture/agency_responsibility_flow.md
    • added explicit functional table: phase -> variation sources -> control outputs -> writable sinks -> forbidden sinks.
  • Added supersession-path clarification in the same table:
    • docs/architecture/agency_responsibility_flow.md
    • made post-dispatch safety override handling explicit as append-only lineage, not commit rewrite.
  • Added token-envelope lineage/supersession metadata guidance:
    • docs/architecture/streams.md
    • clarified required and optional commit_boundary.tokenized_update fields for auditable supersession.
  • Updated claim index phrasing to match the refined scope:
    • docs/claims/claim_index.md
    • MECH-060 description now references the explicit phase control-variation map.

2026-02-25: Applied MECH-062 Promotion and Added Thought-Adjudication Governance Bridge

Overview

Closed a status-application gap where MECH-062 had an approved promotion decision but remained provisional in the claims registry, and added a first-class governance bridge so thought-intake deltas can explicitly surface adjudication follow-through candidates.

What Changed

  • Applied approved MECH-062 promotion into canonical claim status:
    • docs/claims/claims.yaml
    • MECH-062 status set to stable.
  • Aligned architecture status banner with claim registry:
    • docs/architecture/e3.md
    • MECH-062 section status updated to stable; implementation-gap language reframed as hardening/falsifiability work.
  • Recorded applied decision-log entry for traceability:
    • evidence/decisions/decision_log.v1.jsonl
    • appended MECH-062 decision_status=applied for promote_to_stable.
  • Added thought-derived adjudication bridge tooling and governance integration:
    • evidence/planning/scripts/build_thought_adjudication_bridge.py
    • evidence/planning/scripts/run_governance_cycle.py
    • evidence/planning/README.md
    • new outputs: evidence/planning/thought_adjudication_bridge.v1.json, evidence/planning/THOUGHT_ADJUDICATION_BRIDGE.md.

2026-02-25: Processed Determinism-Boundary + Prefrontal-Primitives Thought Intake (Queue to Zero)

Overview

Processed the remaining two thought files in the queue and mapped their deltas into existing canonical claim surfaces. No new claim IDs were added.

What Changed

  • Extended E3 implementation contract with constrained neuro-inspired primitive mapping:
    • docs/architecture/e3.md
    • added non-isomorphic mapping guidance for TaskLoopObject fields and typed-interface separation during tri-loop arbitration.
  • Clarified boundary determinism at commit-token transition:
    • docs/architecture/e3.md
    • added explicit note that stochastic internal rehearsal is compatible with deterministic responsibility-bearing projection.
  • Updated claim source lineage:
    • docs/claims/claims.yaml
    • added thought-source links for ARC-003, MECH-060, MECH-061, MECH-062, and INV-021.
  • Marked thought files as processed with backlinks:
    • docs/thoughts/2026-02-24_determinism_action_gating_boundary.md
    • docs/thoughts/2026-02-24_prefrontal_primitives.md

2026-02-25: Processed Task-Loop Extraction + Signed Social Coupling Thought Intake

Overview

Processed 2026-02-25_task_loop_extraction_and_latent_field_ethics.md into canonical REE docs using existing claim surfaces. The ingestion clarified E3 task-loop extraction contracts, signed social coupling semantics, and deterministic boundary interpretation without adding new claim IDs.

What Changed

  • Clarified E3 extracted-loop representation contract:
    • docs/architecture/e3.md
    • added implementation-facing TaskLoopObject contract and explicit relation to MECH-060/061/062.
  • Clarified social coupling as signed (benefit + harm), not harm-only:
    • docs/architecture/social.md
    • added signed viability coupling note, BENEFIT_j shadow-stream mention, and MECH-036 veto-scope alignment.
  • Clarified determinism scope at commitment boundaries:
    • docs/invariants.md
    • expanded INV-021 interpretation to distinguish internal stochasticity from boundary determinism/auditability.
  • Updated claim source lineage:
    • docs/claims/claims.yaml
    • added thought source links for ARC-003, ARC-010, MECH-052, MECH-061, MECH-062, and INV-021.
  • Marked thought file as processed with backlinks:
    • docs/thoughts/2026-02-25_task_loop_extraction_and_latent_field_ethics.md

2026-02-23: Re-centered Canonical Boundary (REE-First, Convergence-Owned JEPA Playbooks)

Overview

Reinforced repository boundaries so REE_assembly remains REE-core canonical authority while JEPA source-method “how-to integrate” guidance is explicitly convergence-owned.

What Changed

  • Tightened repository boundary language in root docs:
    • README.md
    • explicit REE-first canonical boundary and pointer to REE_convergence/sources/jepa/*.
  • Updated roadmap language away from JEPA-first canonical framing:
    • docs/roadmap.md
    • status snapshot, v2 framing, and immediate queue now encode convergence-owned JEPA integration playbooks.
  • Updated terminology policy from JEPA-first to REE-first canonical:
    • docs/notes/jepa_language_policy.md (IMPL-024)
    • docs/glossary.md (boundary wording section under IMPL-020)
    • docs/architecture/ree_v2_spec.md (terminology mode section)
  • Added explicit boundary note in JEPA substrate contract:
    • docs/architecture/jepa_e1e2_integration_contract.md
  • Strengthened convergence interface/extraction docs with JEPA playbook ownership:
    • evidence/planning/REE_CONVERGENCE_INTERFACE.md
    • evidence/planning/CONVERGENCE_EXTRACTION_MAP.md
  • Added governance-policy note for boundary discipline:
    • evidence/planning/planning_criteria.v1.yaml
    • new model_adjudication.canonical_boundary_policy.
  • Updated claim metadata/index text for IMPL-024 wording shift:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md

2026-02-23: Processed Subjective Skill-Drift Thought Intake (2 files)

Overview

Processed the remaining unprocessed subjective-experience thought pair and mapped them into existing canonical claims without creating new IDs. The update clarifies partial re-engagement behavior, trigger competition, and sleep-driven tactic expansion expectations.

What Changed

  • Extended tri-loop commitment guidance with overlay re-engagement and trigger-field routing:
    • docs/architecture/e3.md
    • added EXECUTE_BASE / OVERLAY_AUDIT / FULL_REPLAN state framing and bounded trigger-field arbitration notes.
  • Extended control-plane contract with bounded trigger-bandwidth policy:
    • docs/architecture/control_plane.md
    • added competitive trigger field guidance and axis-tuned trigger selection dynamics.
  • Extended sleep mechanism expectations for tactic-option expansion:
    • docs/architecture/sleep.md
    • added probe-facing expectation that sleep should improve tactic diversity, not only smoothness.
  • Updated claim source lineage:
    • docs/claims/claims.yaml
    • added thought-source links for MECH-030, MECH-057, MECH-060, MECH-062, and MECH-063.
  • Marked thought files as processed with backlinks:
    • docs/thoughts/2026-02-13_subjective_experience_pre_post_commit.md
    • docs/thoughts/2026-02-23_some_subjective_experience_mapping.md

2026-02-22: Convergence Receipt Loop (Packet Acknowledgment + Adjudication State)

Overview

Promoted convergence handoff from one-way packet submission to a closed-loop, machine-readable process by adding receipt artifacts and queue-level receipt/adjudication visibility.

What Changed

  • Added convergence receipt contract artifacts:
    • evidence/planning/schemas/v1/convergence_packet_receipt.schema.json
    • evidence/planning/CONVERGENCE_PACKET_RECEIPT_TEMPLATE.json
    • evidence/planning/convergence_packets/receipts/.gitkeep
  • Added receipt validator:
    • evidence/planning/scripts/validate_convergence_packet_receipt.py
  • Extended convergence queue builder:
    • evidence/planning/scripts/build_convergence_intake_queue.py
      • now ingests convergence_packets/receipts/*.json
      • annotates queue items with latest receipt queue/adjudication state
      • reports receipt validity/state counts in queue summary
      • emits invalid receipt section in markdown output
  • Updated convergence runbooks and inbox docs:
    • evidence/planning/convergence_packets/README.md
    • evidence/planning/README.md
    • added shared lifecycle state model and receipt requirements for closed-loop handoff.
  • Added first-class source-side roundtrip tooling in companion repo REE_convergence:
    • handoff/README.md
    • handoff/schemas/v1/convergence_promotion_packet.schema.json
    • handoff/schemas/v1/convergence_packet_receipt.schema.json
    • tools/build_convergence_promotion_packet.py
    • tools/validate_convergence_promotion_packet.py
    • tools/validate_convergence_packet_receipt.py
    • tools/run_cross_repo_handoff.py
    • this now establishes packet authoring/validation/submission + receipt/status mirror as a symmetric two-repo flow.

2026-02-22: Convergence Ingestion Route (External Knowledge -> Canonical Review)

Overview

Added a packet-first ingestion route so external architecture/implementation knowledge can be reviewed in a structured queue before any canonical promotion in REE_assembly.

What Changed

  • Added convergence boundary and extraction guidance:
    • evidence/planning/CONVERGENCE_EXTRACTION_MAP.md
    • evidence/planning/REE_CONVERGENCE_INTERFACE.md
  • Added promotion packet contract and template:
    • evidence/planning/schemas/v1/convergence_promotion_packet.schema.json
    • evidence/planning/CONVERGENCE_PROMOTION_PACKET_TEMPLATE.json
    • evidence/planning/convergence_packets/README.md
    • evidence/planning/convergence_packets/inbox/.gitkeep
  • Added packet validation/queue tooling:
    • evidence/planning/scripts/validate_convergence_promotion_packet.py
    • evidence/planning/scripts/build_convergence_intake_queue.py
  • Added first JEPA convergence packet in inbox:
    • evidence/planning/convergence_packets/inbox/2026-02-22_cpkt_jepa_baseline_hygiene.json
  • Integrated convergence queue generation into governance cycle:
    • evidence/planning/scripts/run_governance_cycle.py
    • new flags:
      • --skip-convergence-intake
      • --convergence-fail-on-invalid
  • Hardened convergence packet epistemic controls and risk gates:
    • evidence/planning/schemas/v1/convergence_promotion_packet.schema.json
      • required falsification section (claim under test, disconfirming conditions, test-plan refs)
      • required governance_controls (blast radius, conflict review, probation, rollback conditions/owner)
      • required implementation_readiness block (probe IDs, adapter patch refs, benchmark acceptance criteria)
      • conditional stricter requirements for interface/architecture blast-radius packets
    • evidence/planning/scripts/validate_convergence_promotion_packet.py
      • new flags:
        • --check-gate-readiness
        • --fail-on-gate-failure
    • evidence/planning/scripts/build_convergence_intake_queue.py
      • gate-failure reasons surfaced per packet
      • placeholder evidence token detection (TODO, Needs evidence, example.org) blocks gate-ready status
      • queue summary now tracks blast-radius and gate-failure counts
      • implementation-readiness metadata captured (probe/adapter/benchmark counts)
    • evidence/planning/scripts/run_governance_cycle.py
      • convergence triage now reports gate-failures/placeholder-evidence counts and blast-radius breakdown
      • convergence triage now reports packets carrying implementation-ready plans
  • Pruned canonical JEPA glossary scope and moved extended term-harmonization tables to convergence intake:
    • docs/glossary.md (IMPL-020 canonical core only)
    • REE_convergence/sources/jepa/glossary.md
    • REE_convergence/translations/glossary_master.md
  • Updated planning runbooks:
    • evidence/planning/README.md
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md
    • evidence/planning/planning_criteria.v1.yaml (convergence_ingestion section)

2026-02-21: Applied Epoch Metadata Plumbing (Ingestion + Dispatch)

Overview

Applied epoch propagation and producer-facing epoch tagging so current-epoch evidence is explicitly stamped and future dispatches carry architecture-epoch requirements by default.

