diff --git a/.claude/board/EPIPHANIES.md b/.claude/board/EPIPHANIES.md
index d7811bf79..853b1e8d7 100644
--- a/.claude/board/EPIPHANIES.md
+++ b/.claude/board/EPIPHANIES.md
@@ -1,3 +1,35 @@
+## 2026-09-02 — E-A-GHOST-TRACE-IS-NOT-THE-COUNTERFACTUAL-LANE-1 — the word "ghost" names two semantic families in the tree; the same-word-≠-same-family lesson of the September teardown, caught BEFORE a port this time
+
+**Status:** FINDING + FENCE (verified against source). **Confidence:** High. Raised by an external review of #1137 (GPT-class; held CLAIMED-BY-EXTERNAL per `E-EXTERNAL-REVIEW-ADJUDICATED-1` until checked, then confirmed on both sides in code).
+
+**The two families.**
+1. **Lingering trace.** `thinking_engine::ghosts::{GhostType, Ghost, GhostField}` — a `Ghost` carries `atom`, `ghost_type`, `intensity`, `created_at`, `source_text`; the field decays asymptotically and biases the NEXT cascade (Friston prior). Contract twin `escalation::{GhostEcho, WisdomMarker}` (floor 0.1, decay 0.85). Its eight names are memory residues: Staunen = persistent wonder, Wisdom = deep / harvested knowing, Grief = loss that reshapes topology, Boundary = a limit still felt.
+2. **Non-authoritative counterfactual rung.** `contract::counterfactual::{deposit_counterfactual, CounterfactualMailbox, revise_if_minority_wins}` — the minority pole of a split quorum, kept under `InferenceType::Counterfactual` (−6), never written as observed SPO truth. Its own docs call it a "ghost-tier mailbox" seven times (`counterfactual.rs:16, 111, 229, 243, 247, 340, 402`).
+
+**Why it matters now.** `thinking-engine-harvest-closure-v1` D-TEH-2 ports the trace family into the planner, and `house-differential-style-v1` D-HOUSE-4 consumes it as an anchoring alarm. A reader who meets `GhostType::Wisdom` beside a "ghost-tier mailbox" will plausibly conclude the existing `GhostField` already IS the counterfactual-rung carrier. It is not: one is what a past thought left behind, the other is a road not taken that may still win. The September teardown removed a register precisely because evocative shape-sharing (G24N4) had been read as semantic identity; this is the same failure one word earlier.
+
+**The fence.** The rung MAY consume a trace (a Staunen or Wisdom marker) as a starting prior. The rung is NEVER a trace, a trace is NEVER a rung, and no port under D-TEH-2 or readout under D-HOUSE-1..4 reads `WisdomMarker` as counterfactual state or writes the −6 lane from a ghost field. The wording collision in `counterfactual.rs` is recorded as `TD-GHOST-TIER-NAME-COLLISION-1` (doc-only; rename "ghost-tier" → "counterfactual-tier" when that file is next touched, never as a drive-by).
+
+**Archaeology kept.** The older architecture already carried the epistemic cycle as a DUALITY: Staunen keeps a question alive, Wisdom is what was harvested from it. The shipped `curiosity() = (1/(q+1))·(1−c)` (`exploration.rs`) is a later operational slice of the Staunen end, not the pair. Recorded in the closure plan's idea harvest so the duality survives the crate.
+
+**Also adopted from the same review:** D-HOUSE-1 stays surgical (one causal question, four arms); the resonance, qualia-dye and aperture arms folded in by `575c6d3` are moved OUT to gated follow-on probes D-HOUSE-1b/1c/1d (plan §4b), with three processing-order defects fixed in the move (the aperture knob was reading a quantity that does not exist until two steps later; the dye coloured one candidate; the resonance arm was one shape where the pyramid names two).