What Changed

  • Extended experimental/literature ingestion to carry architecture_epoch into claim-evidence entries:
    • evidence/experiments/scripts/build_experiment_indexes.py
    • defaults post-epoch timestamps to current epoch when explicit tag is absent
  • Updated Experiment Pack contract guidance:
    • evidence/experiments/INTERFACE_CONTRACT.md
    • added recommended manifest.architecture_epoch field and epoch note
  • Updated weekly dispatch emitter:
    • evidence/planning/scripts/emit_weekly_dispatches.py
    • dispatch bundles now include current epoch context and explicit manifest tagging requirements
    • dispatch report now records epoch metadata
  • Regenerated epoch-aware outputs:
    • evidence/experiments/claim_evidence.v1.json
    • evidence/planning/architecture_epoch_applicability.v1.json
    • evidence/planning/governance_agenda.v1.json
    • evidence/planning/GOVERNANCE_AGENDA.md
    • evidence/planning/outbound_dispatches/2026-02-21/*

2026-02-21: Clarified MECH-060 Hierarchy (Non-Supersession)

Overview

Made the commit-boundary claim hierarchy explicit so MECH-060 is interpreted as an umbrella boundary principle, while MECH-062/066/067 are implementation realizations rather than replacements.

What Changed

  • Clarified MECH-060 role and interpretation rule:
    • docs/architecture/agency_responsibility_flow.md
    • added explicit “non-supersession” section linking MECH-060 to MECH-061/062/066/067
  • Clarified MECH-062 relationship to MECH-060:
    • docs/architecture/e3.md
    • added hierarchy note under layered eligibility refinement
  • Updated canonical claim summaries:
    • docs/claims/claim_index.md
    • MECH-060 and MECH-062 descriptions now encode umbrella-vs-realization framing

2026-02-21: Processed Remaining Thought Intake Queue (6 files)

Overview

Ingested the remaining unprocessed thought files into canonical architecture docs using existing claim surfaces (no new claim IDs), and marked each thought as processed with backlinks.

What Changed

  • Updated Papez functional mechanism framing:
    • docs/architecture/papez_circuit.md
    • added affective retrieval-augmentation interpretation and explicit commitment-boundary clarification for MECH-037
  • Updated E3 commitment mechanics:
    • docs/architecture/e3.md
    • added layered eligibility filter ordering and disinhibitory-gating rationale under MECH-062 implementation guidance
  • Updated control-plane wiring narrative:
    • docs/architecture/control_plane_signal_map.md
    • added loop-vector precision (P_motor, P_cognitive, P_value) and associative-loop governance calibration note
  • Added agent-shell compatibility notes to implementation contracts:
    • docs/architecture/hook_surface_contract.md
    • docs/architecture/ree_v2_repo_bootstrap_spec.md
  • Marked thought files as processed with canonical links:
    • docs/thoughts/2026-02-17_papez_RAG.md
    • docs/thoughts/2026-02-18_commitment_as_layered_eligibility_filters.md
    • docs/thoughts/2026-02-18_new_implementation_openclaw.md
    • docs/thoughts/2026-02-19_basal_ganglia_evolutionary_conservation_pull.md
    • docs/thoughts/2026-02-21_meta_critic.md
    • docs/thoughts/2026-02-21_more_control_plane_necessities.md
  • Refreshed sweep outputs:
    • docs/thoughts/thought_sweep.v1.json
    • docs/thoughts/SWEEP_REPORT.md

2026-02-19: Boundary-Conditioned Separation Claims for Commit and Sleep Loci

Overview

Added focused claims to resolve ambiguity around “strict separation” by defining enforceable write-boundary contracts for pre/post-commit learning and offline/online consolidation, without over-claiming full informational isolation.

What Changed

  • Added agency/responsibility mechanism claims:
    • docs/architecture/agency_responsibility_flow.md
    • added MECH-066 (boundary-conditioned channel separation with bounded coupling)
    • added MECH-067 (machine-checkable write-locus permission matrix)
    • expanded related-claims coverage with commit-boundary invariants
  • Added protected offline consolidation architecture claim:
    • docs/architecture/sleep.md
    • added ARC-020 to require typed authority/write boundaries during offline consolidation
  • Added cross-phase isolation invariant:
    • docs/invariants.md
    • added INV-024 (offline/online update-locus isolation for responsibility-bearing writes)
  • Updated claim registry and index:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md
    • added records/navigation for ARC-020, MECH-066, MECH-067, INV-024

2026-02-18: Processed Meta-Invariant Compression Thought

Overview

Processed a new thought proposing consolidation of many lower-level constraints into a smaller meta-invariant layer, with explicit guardrails so compression does not weaken existing architectural safety or responsibility boundaries.

What Changed

  • Added derived meta-invariant layer:
    • docs/invariants.md
    • added INV-019..INV-023:
      • selection compression boundary
      • authority stratification boundary
      • commit-boundary irreversibility
      • heterogeneous trust allocation
      • stability-preserving offline reweighting
    • explicitly marked as a compression lens (not replacement) over existing invariants
  • Updated claim registry/index:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md
    • added claim records and links for INV-019..INV-023
  • Updated control-plane stream contract:
    • docs/architecture/control_plane_signal_map.md
    • added explicit coverage mapping from signal/knob wiring to INV-019..INV-023 and associated failure signatures
  • Updated experiment model contracts:
    • evidence/experiments/stop_criteria.v1.yaml
    • added meta_invariant_compression_audit stop-criteria block
    • evidence/experiments/meta_invariant_compression_audit/experiment.md
    • added minimum metric contract for compressed-layer falsification
  • Marked thought file as processed with canonical links:
    • docs/thoughts/2026-02-18_selection_compression__symmetry_ and_meta_invariant_consolidation.md

2026-02-17: Follow-up Contract Hardening (Commit Token + RC Floor/Hysteresis)

Overview

Applied follow-up hardening from external architecture review feedback to make commitment traceability and reality-conflict calibration requirements explicit and testable.

What Changed

  • Strengthened commit-boundary token contract:
    • docs/architecture/e3.md
    • added minimal hard-floor fields: trajectory_hash, authority_path, policy_hash, caps_hash, verifier_result_hash, uncertainty_snapshot
    • updated Q-015 with explicit minimum field set candidate
  • Strengthened reality-conflict control law:
    • docs/architecture/control_plane_signal_map.md
    • added guarded Pi_rc floor requirement and hysteresis/recovery sketch (theta_high, theta_low, tau_rc_recovery)
    • expanded Q-018 calibration hooks and TODOs for floor + hysteresis contracts
  • Added measurable stop criteria for new probe lanes:
    • evidence/experiments/stop_criteria.v1.yaml
    • added criteria for:
      • claim_probe_arc_017
      • claim_probe_mech_062
      • claim_probe_mech_064
      • claim_probe_mech_065
  • Added required metric contracts in experiment templates:
    • evidence/experiments/claim_probe_arc_017/experiment.md
    • evidence/experiments/claim_probe_mech_062/experiment.md
    • evidence/experiments/claim_probe_mech_064/experiment.md
    • evidence/experiments/claim_probe_mech_065/experiment.md

2026-02-17: Integrated Injection-Resistance Stream Separation Thoughts

Overview

Processed two new thought-intake files on stream separation, typed authority boundaries, and reality-coherence conflict gating; promoted them into canonical claims, signal routing docs, and experiment model templates.

What Changed

  • Updated sensory stream contract:
    • docs/architecture/sensory_stream_tags.md
    • added explicit REALITY_COHERENCE derived stream
    • added typed channel overlay (OBS, INS, POL, ID, CAPS) and exteroceptive write-path constraints
  • Updated control-plane signal map:
    • docs/architecture/control_plane_signal_map.md
    • added S5 reality-coherence conflict lane and hierarchical stream/loop/global precision decomposition
    • added new claim sections:
      • MECH-064 typed authority and control-store separation
      • MECH-065 reality-coherence conflict lane
    • added open question:
      • Q-018 RC-conflict hysteresis calibration
    • added explicit correction for over-absolute assumptions: rapid safety interrupts may bypass deliberation, but still cannot grant privileged policy/identity/capability writes.
  • Updated claim registry/index:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md
    • added MECH-064, MECH-065, Q-018
    • expanded source/dependency links for MECH-004 and ARC-017
  • Added experimental model templates:
    • evidence/experiments/claim_probe_arc_017/experiment.md
    • evidence/experiments/claim_probe_mech_064/experiment.md
    • evidence/experiments/claim_probe_mech_065/experiment.md
    • updated evidence/experiments/claim_probe_mech_062/experiment.md to remove TODOs and align failure signatures with tri-loop + authority constraints
  • Marked thought files as processed with backlinks:
    • docs/thoughts/2026-02-17_control_plane_update.md
    • docs/thoughts/17-02-26_necessary_separations_based_on_considering-prompt_injection.md

2026-02-15: Processed Commit-Gating Refresh Thought

Overview

Processed the refreshed thought on commit gating, pre/post-commit error separation, and expanded control-plane axes, and promoted its unresolved parts into explicit implementation-gap contracts.

What Changed

  • Updated E3 canonical spec:
    • docs/architecture/e3.md
    • added explicit open implementation-gap contract for MECH-061/MECH-062:
      • commit-token wire format/scope
      • cross-gate arbitration policy
      • post-commit credit-assignment interface
      • responsibility-locus storage contract
  • Updated control-plane canonical spec:
    • docs/architecture/control_plane.md
    • added phenomenology-aligned diagnostic profile for “focused but initiation-suppressed” regimes
  • Marked thought as processed:
    • docs/thoughts/2026-02-15_basal_ganglia_commit_gating_control_plane_axes_refresh.md

2026-02-15: Frictionless Task Inbox Sync for Governance Carryover

Overview

Added a markdown checklist inbox that automatically syncs into manual carryover tracking so tasks are captured quickly and disappear from agenda carryover as soon as they are checked done.

What Changed

  • Added task inbox sync script:
    • evidence/planning/scripts/sync_task_inbox.py
    • parses evidence/planning/task_inbox.md checklist items
    • open (- [ ]) items are created/reopened in manual_carryover_items.v1.json
    • done (- [x]) items are marked status=done in carryover
    • done checkbox lines are pruned from inbox by default after sync (--no-prune-completed opt-out)
    • stable matching uses source refs (task_inbox:<sha1>) derived from normalized task text
  • Added default inbox file:
    • evidence/planning/task_inbox.md
  • Integrated task inbox sync into governance cycle:
    • evidence/planning/scripts/run_governance_cycle.py
    • new flags:
      • --skip-task-inbox-sync
      • --task-inbox-dry-run
  • Updated planning docs:
    • evidence/planning/README.md
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md

2026-02-14: Automated Adjudication Cascade Application

Overview

Added post-decision automation that applies adjudication outcomes to the claim registry and reopens dependent claims when structural outcomes (adopt_jepa_structure / retire_ree_claim) are selected.

What Changed

  • Added adjudication cascade script:
    • evidence/planning/scripts/apply_adjudication_cascade.py
    • reads latest decision-log entries by claim
    • applies configured outcome tokens for decision_status=applied
    • updates docs/claims/claims.yaml adjudication/cascade fields
    • emits:
      • evidence/decisions/adjudication_cascade_state.v1.json
      • evidence/planning/ADJUDICATION_CASCADE_PATCH_QUEUE.md
  • Integrated cascade step into governance cycle:
    • evidence/planning/scripts/run_governance_cycle.py
    • new step: adjudication_cascade (enabled by default)
    • new flags:
      • --skip-adjudication-cascade
      • --adjudication-cascade-statuses
      • --adjudication-cascade-dry-run
    • agenda now includes Adjudication Cascade checkpoint and action counts.
  • Updated runbooks:
    • evidence/planning/README.md
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md
    • evidence/decisions/README.md

2026-02-14: External-Precedence Governance Policy (JEPA vs REE)

Overview

Added an explicit governance policy that allows REE claim replacement when matched JEPA evidence is stronger, with cascade handling and anti-lock-in safeguards.

What Changed

  • Extended planning criteria with executable adjudication policy:
    • evidence/planning/planning_criteria.v1.yaml
    • added model_adjudication section:
      • allowed outcomes: retain_ree, hybridize, adopt_jepa_structure, retire_ree_claim
      • cascade policy for dependency reopen on structural replacement
      • temporary override mode for JEPA-internal proxy routing with provenance/calibration/rollback requirements
      • anti-lock-in gate
    • added thresholds for external-precedence detection.
  • Updated planning generation to propagate adjudication/cascade context into backlog and architecture gap outputs:
    • evidence/experiments/scripts/build_experiment_indexes.py
    • external-precedence and anti-lock-in signals are now computed and surfaced in generated planning artifacts.
  • Updated governance agenda builder to report model adjudication checkpoint:
    • evidence/planning/scripts/run_governance_cycle.py
    • includes candidate claims, allowed outcomes, cascade policy, and override mode in agenda outputs.
  • Added claim-level lifecycle/adjudication metadata for core JEPA-sensitive claims:
    • docs/claims/claims.yaml
    • updated MECH-056, MECH-058, MECH-059, MECH-060, and IMPL-022.

2026-02-14: JEPA Control-Signal Mapping Matrix and Proxy Contract

Overview

Hardened JEPA integration wording and interfaces so REE control-plane routing can use internal JEPA proxies without collapsing control ownership into the substrate layer.

What Changed

  • Updated JEPA integration contract with explicit JEPA↔REE control signal mapping matrix:
    • docs/architecture/jepa_e1e2_integration_contract.md
    • clarified that precision is not guaranteed as an explicit calibrated JEPA output channel
    • added execution table for MECH-056, MECH-058, MECH-059, MECH-060
    • added proxy-bank declaration contract and explicit provenance/calibration/separability/ablation checks
  • Updated v2 substrate spec with JEPA control-proxy profile and hard-fail thresholds:
    • docs/architecture/ree_v2_spec.md
    • added jepa_control_proxy_ablation profile
    • added proxy metrics and readiness thresholds (proxy_bank_coverage_rate, calibration ECE, residual separability, ablation utility)
  • Extended adapter signal schema and producer contract for proxy-bank metadata:
    • evidence/experiments/schemas/v1/jepa_adapter_signals.v1.json
    • evidence/experiments/INTERFACE_CONTRACT.md
    • proxy declarations are optional in base v1, but conditional metrics are required when proxy_bank is present

2026-02-14: Connectome Literature Pull Implementation

Overview

Added a generated connectome-oriented literature pull queue to turn architecture pressure signals into concrete, claim-scoped evidence pulls with source-preserving translation and confidence decomposition requirements.

What Changed

  • Added connectome pull generator:
    • evidence/planning/scripts/build_connectome_literature_pull.py
    • outputs:
      • evidence/planning/connectome_literature_pull.v1.json
      • evidence/planning/CONNECTOME_LITERATURE_PULL.md
  • Integrated connectome pull generation into governance cycle:
    • evidence/planning/scripts/run_governance_cycle.py
    • new step: connectome_pull
    • new agenda checkpoint: Connectome Literature Pull
  • Extended literature contract and schema for source-preserving translation and confidence decomposition:
    • evidence/literature/INTERFACE_CONTRACT.md
    • evidence/literature/schemas/v1/literature_evidence.schema.json
    • new optional record fields:
      • mapping (source_claim_statement, ree_translation, mapping_caveat, optional source_context)
      • confidence_components (source_quality, mapping_fidelity, transfer_risk, optional notes)
  • Added planning schema for connectome pull artifact:
    • evidence/planning/schemas/v1/connectome_literature_pull.schema.json

2026-02-14: MECH-059 Terminology Standardization

Overview

Standardized MECH-059 language to avoid ambiguity between uncertainty and precision by using: confidence channel (uncertainty-derived precision), explicitly separate from residual prediction error.

What Changed

  • Updated canonical mechanism wording:
    • docs/architecture/agency_responsibility_flow.md#mech-059
  • Updated JEPA contract wording:
    • docs/architecture/jepa_e1e2_integration_contract.md
  • Updated claim metadata and index label:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md
  • Updated v2 planning/spec references:
    • docs/architecture/ree_v2_spec.md
    • docs/architecture/ree_v2_repo_bootstrap_spec.md
  • Updated governance decision packet phrasing:
    • evidence/decisions/decision_packet_v3_conflicts_2026-02-14.md

2026-02-14: Structure Review Dossier Protocol

Overview

Added a generated human-decision packet for architecture pressure claims so governance can review claim context, REE fit, evidence mix, alternatives, and inspiration options before deciding structural changes.

What Changed

  • Added structure review dossier generator:
    • evidence/planning/scripts/build_structure_review_dossiers.py
    • generates per-claim dossiers from architecture_gap_register.v1.json and claim_evidence.v1.json
    • includes plain-English claim description followed by plain-English REE whole-system fit description
    • includes evidence-mix interpretation, alternative hypotheses with confidence estimates, source wording vs REE translation, and left-field suggestions
  • Added generated dossier outputs:
    • evidence/planning/structure_review/<YYYY-MM-DD>/<CLAIM_ID>/DOSSIER.md
    • evidence/planning/structure_review/<YYYY-MM-DD>/<CLAIM_ID>/dossier.v1.json
    • evidence/planning/structure_review/latest/INDEX.md
    • evidence/planning/structure_review/latest/structure_review_report.v1.json
  • Integrated dossier generation into governance cycle:
    • evidence/planning/scripts/run_governance_cycle.py
    • new step: structure_review
    • new agenda checkpoint: Structure Dossiers
  • Extended planning thresholds to support soft structural trigger logic:
    • evidence/planning/planning_criteria.v1.yaml
    • trigger outcome is consider_new_structure (advisory), not automatic refactor.

2026-02-14: High-Priority Literature Dispatch Pack

Overview

Added an explicit dispatch bundle for the current high-priority literature backlog so governance-critical claims can be worked in a single focused evidence cycle.

What Changed

  • Added literature dispatch pack:
    • evidence/planning/DISPATCH_LITERATURE_HIGH_PRIORITY_2026-02-14.md
  • Includes:
    • proposal queue for LIT-0005, LIT-0008, LIT-0013, LIT-0016, LIT-0018, LIT-0020, LIT-0022
    • acceptance checks and a copy/paste execution prompt for REE_assembly

2026-02-14: REE-v2 Cutover Adjudication Gate

Overview

Added a reusable cutover readiness checker that allows justified ree-v2 vs ree-v1-minimal direction divergence when explicit adjudication criteria pass, instead of treating all mismatches as automatic failures.

What Changed

  • Added cutover checker script:
    • evidence/planning/scripts/check_ree_v2_cutover_readiness.py
    • evaluates latest producer handoff snapshots and latest ingestion-cycle reports
    • replaces strict overlap gate with overlap_sanity_adjudicated_or_resolved
    • enforces per-claim adjudication checks:
      • matched protocol evidence (seeds/conditions/schema version set)
      • ree-v2 CI pass status
      • metric improvement rule (>=3 claim-critical metrics improved on >=2/3 seeds)
      • no contract:* drift signatures
      • no unaddressed P0 falsification for claim in latest lab handoff
      • governance note presence with rationale and rollback trigger
  • Updated planning readme with cutover command:
    • evidence/planning/README.md

2026-02-14: Cross-Version Hook Surface Framework

Overview

Added a versioned hook framework so v2 can expose forward-compatible extension points for v3/v4 without repeated interface churn.

What Changed

  • Added hook contract document:
    • docs/architecture/hook_surface_contract.md#impl-025
  • Added initial hook registry:
    • docs/architecture/hook_registry.v1.json
    • includes v2_required, v3_planned, and v4_plus_reserved tiers
  • Updated v2 spec to consume hook framework:
    • docs/architecture/ree_v2_spec.md#impl-023
  • Added claim/index/mapping entries:
    • docs/claims/claims.yaml (IMPL-025)
    • docs/claims/claim_index.md (IMPL-025)
    • docs/claims/subsystem_map.yaml

2026-02-14: REE-v2 JEPA Source and Cloud Compute Planning

Overview

Extended the REE-v2 bootstrap plan to explicitly require JEPA source provenance locking and a local-vs-cloud execution policy suitable for laptop-constrained development environments.

What Changed

  • Expanded REE-v2 bootstrap spec with:
    • JEPA source acquisition/provenance contract (third_party/jepa_sources.lock.v1.json)
    • local-vs-cloud offload gate and remote export/import gate
    • hobby-mode runtime placement baseline updated to 360m/360m for macbook_air_m2_2022
    • local hardware options/cost visibility gate (docs/ops/local_compute_options.md)
    • explicit hobby-mode buy/hold thresholds in EUR for structured decisions
    • migration stage updates and cutover acceptance updates
    • docs/architecture/ree_v2_repo_bootstrap_spec.md
  • Updated REE-v2 bootstrap dispatch prompt to include:
    • MacBook Air M2 local constraints
    • required remote execution scripts and dry-run acceptance checks
    • JEPA provenance fields required in manifest.scenario
    • required local compute options/cost sheet deliverable and output table
    • evidence/planning/DISPATCH_REE_V2_BOOTSTRAP.md
  • Extended weekly handoff template and policy to report compute placement:
    • execution_mode, compute_backend, runtime_minutes
    • remote export/import CI gate status
    • ree-v2 local compute options watch section for buy/hold decisions
    • evidence/planning/WEEKLY_HANDOFF_TEMPLATE.md
    • evidence/experiments/CROSS_REPO_SYNC_POLICY.md
    • evidence/planning/DISPATCH_WEEKLY_HANDOFF_FORMAT_UPDATE_2026-02-14.md
  • Added structured operator checklist for spare-time workflow:
    • evidence/planning/HOBBY_OPERATOR_PLAYBOOK.md

2026-02-14: Local Cadence Automation Scripts

Overview

Added local-scheduler automation for weekly producer handoff sync/import and outbound dispatch generation.

What Changed

  • Added cadence automation config:
    • evidence/planning/cadence_automation.v1.json
  • Added handoff sync/import script:
    • evidence/planning/scripts/sync_weekly_handoffs.py
    • supports day-based repo selection, handoff validation, run-pack import, and optional ingestion/governance steps
  • Added weekly dispatch emission script:
    • evidence/planning/scripts/emit_weekly_dispatches.py
    • builds per-repo dispatch markdown bundles from experiment_proposals.v1.json
  • Added local scheduling runbook:
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md
  • Updated planning docs for discoverability and handoff-path guidance:
    • evidence/planning/README.md
    • evidence/planning/WEEKLY_HANDOFF_TEMPLATE.md
    • evidence/experiments/CROSS_REPO_SYNC_POLICY.md

2026-02-14: Weekly Cross-Repo Handoff Standardization

Overview

Standardized weekly producer handoff format across qualification, stress, and parity lanes and added dispatch prompts to upgrade producer repos to the shared format.

What Changed

  • Updated cross-repo sync policy to include all producer repos and staggered weekly cadence:
    • evidence/experiments/CROSS_REPO_SYNC_POLICY.md
  • Added shared weekly handoff template:
    • evidence/planning/WEEKLY_HANDOFF_TEMPLATE.md
  • Added copy/paste dispatch bundle for producer format upgrades:
    • evidence/planning/DISPATCH_WEEKLY_HANDOFF_FORMAT_UPDATE_2026-02-14.md
  • Updated planning README:
    • evidence/planning/README.md

2026-02-14: Qualification Lane Ownership Update (ree-v2)

Overview

Updated v2 planning docs so ree-v2 (not ree-v1-minimal) is the intended qualification lane, with ree-v1-minimal retained as a transitional baseline during migration.