+
+Refs: `E-SIX-SEMANTIC-FAMILIES-MUST-NOT-IMPERSONATE-EACH-OTHER-1`, `E-EXTERNAL-REVIEW-ADJUDICATED-1`, `TD-GHOST-ECHO-DUP-1`, `E-QUALIA-IS-RANK-INERT-AT-THE-FRONTIER-AND-POPULATION-LOSES-TO-COUNTING-1` (the attribution lesson), plans `house-differential-style-v1` §4b, `thinking-engine-harvest-closure-v1` §1c/§4.
+
+## 2026-09-02 — E-THINKING-ENGINE-LIVE-FOOTPRINT-IS-ONE-TRAIT-AND-HOUSE-IS-SHIPPED-IN-PIECES-1 — the 51-file excluded crate binds the spine through ONE trait module; the differential-diagnosis style the operator asked for already exists as seven scattered shipped mechanisms and zero names
+
+**Status:** FINDING (measured on `94543a5`, read-only). **Confidence:** High for both measurements. Plans: `house-differential-style-v1`, `thinking-engine-harvest-closure-v1`.
+
+**Measurement 1 — the crate's live footprint.** `grep thinking-engine crates/*/Cargo.toml` names two consumers. `lance-graph-callcenter` requires it and uses exactly `bridge_gate::{CognitiveBridgeGate, CognitiveAuthResult, …}` — a trait-only module whose own doc says it lives low so callcenter can implement it. `cognitive-shader-driver` takes it OPTIONALLY behind `with-engine` (default off) and reaches `dto::BusDto`, two `cognitive_stack` thresholds, and lens lookups in doc/test paths. Everything else in the 51 files is internal or example-only. Consequence: the chapter closes by moving one trait to the zero-dep contract and ruling the DTO ladder's home under D-TTV-1 — not by a rewrite. The MODULE-TABLE census's gem-status column, COMPONENT-MAP §3 and three TD rows already agreed on the file-level verdicts; nobody had measured the dependency edge, so the closure kept looking larger than it is.
+
+**Measurement 2 — the House question.** A grep for the pattern's name (and for "differential diagnosis") over `.claude/`, thinking-engine, contract and planner returns nothing. The mechanics are shipped in seven places: `rcr_abduce` (the ranked board), `asc_challenge` with the disjoint-stamp guard (attack the leader), `cr_synthesize` (thesis/antithesis, depth preserved), `InnerCouncil` + `quorum_project` + `deposit_counterfactual` (minority kept as −6), `peripheral_sample_where` + `peripheral_dissent` (the test nobody ordered; never decides), `Recipe::min_rung` (cheap passes first), `WisdomMarker`/`GhostEcho` (memory, without the field). MedCare-rs already carries the consumer shape (`differential() -> Vec<DiffRow>`, private repo). What is missing is not a mechanism but two READOUTS (discriminating evidence; elimination as a predicate over `truth`/`contradiction`/`stamp`) and the O2 style→recipe edge. Consequence: the right first artifact is a falsifier, not a struct — PROBE-HOUSE-DIFFERENTIAL-1 (planted-cause recovery, shuffle null, periphery can-fire AND can-stay-silent).
+
+**Two stale rows found on the way.** D-PERSONA-1 still read "In progress" on a branch name; its content (`contract::escalation`, planner `mul/escalation.rs`) is on main. D-REUNIFY-2 still read "Backlog"; it shipped as D-CSV-9 in #387. Both regraded in the same commit, status cells only.
+
+**Rule extracted.** Before sizing a retirement, measure the DEPENDENCY EDGE (which symbols cross the crate boundary), not the file count. A 51-file crate with a one-trait edge is a one-trait retirement plus a harvest list; the harvest list is then governed by consumers and falsifiers, one port at a time.
+
+Refs: `TD-THINKING-ENGINE-EXCLUDED-DEBT-1`, `TD-GHOST-ECHO-DUP-1`, `E-RUNG-ASCENT-WIRED-1` (the crate named orphaned), `E-PERIPHERAL-DISSENT-GUARDS-THE-STRATIFICATION-1`, `E-DIALECTIC-V1-TACTICS-IN-PLANNER-1`, `persona-vs-rung-ladder.md` O2/O3.