What Changed

  • Updated qualification-lane wording in:
    • docs/architecture/ree_v2_spec.md#impl-023
  • Updated repository-role framing in:
    • docs/roadmap.md#impl-008

2026-02-14: REE-v2 Spec Expansion (Implementation Detail Pass)

Overview

Expanded the v2 spec from framing-level draft into an implementation-ready contract with explicit module layout, configuration requirements, qualification profiles, milestones, and go/no-go criteria.

What Changed

  • Expanded:
    • docs/architecture/ree_v2_spec.md#impl-023
  • Added concrete sections for:
    • implementation objectives,
    • reference module layout,
    • interface detail and configuration contract,
    • qualification profiles (MECH-058/059/060),
    • failure thresholds and milestone gates (M0-M4),
    • acceptance checklist for v2 completion.

2026-02-14: v2 JEPA-First Wording Policy

Overview

Added explicit v2 terminology policy so docs and user-facing interactions use JEPA-first language with inline REE translation, while keeping machine contracts stable.

What Changed

  • Added policy doc:
    • docs/notes/jepa_language_policy.md#impl-024
  • Wired v2 spec to policy:
    • docs/architecture/ree_v2_spec.md#impl-023
  • Added operating-procedure rule:
    • docs/README.md#impl-013
  • Added glossary policy note:
    • docs/glossary.md#impl-020
  • Added claim/index/mapping entries:
    • docs/claims/claims.yaml (IMPL-024)
    • docs/claims/claim_index.md (IMPL-024)
    • docs/claims/subsystem_map.yaml

2026-02-14: Roadmap Refresh and REE-v2 Spec Draft

Overview

Refreshed the roadmap to reflect current repository roles and moved program framing to a substrate-first v2 followed by control-completion v3. Added a dedicated v2 implementation spec document.

What Changed

  • Rewrote roadmap:
    • docs/roadmap.md#impl-008
    • reframed v1 as qualification harness outcome
    • defined v2 as JEPA-aligned substrate phase
    • defined v3 as control-plane/hippocampal/E3 completion phase
  • Added v2 spec:
    • docs/architecture/ree_v2_spec.md#impl-023
  • Added claims/index mapping for v2 spec:
    • docs/claims/claims.yaml (IMPL-023)
    • docs/claims/claim_index.md (IMPL-023)
    • docs/claims/subsystem_map.yaml (JEPA-REE alignment mapping includes v2 spec)
  • Added literature dispatch pack for v3-critical evidence:
    • evidence/planning/DISPATCH_LIT_V3_HIPPO_PFC_JEPA.md

2026-02-13: Invariant Rewording for Consequence Persistence (INV-004/INV-006)

Overview

Reworded persistence invariants to be mechanism-agnostic now that consequence retention is distributed across more specific architecture claims (including but not limited to residue geometry).

What Changed

  • Updated invariant language in:
    • docs/invariants.md
    • INV-004 now: post-commit consequence traces are persistent
    • INV-006 now: post-commit consequence traces cannot be erased
  • Updated claim registry/index wording:
    • docs/claims/claims.yaml (subjects for INV-004, INV-006)
    • docs/claims/claim_index.md
  • Updated dependency wording in:
    • docs/architecture/residue_geometry.md

2026-02-13: Targeted Literature Entries for MECH-058/059/060

Overview

Added first-pass literature triangulation entries for new JEPA follow-on claims and refreshed evidence indexes.

What Changed

  • Added literature evidence entries:
    • evidence/literature/targeted_review_mech_058/entries/2026-02-13_mech058_ijepa_target_encoder_anchor/record.json
    • evidence/literature/targeted_review_mech_058/entries/2026-02-13_mech058_ijepa_target_encoder_anchor/summary.md
    • evidence/literature/targeted_review_mech_059/entries/2026-02-13_mech059_vjvcr_uncertainty_channeling/record.json
    • evidence/literature/targeted_review_mech_059/entries/2026-02-13_mech059_vjvcr_uncertainty_channeling/summary.md
    • evidence/literature/targeted_review_mech_060/entries/2026-02-13_mech060_vjepa2_dual_loss_channels/record.json
    • evidence/literature/targeted_review_mech_060/entries/2026-02-13_mech060_vjepa2_dual_loss_channels/summary.md
  • Regenerated evidence outputs via:
    • python3 evidence/experiments/scripts/build_experiment_indexes.py

2026-02-13: Dispatch Bundle for MECH-058/059/060

Overview

Added a focused cross-repo dispatch bundle for first-pass evidence generation on the new JEPA follow-on mechanism claims.

What Changed

  • Added dispatch bundle with copy/paste prompts:
    • evidence/planning/DISPATCH_JEPA_MECH_058_060.md
  • Updated planning README to document curated dispatch files:
    • evidence/planning/README.md

2026-02-13: JEPA Follow-On Claim Stubs (MECH-058/059/060, Q-013/014)

Overview

Added draft claim stubs capturing JEPA-driven follow-on mechanisms and open questions for governance/evidence routing.

What Changed

  • Added mechanism stubs in:
    • docs/architecture/agency_responsibility_flow.md
    • MECH-058 slow target-anchor timescale separation
    • MECH-059 latent uncertainty stream separation
    • MECH-060 pre-commit vs post-commit dual error channels
  • Added open questions in:
    • docs/architecture/agency_responsibility_flow.md
    • Q-013 deterministic vs stochastic uncertainty calibration
    • Q-014 invariance blind-spot risk for ethical attribution
  • Added claim registry and index entries:
    • docs/claims/claims.yaml
    • docs/claims/claim_index.md

2026-02-13: Cross-Repo Contract Sync Policy

Overview

Added a minimal lockstep policy so producer repos can stay aligned with REE_assembly experiment contract schemas and fail fast on drift.

What Changed

  • Added policy file:
    • evidence/experiments/CROSS_REPO_SYNC_POLICY.md
  • Updated evidence README index:
    • evidence/experiments/README.md

2026-02-13: JEPA Adapter Signals Schema + Ingestion Validation

Overview

Added a machine-readable JEPA adapter signal contract and wired ingestion to validate declared adapter artifacts so contract drift is surfaced automatically as FAIL evidence.

What Changed

  • Added schema:
    • evidence/experiments/schemas/v1/jepa_adapter_signals.v1.json
  • Extended manifest artifact schema:
    • evidence/experiments/schemas/v1/manifest.schema.json (artifacts.adapter_signals_path)
  • Extended ingestion:
    • evidence/experiments/scripts/build_experiment_indexes.py
    • validates adapter signal files when declared in manifest
    • marks invalid/missing adapter artifacts as FAIL with contract:jepa_adapter_signals_*
    • shows adapter contract status in per-experiment indexes
  • Updated experiment contract docs:
    • evidence/experiments/INTERFACE_CONTRACT.md
    • evidence/experiments/README.md
  • Added example adapter artifacts in dummy run packs:
    • evidence/experiments/trajectory_integrity/runs/*/jepa_adapter_signals.v1.json

2026-02-13: Hybrid JEPA to REE Diagram Spec + JEPA Literature Pull

Overview

Added a canonical hybrid-diagram contract so JEPA↔REE alignment can be discussed with consistent structure, then ingested targeted JEPA-paper evidence into the literature pipeline for MECH-057/Q-012.

What Changed

  • Added hybrid diagram contract:
    • docs/architecture/jepa_ree_hybrid_diagram_spec.md#impl-021
  • Added JEPA-as-E1/E2 integration contract:
    • docs/architecture/jepa_e1e2_integration_contract.md#impl-022
  • Added claim registry/index entries:
    • docs/claims/claims.yaml (IMPL-021, IMPL-022)
    • docs/claims/claim_index.md (IMPL-021, IMPL-022)
    • docs/claims/subsystem_map.yaml (new JEPA-REE Hybrid Alignment mapping)
  • Added targeted JEPA literature entries:
    • evidence/literature/targeted_review_mech_057/entries/*
  • Refreshed evidence indexes/matrices using:
    • evidence/experiments/scripts/build_experiment_indexes.py

2026-02-13: IMPL-022 Signal-Contract Tightening (JEPA Native Streams)

Overview

Tightened the JEPA integration contract so adapter outputs are explicit about what is native to JEPA versus what must be derived for REE precision-control usage.

What Changed

  • Updated docs/architecture/jepa_e1e2_integration_contract.md#impl-022 with:
    • JEPA-native signal map table (z, z_hat, residuals, uncertainty, trace fields),
    • signed-PE boundary rule (adapter emits unsigned substrate streams only),
    • additional required knobs for residual/uncertainty export mode,
    • extra metrics/checks for uncertainty provenance and precision-input completeness.

2026-02-13: Thought Processing - Formal JEPA to REE Alignment Glossary

Overview

Processed a dedicated JEPA↔REE terminology alignment thought into a formal interoperability glossary section. This pass is documentation-level alignment: it does not claim equivalence of objectives, only consistent mapping between representation/prediction/control terms across both frames.

What Changed

  • Expanded and formalized JEPA↔REE mapping in:
    • docs/glossary.md#impl-020
  • Added claim registry entry:
    • docs/claims/claims.yaml (IMPL-020)
  • Added claim index entry:
    • docs/claims/claim_index.md (IMPL-020)
  • Marked thought as processed with canonical links:
    • docs/thoughts/2026-02-13_jepa_ree_formal_alignment_glossary.md

2026-02-13: Thought Processing - JEPA to REE Convergence

Overview

Processed the JEPA↔REE convergence thought into a new candidate mechanism and an explicit falsifiability question. The integration captures a minimal architectural claim (latent predictive geometry plus control completion pressure) without committing to the strong claim as settled.

What Changed

  • Added mechanism MECH-057 in:
    • docs/architecture/agency_responsibility_flow.md#mech-057
  • Added open question Q-012 in:
    • docs/architecture/agency_responsibility_flow.md#q-012
  • Added JEPA↔REE vocabulary bridge in:
    • docs/glossary.md
  • Added claim registry entries:
    • docs/claims/claims.yaml (MECH-057, Q-012)
  • Added claim index entries:
    • docs/claims/claim_index.md (MECH-057, Q-012)
  • Marked thought as processed with canonical links:
    • docs/thoughts/2026-02-13_LeCun_developed_lots_of_REE.md

2026-02-13: Thought Processing - Self-First, Social-Later Roadmap Heuristic

Overview

Processed the developmental-roadmap thought and promoted it as a governance/testing heuristic without converting it into a new architectural commitment.

What Changed

  • Added implementation note IMPL-019 in:
    • docs/architecture/developmental_curriculum.md#impl-019
  • Added claim registry entry:
    • docs/claims/claims.yaml (IMPL-019)
  • Added claim index entry:
    • docs/claims/claim_index.md (IMPL-019)
  • Marked thought as processed with canonical links:
    • docs/thoughts/DEV-ROADMAP-SELF-FIRST-SOCIAL-LATER.md

2026-02-13: Claims Explorer Governance Dashboard Tab

Overview

Extended the HTML claims explorer so it can act as a governance dashboard for the new evidence loop.

What Changed

  • Updated docs/claims/explorer.html with a new Governance view tab.
  • Added dashboard rendering for:
    • governance cycle summary cards,
    • thought-intake queue,
    • conflict queue,
    • decision queue,
    • high-priority backlog and proposal snapshots,
    • dispatch-by-target-repo,
    • and maintenance flags/warnings.
  • Added staleness indicator based on evidence/planning/governance_agenda.v1.json.
  • Added copy-command helpers (manual CLI flow):
    • run governance cycle,
    • strict thought-gate cycle,
    • record-decision template command.

2026-02-13: Governance Cycle Orchestrator

Overview

Added a single helper that runs non-decision upkeep steps and emits a structured conversation agenda.

What Changed

  • Added script:
    • evidence/planning/scripts/run_governance_cycle.py
  • Script executes:
    • docs/thoughts/scripts/thought_sweep.py
    • evidence/experiments/scripts/build_experiment_indexes.py
  • Added generated agenda outputs:
    • evidence/planning/governance_agenda.v1.json
    • evidence/planning/GOVERNANCE_AGENDA.md
  • Updated planning docs:
    • evidence/planning/README.md

2026-02-13: Deterministic Thought Sweep Helper

Overview

Added a small utility to make thought-intake sweeps deterministic and repeatable.

What Changed

  • Added script:
    • docs/thoughts/scripts/thought_sweep.py
  • Added generated sweep outputs:
    • docs/thoughts/thought_sweep.v1.json
    • docs/thoughts/SWEEP_REPORT.md
  • Updated usage guidance in:
    • docs/thoughts/README.md

2026-02-13: Human-in-the-Loop Decision Queue, Confidence Channels, and Literature Merge

Overview

Extended the evidence subsystem so promotion/demotion decisions are surfaced as explicit review items, claim confidence is channelized by source, and literature evidence can be ingested alongside experiments.

What Changed

  • Added decision thresholds file evidence/experiments/decision_criteria.v1.yaml.
  • Added persistent decision governance files:
    • evidence/decisions/decision_log.v1.jsonl (append-only)
    • evidence/decisions/decision_state.v1.json (generated snapshot)
    • evidence/decisions/schemas/v1/decision_log_entry.schema.json
  • Extended ingestion script evidence/experiments/scripts/build_experiment_indexes.py to:
    • merge evidence/literature/**/entries/**/record.json with experiment evidence,
    • compute experimental_confidence, literature_confidence, and overall_confidence,
    • generate evidence/experiments/conflicts.md,
    • generate evidence/experiments/promotion_demotion_recommendations.md with human decision status,
    • read decision history from evidence/decisions/decision_log.v1.jsonl,
    • and generate decision state snapshot evidence/decisions/decision_state.v1.json,
    • and refresh evidence/literature/INDEX.md.
  • Added decision logging helper script:
    • evidence/experiments/scripts/record_decision.py
  • Added planning subsystem and generated loop outputs:
    • evidence/planning/planning_criteria.v1.yaml
    • evidence/planning/evidence_backlog.v1.json
    • evidence/planning/experiment_proposals.v1.json
    • evidence/planning/INDEX.md
    • evidence/planning/schemas/v1/evidence_backlog.schema.json
    • evidence/planning/schemas/v1/experiment_proposals.schema.json
  • Extended ingestion script to generate planning backlog/proposals directly from matrix + conflicts + decision log.
  • Extended claim-evidence schema evidence/experiments/schemas/v1/claim_evidence.schema.json with:
    • source_type split (experimental vs literature),
    • entry-level confidence fields,
    • claim-level confidence channels and source counts.
  • Added literature subsystem docs and schema:
    • evidence/literature/README.md
    • evidence/literature/INTERFACE_CONTRACT.md
    • evidence/literature/schemas/v1/literature_evidence.schema.json
  • Added example literature evidence record:
    • evidence/literature/neuro_pe_habenula_da/entries/2026-02-13_habenula_da_signed_pe_review/record.json
    • evidence/literature/neuro_pe_habenula_da/entries/2026-02-13_habenula_da_signed_pe_review/summary.md

2026-02-13: Experimental Evidence Ingestion Pipeline

Overview

Added a repository-level experiment ingestion subsystem so external implementation runs can be indexed in REE_assembly and translated into design-facing implications and TODOs.

What Changed

  • Added evidence/experiments/ with a versioned Experiment Pack contract:
    • manifest.json, metrics.json, summary.md, optional traces/ and media/.
  • Added schemas in evidence/experiments/schemas/v1/ for manifest and metrics files.
  • Added claim-evidence matrix schema in evidence/experiments/schemas/v1/claim_evidence.schema.json.
  • Added versioned stop criteria in evidence/experiments/stop_criteria.v1.yaml.
  • Added ingestion/index script evidence/experiments/scripts/build_experiment_indexes.py (standard library only) that:
    • scans evidence/experiments/**/runs/**/manifest.json,
    • flags FAIL runs using manifest status plus stop-criteria checks,
    • generates per-experiment and top-level indexes,
    • computes key metric deltas across runs,
    • populates evidence/experiments/claim_evidence.v1.json from claim_ids_tested,
    • auto-updates design implications in experiment templates,
    • and refreshes a failure-driven TODO queue.
  • Added strict producer-facing interface contract in evidence/experiments/INTERFACE_CONTRACT.md.
  • Added usage/readme documentation in evidence/experiments/README.md.
  • Added example experiment type trajectory_integrity with dummy PASS/FAIL runs demonstrating criteria-triggered failure indexing.

2026-02-12: Consistency Sweep (Dependencies, Conflicts, Open Questions)

Overview

Ran a repo-wide documentation consistency pass to tighten dependency symmetry, conflict tracking coverage, and open-question navigation.

What Changed

  • Added missing conflict records:
    • docs/conflicts/valence_vectors_vs_mu_kappa_overlays.md
    • docs/conflicts/care_override_vs_other_harm_veto.md
    • docs/conflicts/rollout_entropy_floor_vs_residue_persistence.md
  • Added missing resolution note and backlink for the previously resolved ethics conflict:
    • docs/conflicts/resolutions/2026-02-08_ethics-module-vs-cost-term.md
    • docs/conflicts/ethics_module_vs_cost_term.md
  • Updated docs/conflicts/README.md to index all active conflicts and map Q-008/Q-009/Q-011 to their trackers.
  • Added tracked-conflict links in open-question sections of:
    • docs/architecture/control_plane.md (Q-008)
    • docs/architecture/social.md (Q-009)
    • docs/architecture/hippocampal_systems.md (Q-011)
  • Updated docs/claims/claim_index.md for status symmetry:
    • Marked MECH-008 and MECH-020 as legacy in summaries.
    • Added conflict-tracker references to Q-008/Q-009/Q-011 index entries.
  • Updated docs/claims/claims.yaml so IMPL-017 depends on Q-008/Q-009/Q-011 and cites new conflict sources.
  • Logged an explicit open-question registry sweep in docs/notes/synthesis_passes.md (Q-001 through Q-011 coverage/triage).

2026-02-12: Justification Invariants and Operational Control-State Hooks

Overview

Converted prior discussion points into explicit architecture contracts: justification-gate invariants, operational control-state equations, and focused promotion criteria for the next claim-status pass.

What Changed

  • Added a formal Justification gate invariants section to docs/architecture/control_plane.md (necessity, imminence, proportionality, explainability, accountability) with decision inequalities over rollout sets.
  • Added operational control-state variables and illustrative equations to docs/architecture/precision_control.md (A_t, N_t, R_t, C_t; plus K1, K2, K2_H, K2_B, K10 update hooks).
  • Added telemetry requirements in docs/architecture/precision_control.md to log tonic/phasic control dynamics and interrupt triggers.
  • Added a focused promotion-gate entry in docs/notes/synthesis_passes.md for MECH-004, MECH-053, MECH-054, and MECH-055 with explicit promotion criteria.

2026-02-12: Thought Sweep - Failure Geometry and Residue Placement

Overview

Processed new 2026-02-12 thought files by promoting trajectory-vs-representation residue placement into a candidate mechanism, adding a rollout-collapse open question, and extending the implementation failure taxonomy.

What Changed

  • Added MECH-056 (trajectory-first residue placement) to docs/architecture/residue_geometry.md.
  • Added Q-011 (minimum rollout-diversity floor under repeated harm) to docs/architecture/hippocampal_systems.md.
  • Extended docs/REE_failure_modes.md with trajectory-space collapse and a failure-vector coordinate framing.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md for MECH-056, Q-011, and IMPL-005 source/dependency updates.
  • Marked 2026-02-12 thought files as processed with canonical links:
    • docs/thoughts/2026-02-12_DEPRESSIVE-PATH-PRUNING-HIPPOCAMPAL-ROLLBACK.md
    • docs/thoughts/2026-02-12_TRAJECTORY-RESIDUE-VS-REPRESENTATIONAL-DISTORTION.md
    • docs/thoughts/FAILURE-2026-02-12_COORDINATE-SYSTEM-FOR-COGNITIVE-PATHOLOGY.md

2026-02-11: Mode Priors, Stability Overlays, and Phase Separation

Overview

Integrated new thought intakes on amygdala‑style mode priors, μ/κ stability overlays, phase compartmentalisation, columnar motif vs geometry, and control‑plane math sketches.

What Changed

  • Added MECH‑046 (amygdala analogue mode priors) and MECH‑048 (μ/κ stability overlays) to docs/architecture/control_plane.md.
  • Added MECH‑047 (pre‑commitment mode manager with hysteresis) and a non‑binding math sketch to docs/architecture/mode_manager.md.
  • Added MECH‑049 (temporal phase compartmentalisation for ethical stability) to docs/architecture/temporal_dynamics.md.
  • Added MECH‑050 (functional locality without column geometry) to docs/architecture/entities_and_binding.md.
  • Added MECH‑051 (oxytocin/vasopressin relational topology modulation) and MECH‑052 (prolactin care persistence) to docs/architecture/social.md.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md.
  • Marked new 2026‑02‑11 thought files as processed with canonical links.
  • Re‑scoped MECH‑008 as legacy and added open questions Q‑008 (valence vs μ/κ overlays) and Q‑009 (care weights vs other‑harm veto).
  • Promoted MECH‑033, MECH‑034, MECH‑039, and MECH‑040 from candidate to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Promoted core control‑plane, modes, social, and agency mechanisms (MECH‑002, MECH‑005, MECH‑022 through MECH‑031, MECH‑036, MECH‑041) to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Promoted MECH‑021, MECH‑043, MECH‑044, and MECH‑045 to provisional and demoted MECH‑020 to legacy, logged in docs/notes/synthesis_passes.md.
  • Promoted MECH‑003 and MECH‑032 to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Promoted MECH‑037 to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Promoted MECH‑046, MECH‑047, and MECH‑048 to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Promoted MECH‑006 and MECH‑007 to provisional and logged the review in docs/notes/synthesis_passes.md.
  • Added MECH‑053 (habenula‑like aversive PE gate) and MECH‑054 (signed harm/benefit PE precision), plus Q‑010 on hedonic/valence/PE separation; updated control‑plane signal map.
  • Added a signed harm/benefit PE precision schema to docs/architecture/precision_control.md aligned with MECH‑054.
  • Added MECH‑055 (affective channel separation) and marked Q‑010 as legacy (resolved into MECH‑055).
  • Added a calibration note to docs/architecture/precision_control.md deferring tuning rules until the signal map and astrocyte stack are clarified.
  • Expanded signed harm/benefit PE routing details in docs/architecture/control_plane_signal_map.md (K2_H/K2_B plus aversive-gate elevation of S3/K10).

2026-02-10: Control-Plane Channel Clarification

Overview

Clarified control-plane channels vs modes, separated safety baseline from volatility, and updated evidence anchors.