+
 ## 2026-09-02 — E-QUALIA-IS-RANK-INERT-AT-THE-FRONTIER-AND-POPULATION-LOSES-TO-COUNTING-1
 
 **Status:** MEASURED (PROBE-POP-READOUT-1 / D-POP-1, `pop_readout.rs`, whole
diff --git a/.claude/board/INTEGRATION_PLANS.md b/.claude/board/INTEGRATION_PLANS.md
index ab2242487..c9fd306ac 100644
--- a/.claude/board/INTEGRATION_PLANS.md
+++ b/.claude/board/INTEGRATION_PLANS.md
@@ -1,3 +1,34 @@
+## 2026-09-02 — `house-differential-style-v1` (PROPOSAL, plan-only) + `thinking-engine-harvest-closure-v1` (PROPOSAL, plan-only)
+
+`.claude/plans/house-differential-style-v1.md`. A differential-diagnosis
+style program named as a COMPOSITION of shipped rung-3 tactics — RCR
+frontier, periphery watchers (`peripheral_sample_where` +
+`peripheral_dissent`), ASC challenge with the disjoint-stamp guard, CR
+synthesis, council split → −6 counterfactual fork — plus two readouts it
+asks for (discriminating evidence; elimination as a predicate over arena
+entries, never a new state) and an anchoring alarm gated on the ghost-prior
+harvest. Nothing in the parts list is modified. Gate: PROBE-HOUSE-
+DIFFERENTIAL-1 (D-HOUSE-1), planted-cause recovery on synthetic arenas with
+shared-predicate distractors, size-preserving shuffle null, can-fire AND
+can-stay-silent bounds on the periphery. No opcode, tenant, ClassView, axis,
+struct, persona wiring, or clinical content. The working name never enters
+code.
+
+`.claude/plans/thinking-engine-harvest-closure-v1.md`. Fate of every one of
+the 51 files in the workspace-excluded crate (HARVESTED / HOT-VIA-FEATURE /
+GEM→landing / LAB / RESIDUE), built on the MODULE-TABLE census and
+COMPONENT-MAP §3. Measured live footprint: one required consumer
+(`lance-graph-callcenter`, ONE trait module `bridge_gate`) and one optional
+(`cognitive-shader-driver` `with-engine`: the DTO ladder + two thresholds).
+Reconciles every open row still pointing at the crate (D-PERSONA-1 stale →
+Shipped; D-PERSONA-5 proposed RETIRE under E-NOBODY-WAITS-1; D-REUNIFY-2
+already shipped as D-CSV-9; D-REUNIFY-4 done in `think.rs`; three TDs
+paid by named waves). Closure in four waves — cut the hard dependency,
+harvest gems WITH consumers and falsifiers, M8 engine collapse, retire and
+rename to a lab crate — with stop rules (no port without a consumer, no
+singleton field, no handoff-type revival, no asserted axis basis). §4 keeps
+the chapter's ideas as doctrine with named homes.
+
 ## 2026-09-02 — `post-teardown-buildup-survey-v1` (SURVEY, plan-only)
 
 `.claude/plans/post-teardown-buildup-survey-v1.md`. First step after the
diff --git a/.claude/board/PR_ARC_INVENTORY.md b/.claude/board/PR_ARC_INVENTORY.md
index 79e7218e3..303066469 100644
--- a/.claude/board/PR_ARC_INVENTORY.md
+++ b/.claude/board/PR_ARC_INVENTORY.md
@@ -10,6 +10,52 @@
 > census §8.3 trap 10: read the body FIRST, then open for write — never
 > inline both in one expression.