What Changed

  • Added control-plane channel vs mode framing and safety baseline/volatility split in docs/architecture/control_plane.md.
  • Threaded baseline/volatility channels into docs/architecture/control_plane_signal_map.md.
  • Added a functional brainstem‑to‑control‑plane mapping table in docs/architecture/control_plane_signal_map.md.
  • Added channel‑space note to docs/architecture/modes_of_cognition.md.
  • Added MECH-039 and MECH-040 to the claim registry and index.
  • Expanded PAG evidence to include columnar defensive repertoires and flagged P28 as a safety extension in docs/notes/evidence_map.md.
  • Expanded evidence‑map implications for basal ganglia gating, hyperdirect stopping, LC‑NE baseline/volatility, PPN readiness, ARAS distribution, DRN arousal gating, and PAG defensive repertoires in docs/notes/evidence_map.md.
  • Added a control‑plane evidence summary to docs/notes/evidence_map.md.
  • Added a control‑plane synthesis pass entry in docs/notes/synthesis_passes.md and promoted MECH‑019 to provisional.
  • Added a hippocampal systems synthesis pass entry and subsystem abstract in docs/architecture/hippocampal_systems.md.
  • Added a modes of cognition synthesis pass entry and subsystem abstract in docs/architecture/modes_of_cognition.md.
  • Added an E‑stack synthesis pass entry and subsystem abstracts for docs/architecture/e1.md, docs/architecture/e2.md, and docs/architecture/e3.md.
  • Added an L‑space synthesis pass entry and subsystem abstract in docs/architecture/l_space.md.
  • Added oscillation/coupling evidence anchors P30–P36 to docs/notes/evidence_map.md for L‑space depth coordination and EEG caveats.
  • Abstracted oscillatory evidence implications in docs/architecture/l_space.md to keep coordination non‑anatomical.
  • Clarified residue vs viability mapping and online vs offline residue integration in docs/architecture/residue_geometry.md.
  • Added residue cross‑links in docs/architecture/sleep.md and docs/architecture/hippocampal_systems.md.
  • Added residue integration anchors P37–P39 in docs/notes/evidence_map.md and a neuro‑anchored consolidation note in docs/architecture/residue_geometry.md.
  • Added implementation hint block for residue integration in docs/architecture/residue_geometry.md.
  • Added a clarification block in docs/architecture/control_plane.md defining emotion as a composite control‑plane regime.
  • Added open question Q‑007 on universal expression ↔ control‑channel mapping in docs/architecture/control_plane.md and docs/claims/claims.yaml.
  • Added MECH‑041 (affective expression as mode broadcast) in docs/architecture/social.md and recorded thought intake in docs/thoughts/2026-02-10_affective_expression_mode_broadcast.md.
  • Added cranial‑nerve / brainstem evidence anchors P40–P43 to docs/notes/evidence_map.md and linked them from docs/architecture/social.md.
  • Added an evidence summary and safety‑function note for MECH‑041 in docs/architecture/social.md.
  • Added MECH‑042 (control‑plane telemetry exposure) in docs/architecture/control_plane.md and recorded thought intake in docs/thoughts/2026-02-10_control_plane_telemetry.md.
  • Expanded MECH‑042 with a developmental safety rationale (diagnostics without trauma).
  • Expanded telemetry thought intake with preverbal development rationale in docs/thoughts/2026-02-10_control_plane_telemetry.md.
  • Added a social subsystem synthesis pass entry and subsystem abstract in docs/architecture/social.md.
  • Added an offline integration synthesis pass entry and subsystem abstract in docs/architecture/sleep.md.
  • Added subsystem abstracts and synthesis pass entries for residue geometry, default mode, and agency/responsibility flow.
  • Added representation‑consolidation (refactor analog) note in docs/architecture/sleep.md and evidence anchors P44–P46 in docs/notes/evidence_map.md.
  • Added evidence anchors P47–P55 for entities/binding and precision control/scoping in docs/notes/evidence_map.md.
  • Formalized ACh/NE expected vs unexpected uncertainty split in docs/architecture/control_plane_signal_map.md, and added MECH‑043 (dopamine precision weighting) plus MECH‑044 (hippocampal relational binding) with supporting notes in docs/architecture/precision_control.md, docs/architecture/precision_scoping.md, and docs/architecture/entities_and_binding.md.
  • Formalized object‑file‑like persistence as MECH‑045 in docs/architecture/entities_and_binding.md and recorded thought intake in docs/thoughts/2026-02-10_object_file_persistence.md.
  • Clarified that OTHER_SELFLIKE detection depends on object persistence/binding in docs/architecture/social.md.
  • Added docs/architecture/developmental_curriculum.md as ARC‑019 to formalize staged training and curriculum gates.
  • Added per‑stage evidence anchors (P‑numbers) to docs/architecture/developmental_curriculum.md.
  • Documented evidence‑anchor usage style in docs/notes/evidence_map.md and aligned section‑level anchor phrasing in docs/architecture/control_plane_signal_map.md and docs/architecture/social.md.
  • Aligned evidence‑anchor usage in docs/architecture/entities_and_binding.md, docs/architecture/residue_geometry.md, and docs/architecture/sleep.md (section‑level anchors, no inline mixing).
  • Removed ARC‑003 as a dependency of ARC‑005 to break the E‑stack/control‑plane cycle.
  • Clarified the non‑cyclic relationship between control plane and E3 in docs/architecture/control_plane.md.

2026-02-09: Subsystem Map and Rollout Wording Alignment

Overview

Aligned subsystem labels and clarified hippocampal rollout wording in legacy‑adjacent docs and roadmap.

What Changed

  • Renamed subsystem map groups in docs/claims/subsystem_map.yaml.
  • Clarified hippocampal rollout wording in docs/architecture/trajectory_selection.md, docs/architecture/precision_control.md, docs/MIGRATION.md, and docs/roadmap.md.
  • Added rollout notes to docs/architecture/overview.md and docs/REE_overview.md.

2026-02-09: Rollout Terminology Clarification

Overview

Defined rollout vs forward prediction to prevent E2/HPC terminology collisions.

What Changed

  • Added glossary entries to docs/glossary.md.
  • Clarified terminology in docs/architecture/e2.md, docs/architecture/trajectory_selection.md, docs/architecture/e3.md, docs/architecture/l_space.md, docs/architecture/control_plane.md, docs/architecture/control_plane_signal_map.md, and docs/architecture/overview.md.
  • Updated rollout references in docs/REE_MIN_SPEC.md, docs/REE_overview.md, docs/examples/toy_world_scoring_functions.md, and docs/architecture/language.md.

2026-02-09: Empathy Calibration Note

Overview

Clarified that OTHER_SELFLIKE detection should favor high recall (tolerating false positives) over false negatives.

What Changed

  • Added calibration note to docs/architecture/social.md.
  • Added MECH-032 to formalize the high-recall bias.

2026-02-09: Thought Intake — Kernel, Viability, Valence, and Other-Harm Gating

Overview

Integrated four refinement thoughts: the E2 kernel → hippocampal rollout interface, viability vs residue updates, vector-valued valence ranking, and other-harm gating thresholds.

What Changed

  • Added MECH-033 to docs/architecture/hippocampal_systems.md (kernel chaining interface).
  • Added MECH-034 to docs/architecture/residue_geometry.md (viability vs residue updates).
  • Added MECH-035 to docs/architecture/sensory_stream_tags.md (vector-valued valence ranking).
  • Added MECH-036 to docs/architecture/social.md (other-harm veto threshold).
  • Added thought files under docs/thoughts/2026-02-09_*.md.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md.

2026-02-09: Papez-Like Provenance Gating

Overview

Added a Papez-like provenance-gating mechanism to constrain confabulation and require hippocampal trace support before commitment.

What Changed

  • Added docs/architecture/papez_circuit.md (MECH-037).
  • Added docs/thoughts/2026-02-09_papez_circuit_reality_filtering.md.
  • Updated docs/architecture/hippocampal_systems.md with a Papez-loop cross-reference.
  • Updated docs/claims/claims.yaml, docs/claims/claim_index.md, and docs/claims/subsystem_map.yaml.
  • Added evidence-map anchors for Papez circuit and confabulation in docs/notes/evidence_map.md.
  • Added an operational checklist to docs/architecture/papez_circuit.md.

2026-02-09: Arcuate Fasciculus Language Nudges

Overview

Added an arcuate-fasciculus functional analog to capture dorsal sequence→motor language nudges without anatomy claims.

What Changed

  • Added docs/architecture/arcuate_fasciculus.md (MECH-038).
  • Added docs/thoughts/2026-02-09_arcuate_fasciculus_language_nudges.md.
  • Updated docs/architecture/language.md and docs/claims/claims.yaml.
  • Updated docs/architecture/language/emergence_and_bootstrapping.md with a dual-stream nudge.
  • Updated docs/claims/claim_index.md and docs/claims/subsystem_map.yaml.
  • Added evidence-map anchors for arcuate fasciculus and dual-stream language pathways in docs/notes/evidence_map.md.

Overview

Expanded the evidence map with additional neuroscience and psychiatry anchors to guide REE refinement.

What Changed

  • Added sections P14–P23 to docs/notes/evidence_map.md (hippocampal replay dynamics, basal ganglia gating, stop-signal hyperdirect pathway, LC‑NE adaptive gain, precision‑weighting in psychosis, efference copy for inner speech, and DMN autobiographical memory).
  • Added sections P24–P29 (PPN brainstem hub, STN interrupt, ARAS arousal pathways, dorsal raphe arousal, PAG defensive control, and basal ganglia pathway dynamics).

2026-02-09: Evidence Map — Neuroscience Anchors

Overview

Added a lightweight evidence map linking REE prompts to primary neuroscience sources.

What Changed

  • Added docs/notes/evidence_map.md.

2026-02-09: Claim Index — New Claims Added

Overview

Added explicit claim entries for newly introduced architectural and social mechanism clarifications.

What Changed

  • Added ARC-018 (hippocampal rollouts + viability mapping) to docs/architecture/hippocampal_systems.md.
  • Added MECH-031 (derived social tags + empathy coupling) to docs/architecture/social.md.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md.

2026-02-09: Thought Intake — Empathy as Self/Other Coupling

Overview

Clarified empathy as control-plane coupling over derived social tags (AGENCY, OTHER_SELFLIKE), adding shadow bundles and coupling knobs without a new subsystem.

What Changed

  • Updated docs/architecture/social.md with derived social tags and fast empathy routing.
  • Added optional social coupling knobs to docs/architecture/control_plane.md.
  • Added social extension tags to docs/architecture/sensory_stream_tags.md.
  • Added docs/thoughts/2026-02-09_empathy.md (processed).
  • Updated docs/claims/claims.yaml sources for ARC-010, ARC-005, ARC-017.

2026-02-09: Thought Intake — Language as Compression and Coordination

Overview

Positioned language as a compression layer emerging from joint attention and simulation overhead; no new architecture.

What Changed

  • Updated docs/architecture/language.md with joint attention + compression pressure framing.
  • Updated docs/architecture/language/emergence_and_bootstrapping.md with joint attention and minimal nudges.
  • Added docs/thoughts/2026-02-09_language.md (processed).
  • Updated docs/claims/claims.yaml sources for ARC-009 and MECH-010.

2026-02-09: Thought Intake — Starting With Sensory Streams (Rollouts + HPC Mapping)

Overview

Reaffirmed minimal stream tags, mapped to E1/E2/E3, and clarified hippocampus as the sole multi-step rollout engine with post-commitment viability mapping.

What Changed

  • Updated docs/architecture/sensory_stream_tags.md with engine mapping and reward-readonly note.
  • Clarified hippocampal rollouts in docs/architecture/e2.md, docs/architecture/e3.md, docs/architecture/hippocampal_systems.md, docs/architecture/trajectory_selection.md, and docs/architecture/control_plane_signal_map.md.
  • Added rollout tuning clarification to docs/architecture/control_plane.md.
  • Added docs/thoughts/2026-02-09_starting_with_sensory_streams.md (processed).
  • Updated docs/claims/claims.yaml sources for ARC-002, ARC-003, ARC-005, ARC-007, ARC-017, MECH-004, IMPL-016.

2026-02-08: Thought Intake — Minimal Sensory Stream Tags

Overview

Captured a minimal tagging scheme for sensory, precision, and self-impact streams, framing reward-like signals as emergent rather than primitive.

What Changed

  • Added docs/architecture/sensory_stream_tags.md (ARC-017).
  • Added docs/thoughts/2026-02-08_sensory_stream_tags.md.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md.

2026-02-08: Thought Intake — Residue as Cognitive Map

Overview

Captured a clarification that residue geometry is both a terrain and a recorded path, framing hippocampal traces as a cognitive map of thought trajectories.

What Changed

  • Updated docs/architecture/residue_geometry.md (ARC-013) to describe residue as terrain plus recorded traversal paths.
  • Added docs/thoughts/2026-02-08_residue_paths_cognitive_map.md.
  • Updated docs/claims/claim_index.md summary for ARC-013.

2026-02-08: Legacy M Alignment and Claim Triage

Overview

Aligned canonical docs to treat the explicit ethical cost term (M) as legacy-only, and triaged open-question statuses.

What Changed

  • Marked explicit (M) references as legacy/evaluation-only in docs/glossary.md, docs/REE_MIN_SPEC.md, docs/examples/toy_world_environment.md, docs/examples/toy_world_scoring_functions.md, docs/architecture/trajectory_selection.md, docs/architecture/social.md, and docs/REE_overview.md.
  • Clarified failure mode #5 in docs/REE_failure_modes.md to reference legacy (M) proxies or residue penalties rather than an explicit cost term.
  • Updated docs/claims/claims.yaml to mark open questions (Q-001..Q-006) as active.

2026-02-08: Candidate Claim Promotions

Overview

Promoted selected candidate claims to stable or provisional based on maturity and architectural scope.

What Changed

  • Promoted INV-018 to stable (agency required).
  • Promoted ARC-006, ARC-008, ARC-015, and ARC-016 to provisional.
  • Updated claim registry and aligned section headers where needed.

2026-02-08: Typed-Claims Docs Rebuild

Overview

Rebuilt canonical REE documentation as a typed-claims system assembled from preserved sources with minimal glue text.

What Changed

  • Added docs/processed/legacy_tree/ with a verbatim mirror of all documentation-like files and a manifest.
  • Refreshed canonical docs under docs/ with claim IDs, typed headers, and required end sections.
  • Added new canonical files: docs/architecture/hippocampal_systems.md and docs/architecture/temporal_dynamics.md.
  • Rebuilt docs/claims/claims.yaml and docs/claims/claim_index.md to index all current claims.
  • Added docs/conflicts/ethics_module_vs_cost_term.md to capture a documented tension.
  • Rewrote docs/README.md as executable agent guidance with verbatim prompts.

Legacy Preservation

All prior documentation-like files are preserved verbatim under docs/processed/legacy_tree/ with their original paths.

Conflicts Recorded

  • docs/conflicts/ethics_module_vs_cost_term.md

2026-02-08: Subsystem Promotions and Conflict Resolution

Overview

Promoted language, social, and sleep subsystems into canonical typed-claims docs; added astrocyte-aware control-plane hypothesis; resolved the ethics-cost-term conflict.

What Changed

  • Added canonical subsystem docs:
    • docs/architecture/language.md
    • docs/architecture/social.md
    • docs/architecture/sleep.md
  • Added docs/architecture/astrocyte_regulatory_stack.md as a control-plane mechanism hypothesis.
  • Updated docs/claims/claims.yaml and docs/claims/claim_index.md with new ARC/MECH/Q claims.
  • Resolved docs/conflicts/ethics_module_vs_cost_term.md and updated canonical E3 framing to remove explicit cost terms.

2026-02-08: Minimum Spec Alignment

Overview

Aligned the minimum specification with the current canonical framing that removes explicit ethical cost terms.

What Changed

  • Added canonical notes to docs/REE_MIN_SPEC.md clarifying that explicit ethical cost terms are legacy.

2026-02-08: Extended Preservation Sweep

Overview

Extended preservation to non-markdown REE-relevant files and extracted code comment fragments.

What Changed

  • Added CITATION.cff and .github/ISSUE_TEMPLATE/*.yml to docs/processed/legacy_tree/.
  • Added docs/processed/extracted_fragments/code_comments.md with verbatim excerpts from src/cathedral.py.

2026-02-08: Astrocyte Stack Relocation

Overview

Relocated the astrocyte-aware regulatory stack subsystem out of docs/ to keep canonical docs focused on typed-claims files.

What Changed

  • Moved docs/astrocyte_aware_regulatory_stack/ to architecture/astrocyte_aware_regulatory_stack/.

2026-02-08: Full Legacy Capture and Intake Workflow

Overview

Captured all original documentation-like sources into canonical typed-claims files or indexed claims, and added a thoughts/resolution intake workflow.

What Changed

  • Added canonical docs for remaining architecture, language, sleep, and example files.
  • Added docs/thoughts/ and docs/conflicts/resolutions/ with workflow guidance.
  • Added repository meta, roadmap, and notes docs to the canonical set.
  • Updated claim registry and claim index to include all remaining sources.

2026-02-08: Thought Intake — Control Plane, Modes, Responsibility

Overview

Processed a new thought on control-plane mode shaping, temporal “now” construction, hippocampal hypothesis injection, and responsibility flow.

What Changed

  • Added docs/thoughts/2026-02-08_control_plane_modes_responsibility_flow.md.
  • Added docs/architecture/agency_responsibility_flow.md and updated control-plane, mode-manager, temporal-dynamics, and hippocampal systems docs.
  • Added INV-018, ARC-015, MECH-019..MECH-024, and Q-006 to the claim registry and index.

2026-02-08: Thought Intake — Modes as Control-Plane Regimes

Overview

Processed a new thought detailing control-plane regimes for action, vigilance, reflective/DMN cognition, sleep modes, and pathological extremes.

What Changed

  • Added docs/thoughts/2026-02-08_modes_of_cognition_control_plane_regimes.md.
  • Updated docs/architecture/mode_manager.md, docs/architecture/default_mode.md, and docs/architecture/sleep.md.
  • Added ARC-016 and MECH-025..MECH-030 to the claim registry and index.

Overview

Separated the control-plane regime taxonomy into its own architecture file and cross-linked regime pathologies into the failure-mode taxonomy.

What Changed

  • Added docs/architecture/modes_of_cognition.md and trimmed regime details from docs/architecture/mode_manager.md.
  • Updated claim registry/index links for ARC-016 and MECH-025..MECH-028.
  • Linked regime pathologies into docs/REE_failure_modes.md.

2026-02-07: Systematic Documentation Refactoring

Overview

Completed a comprehensive 8-phase refactoring of REE documentation to create a structured, dependency-aware system that explicitly surfaces inconsistencies rather than resolving them silently.

Phase 1: Inventory ✓

Completed:

  • Scanned all documentation files (.md format)
  • Identified 56+ markdown files across repository
  • Catalogued core concepts: E1, E2, E3, L-space, control plane, residue, entities, binding
  • Located repeated definitions and potential conflicts

Key Findings:

  • Core architecture well-documented in architecture/ directory
  • Three critical root documents: README.md, REE_CORE.md, DANIEL_README.md
  • Subsystems organized under: sleep/, language/, social/
  • Existing docs/ directory with REE_MIN_SPEC.md and other specifications

Phase 2: Canonical Structure ✓

Created:

  • docs/README.md — Navigation guide explaining structure and reading paths
  • docs/invariants.md — 17 non-negotiable architectural invariants
  • docs/architecture/ — Canonical architecture documents
    • e1.md, e2.md, e3.md
    • l_space.md (renamed from latent_stack.md)
    • control_plane.md
    • residue_geometry.md
    • default_mode.md
    • hippocampal_braid.md
    • entities_and_binding.md (newly synthesized)
  • docs/claims/ — Claim registry directory
  • docs/conflicts/ — Conflict documentation directory (prepared)
  • docs/archive/ — Archive directory (prepared)

Directory Structure:

docs/
  README.md
  invariants.md
  glossary.md (existing, to be updated)
  MIGRATION.md
  changelog.md (this file)
  architecture/
    e1.md, e2.md, e3.md
    l_space.md
    control_plane.md
    entities_and_binding.md
    residue_geometry.md
    default_mode.md
    hippocampal_braid.md
  claims/
    claim_index.md
    claims.yaml
  conflicts/ (empty)
  archive/ (empty)

Phase 3: Claim Typing ✓

Applied to 9 canonical architecture files:

Standard header format:

**Claim Type:** [invariant | architectural_commitment | mechanism_hypothesis | open_question | implementation_note]
**Scope:** [Brief description]
**Depends On:** [List of dependencies]
**Status:** [stable | provisional | legacy | candidate]

Files Updated:

  • All 9 canonical architecture files received claim typing headers
  • Dependencies explicitly linked
  • Status clearly marked

Phase 4: Content Migration (Partial)

Completed:

  • Migrated core architecture files to canonical locations
  • Preserved original files in architecture/ for compatibility
  • No content deleted or archived yet (per global rules)

Deferred:

  • Full content consolidation from supplementary files
  • Archive migration for superseded content
  • Detailed duplicate removal

Rationale: Prioritized structure and claim typing over content consolidation to establish foundation first.

Phase 5: Claim Registry ✓

Created:

  • docs/claims/claims.yaml — Machine-readable registry with full metadata
  • docs/claims/claim_index.md — Human-readable indexed list

Claims Documented:

  • 17 Invariants (INV-001 to INV-017)
    • Core thesis invariants (3): No ethics module, temporal binding, language emergence
    • Safety invariants (4): Persistent ethics, mirror modeling, residue persistence, language constraint
    • Control invariants (2): Routed precision, attention via modulation
    • Mode invariants (3): Offline integration, imagination safety, commitment gates
    • Foundational invariants (5): Predictive cognition, content/control separation, constraint-based ethics, stability priority, control failures
  • 9 Architectural Commitments (ARC-001 to ARC-009)
    • Core components (3): E1, E2, E3
    • Latent/control (3): L-space, control plane, residue geometry
    • Modes/memory (2): Default mode, hippocampal braid
    • Entity representation (1): Entities and binding
  • 2 Mechanism Hypotheses (MECH-001 to MECH-002)
    • Phase compatibility binding (candidate)
    • τ-scoped precision (candidate)

Total: 28 claims

Phase 6: Consistency Pass ✓

Automated Analysis Results:

  • ✓ No circular dependencies
  • ✓ All invariants depend only on other invariants
  • ✓ No contradictions detected in same-subject claims
  • ✓ No documented conflicts

Dependency Graph Verified:

  • 11 axiomatic invariants (no dependencies)
  • 6 derived invariants (depend on other invariants)
  • 9 architectural commitments (depend on invariants and other commitments)
  • 2 mechanism hypotheses (depend on invariants and commitments)

Classification Safety:

  • No invariants depend on hypotheses or open questions ✓
  • No architectural commitments promote mechanism hypotheses without justification ✓
  • Provisional and candidate statuses appropriately marked ✓

Phase 7: Linking and Hygiene (Partial)

Completed:

  • Created docs/MIGRATION.md with path mapping
  • Internal links within canonical docs point to canonical locations
  • Cross-references added to architecture files

Remaining:

  • Update root documents (README.md, REE_CORE.md) to reference canonical locations
  • Add redirect notes to original architecture/ files
  • Verify all orphaned sections

Open Questions

None noted in preserved sources.

  • IMPL-014

References / Source Fragments

  • docs/processed/legacy_tree/docs/changelog.md

Phase 8: Final Output ✓

See sections below.

Documentation Tree 

/home/runner/work/Reflective-Ethical-Engine-assembled/Reflective-Ethical-Engine-assembled/
├── README.md (root, to be updated with canonical links)
├── REE_CORE.md (canonical spine, preserved)
├── DANIEL_README.md (refinement process, preserved)
├── CONTRIBUTING.md
├── docs/
│   ├── README.md ✓ NEW
│   ├── invariants.md ✓ NEW
│   ├── glossary.md (existing, to be updated)
│   ├── MIGRATION.md ✓ NEW
│   ├── changelog.md ✓ NEW (this file)
│   ├── REE_MIN_SPEC.md (existing, preserved)
│   ├── REE_overview.md (existing, to be reviewed)
│   ├── REE_failure_modes.md (existing, to be reviewed)
│   ├── architecture/
│   │   ├── e1.md ✓ MIGRATED
│   │   ├── e2.md ✓ MIGRATED
│   │   ├── e3.md ✓ MIGRATED
│   │   ├── l_space.md ✓ MIGRATED
│   │   ├── control_plane.md ✓ MIGRATED
│   │   ├── residue_geometry.md ✓ MIGRATED
│   │   ├── default_mode.md ✓ MIGRATED
│   │   ├── hippocampal_braid.md ✓ MIGRATED
│   │   └── entities_and_binding.md ✓ NEW
│   ├── claims/
│   │   ├── claims.yaml ✓ NEW
│   │   └── claim_index.md ✓ NEW
│   ├── conflicts/ (prepared, empty)
│   ├── archive/ (prepared, empty)
│   └── astrocyte_aware_regulatory_stack/ (existing subsystem)
├── architecture/ (original files preserved for compatibility)
│   ├── E1.md, E2.md, E3.md (originals)
│   ├── latent_stack.md, control_plane.md, etc.
│   ├── trajectory_selection.md (supplementary)
│   ├── depth_phase_spec.md (mechanism hypothesis)
│   ├── precision_scoping.md (mechanism hypothesis)
│   ├── sleep/, language/, social/ (subsystems)
│   └── ... (other supplementary files)
└── ... (other repository content)

Migration Summary

Newly Created Files (11)

  1. docs/README.md
  2. docs/invariants.md
  3. docs/MIGRATION.md
  4. docs/changelog.md
  5. docs/architecture/e1.md
  6. docs/architecture/e2.md
  7. docs/architecture/e3.md
  8. docs/architecture/l_space.md
  9. docs/architecture/control_plane.md
  10. docs/architecture/residue_geometry.md
  11. docs/architecture/default_mode.md
  12. docs/architecture/hippocampal_braid.md
  13. docs/architecture/entities_and_binding.md
  14. docs/claims/claims.yaml
  15. docs/claims/claim_index.md

Modified Files (0)

  • No existing files modified (preservation principle)

Archived Files (0)

  • No files archived yet (all content preserved in original locations)

Detected Conflicts

Count: 0

No formal conflicts detected during consistency analysis. All claims are compatible within their dependency structure.

Potential Semantic Variations (For Future Review)

  1. E1 Implementation Details
    • DANIEL_README.md describes E1 as “predictive field, not deep network; shallow, recurrent, multi-rate”
    • This detail not prominently featured in canonical e1.md
    • Action: Consider adding as MECH claim or architectural note
  2. Self Definition Prominence
    • Strong operational definition in REE_CORE.md: “Self is currently committed trajectory prefix”
    • Should be elevated to architectural commitment or invariant
    • Action: Add as ARC-010 or INV-018
  3. Phase Compatibility Mechanism
    • depth_phase_spec.md provides detailed phase-coordinate system
    • precision_scoping.md provides τ-scoped precision rules
    • Both are mechanism hypotheses (MECH-001, MECH-002)
    • Action: Ensure clear status as candidate mechanisms, not committed architecture

Classification Uncertainties

Items Flagged for Review

  1. Entities and Binding (ARC-009)
    • Status: Provisional
    • Reason: Extracted from DANIEL_README.md Layer 3; needs fuller specification
    • Recommendation: Expand with detailed emergence mechanisms and binding constraints
  2. Phase Compatibility (MECH-001)
    • Status: Candidate
    • Reason: Detailed specification exists but marked as hypothesis
    • Recommendation: Test and validate before promoting to architectural commitment
  3. τ-Scoped Precision (MECH-002)
    • Status: Candidate
    • Reason: Strong normative language (“MUST”) but still marked as hypothesis
    • Recommendation: Clarify if this is mechanism hypothesis or architectural requirement

Items Successfully Classified

All 17 invariants have clear, conservative classification:

  • No borderline cases
  • No conflicting invariants requiring downgrade
  • All dependencies valid

All 9 architectural commitments have stable or provisional status:

  • 8 stable
  • 1 provisional (entities_and_binding)

Statistics

  • Total Documentation Files: 56+ markdown files
  • Canonical Architecture Files: 9
  • Total Claims: 28
    • Invariants: 17 (60.7%)
    • Architectural Commitments: 9 (32.1%)
    • Mechanism Hypotheses: 2 (7.2%)
  • Claim Statuses:
    • Active: 17 (invariants)
    • Stable: 8 (architectural commitments)
    • Provisional: 1 (entities_and_binding)
    • Candidate: 2 (mechanism hypotheses)
  • Dependency Relationships: 35+ documented dependencies
  • Circular Dependencies: 0
  • Documented Conflicts: 0

Principles Followed

All global rules were respected:

  1. ✓ No content deleted
  2. ✓ Original wording preserved where possible
  3. ✓ No silent meaning changes
  4. ✓ No silent claim type changes
  5. ✓ No contradictions resolved by choosing “winner”
  6. ✓ Uncertain classifications marked as provisional/candidate
  7. ✓ Invariants treated conservatively

Next Steps (Deferred)

The following items were identified but deferred for future work:

  1. Decide promotion/demotion for remaining candidate claims (IMPL-008, IMPL-016, MECH-001..MECH-030).
  2. Expand docs/architecture/entities_and_binding.md with emergence/binding mechanisms and resolve Q-001.
  3. Add Self operational definition as a formal claim and link dependencies (candidate ARC or INV).
  4. Formalize the E1 implementation constraint from DANIEL_README.md as a mechanism hypothesis.
  5. Review and consolidate duplicate content across supplementary files and legacy references.
  6. Decide whether to split mechanism hypotheses into a dedicated directory if the count continues to grow.

Conclusion

This refactoring establishes a solid foundation for REE documentation:

  • Clear dependency structure
  • Explicit claim typing
  • No hidden conflicts
  • Preserved content safety
  • Machine-readable registry
  • Human-readable navigation

The documentation is now ready for:

  • Systematic evolution via the claim registry
  • Safe refinement following DANIEL_README.md process
  • Falsifiable architectural testing
  • Multi-contributor collaboration with clear change boundaries

2026-02-14: REE-v2 Qualification Routing Cutover

Overview

Adopted adjudicated-divergence cutover policy and switched default experimental routing to ree-v2 after all readiness gates passed.

What Changed

  • Updated planning routing default:
    • evidence/planning/planning_criteria.v1.yaml (experimental_default_repo=ree-v2)
  • Generated cutover readiness artifacts:
    • evidence/planning/CUTOVER_REE_V2_READINESS.md
    • evidence/planning/CUTOVER_REE_V2_READINESS.v1.json

2026-02-15: Focused Dependency Retune (Commit Boundary + Tri-Loop + Control Axes)

Overview

Performed a focused dependency and claim-consistency retune for the commit-boundary/gating/control-axis bundle and aligned hook-surface dependencies with the new bridge commitments.

What Changed

  • Status alignment:
    • docs/architecture/e3.md: MECH-061 and MECH-062 status headers set to candidate to match registry.
    • docs/architecture/control_plane.md: MECH-063 status header set to candidate to match registry.
  • Hook-surface dependency tightening:
    • docs/architecture/hook_surface_contract.md now explicitly depends on MECH-061, MECH-062, MECH-063.
    • Added explicit v2_required bridge hook families for:
      • commit-boundary token envelope fields,
      • tri-loop arbitration traces,
      • orthogonal tonic/phasic control-axis telemetry.
  • Claim registry/index updates:
    • docs/claims/claims.yaml: IMPL-025.depends_on extended with MECH-061, MECH-062, MECH-063.
    • docs/claims/claim_index.md: IMPL-025 summary updated to mention bridge families explicitly.
  • Audit note added:
    • docs/notes/2026-02-15_dependency_retune_audit.md

Validation

  • Claim graph check passed:
    • total claims: 141
    • missing dependencies: 0
    • missing location/anchor: 0
    • cycles: 0

2026-02-15: Sensory-to-Learning Loop Conformance Retune

Overview

Ran a strict sensory-origin information-flow conformance pass and aligned v2 hook registry obligations with the bridge families already required by the hook-surface contract.

What Changed

  • Hook registry updates:
    • docs/architecture/hook_registry.v1.json
    • Added v2_required bridge hooks:
      • HK-007 commit_boundary_token_export (MECH-061 bridge)
      • HK-008 tri_loop_gate_trace_export (MECH-062 bridge)
      • HK-009 control_axis_telemetry_export (MECH-063 bridge)
  • v2 spec alignment:
    • docs/architecture/ree_v2_spec.md
    • Added explicit v2_required bridge-hook expectations in the cross-version hook section.
  • Audit note:
    • docs/notes/2026-02-15_sensory_loop_conformance_audit.md
    • Captures resolved mismatch and residual interface gaps for next pass.

Residual Gaps Logged

  • No dedicated payload schema doc yet for HK-007/008/009 envelopes.
  • Adapter-signal schema does not yet enforce sensory-tag provenance/quality fields for ARC-017.
  • No direct readiness metric yet for commit-boundary join completeness across post-commit attribution updates.

2026-02-15: Weekly Cadence and Full-Run Sync Hardening

Overview

Hardened handoff-sync automation so cross-repo syncing of ree-v2, ree-v1-minimal, and ree-experiments-lab is guaranteed both in weekly cadence and via explicit full-run invocation.

What Changed

  • Updated cadence config:
    • evidence/planning/cadence_automation.v1.json
    • THURSDAY now includes all three producer repos.
    • Added full_run_repos list for explicit full-run target set.
  • Extended sync script:
    • evidence/planning/scripts/sync_weekly_handoffs.py
    • Added --full-run flag to select full_run_repos from config.
  • Updated operator docs:
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md now uses --full-run --run-ingestion for Thursday/full sync.
    • evidence/planning/README.md now documents on-demand full-run command.

2026-02-15: Governance Agenda Carryover Persistence

Overview

Added a persistent manual carryover queue so unfinished governance items survive agenda regeneration across full-run and weekly cadence cycles.

What Changed

  • Added carryover source file:
    • evidence/planning/manual_carryover_items.v1.json
  • Updated governance cycle script:
    • evidence/planning/scripts/run_governance_cycle.py
    • loads manual_carryover_items.v1.json when present,
    • includes all items with status != done in agenda checkpoint manual_carryover,
    • adds summary counts for total/open carryover items.
  • Updated planning docs:
    • evidence/planning/README.md
    • evidence/planning/LOCAL_CADENCE_AUTOMATION.md
  • Added quick-capture helper:
    • evidence/planning/scripts/capture_carryover_item.py
    • supports add, list, and done actions for carryover tasks.

2026-02-18: Conflict Resolution Pass (Q-008, Q-009, Q-011)

Overview

Resolved three explicitly documented architecture conflicts by clarifying separation boundaries and legal update loci, while preserving historical question IDs as legacy entries.

What Changed

  • Added resolution notes:
    • docs/conflicts/resolutions/2026-02-18_valence-vs-mu-kappa.md
    • docs/conflicts/resolutions/2026-02-18_care-override-vs-other-harm-veto.md
    • docs/conflicts/resolutions/2026-02-18_rollout-entropy-floor-vs-residue-persistence.md
  • Updated conflict records to resolved status:
    • docs/conflicts/valence_vectors_vs_mu_kappa_overlays.md
    • docs/conflicts/care_override_vs_other_harm_veto.md
    • docs/conflicts/rollout_entropy_floor_vs_residue_persistence.md
    • docs/conflicts/README.md
    • docs/conflicts/resolutions/README.md
  • Updated canonical architecture text:
    • docs/architecture/control_plane.md (Q-008 marked legacy with resolution pointer)
    • docs/architecture/social.md (Q-009 marked legacy with bounded override policy)
    • docs/architecture/hippocampal_systems.md (Q-011 marked legacy with update-locus separation)
  • Updated claim registry/index:
    • docs/claims/claims.yaml (Q-008, Q-009, Q-011 status set to legacy)
    • docs/claims/claim_index.md legacy summaries updated for Q-008, Q-009, Q-011

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