 
+## 2026-09-02 — lance-graph branch `claude/medcare-rs-continue-6nhbxn` (plans PR, after #1136) — House differential style v1 + thinking-engine harvest & closure v1
+
+- **Added:** `.claude/plans/house-differential-style-v1.md` (223 lines: parts
+  list of shipped surfaces file-cited, the composition as a recipe-level
+  program, ladder placement per `persona-vs-rung-ladder.md`, PROBE-HOUSE-
+  DIFFERENTIAL-1 pre-registered, D-HOUSE-0..6);
+  `.claude/plans/thinking-engine-harvest-closure-v1.md` (206 lines: measured
+  live footprint, 51-file fate table, open-row reconciliation, four-wave
+  closure with stop rules, idea harvest, D-TEH-0..5, four rulings asked);
+  INTEGRATION_PLANS entry; STATUS_BOARD sections for both plans; EPIPHANIES
+  entry `E-THINKING-ENGINE-LIVE-FOOTPRINT-IS-ONE-TRAIT-AND-HOUSE-IS-SHIPPED-IN-PIECES-1`;
+  SUPERSESSION-INDEX regenerated last.
+- **Regraded (status cells only, evidence cited):** D-PERSONA-1 In progress →
+  Shipped (`contract::escalation` + planner `mul/escalation.rs` on main);
+  D-PERSONA-5 annotated "proposed RETIRE, ruling needed"; D-REUNIFY-2
+  annotated "shipped as D-CSV-9 (#387)".
+- **Locked:** nothing — plan-only; no code, no Cargo change, no type.
+- **Deferred:** every code wave (D-HOUSE-1, D-TEH-1..5) behind its gate; four
+  rulings listed in the closure plan §6.
+- **Confidence:** High for the footprint measurement (two `Cargo.toml`
+  consumers, symbols grepped) and the parts list (each row file:line);
+  Medium for the fate verdicts that name a landing without a consumer yet
+  (they are marked "park" rather than "port").
+- **Correction 2026-09-02 (same PR, third commit):** `575c6d3` had folded three
+  extra arms into D-HOUSE-1; an external review objected (attribution — one
+  causal question per harness) and the objection was adopted: arms moved out
+  to gated D-HOUSE-1b/1c/1d (plan §4b) with three processing-order defects
+  fixed; ghost-trace vs counterfactual-lane fence banked
+  (`E-A-GHOST-TRACE-IS-NOT-THE-COUNTERFACTUAL-LANE-1`,
+  `TD-GHOST-TIER-NAME-COLLISION-1`).
+- **Review pass 2026-09-02 (fourth commit):** 6 Codex + 8 CodeRabbit findings
+  on `84f2885`, all verified in source, all fixed. The three Codex P1s were
+  real probe-design errors: the fixture direction meant `rcr_abduce`
+  (`{P→M, S→M} ⊢ S→P`) could never emit the cause as a candidate (fixed:
+  `cause→feature` rules + `case→feature` observations); candidates were
+  never admitted before ASC (`admit_derived` step added, else
+  `AscOutcome::NoTarget`); `quorum_project` is an unconditional `todo!`
+  (removed from the path; council verdict → `deposit_counterfactual`
+  directly). Also: A1c control arm for the `k = 0` guard; swap-only,
+  `None`-defined discriminating-evidence readout; role_tables/jina_lens
+  one fate each; bridge_gate migration symbol-complete (7 items) with a
+  W1 compatibility contract incl. callcenter's lens-touching test; ghost
+  floor mismatch (0.001-prune vs 0.1-clamp) declared an intentional change
+  with a calibration gate; M8 gate = tolerances + invariants, not
+  bit-parity; stop rule scoped to §1b+§1c; generator singular agreement.
+
 ## 2026-09-02 — lance-graph branch `claude/medcare-rs-continue-6nhbxn` (D-POP-1 result PR, after #1135) — PROBE-POP-READOUT-1
 
 - **Added:** `crates/deepnsm-v2/examples/pop_readout.rs` (probe only, no library
diff --git a/.claude/board/STATUS_BOARD.md b/.claude/board/STATUS_BOARD.md
index adaf69dbe..12ba5d7c3 100644
--- a/.claude/board/STATUS_BOARD.md
+++ b/.claude/board/STATUS_BOARD.md
@@ -1,3 +1,29 @@
+## house-differential-style-v1
+
+| D-id | deliverable | status |
+|---|---|---|
+| D-HOUSE-0 | the plan: parts list of shipped surfaces (file-cited), the composition as a recipe-level style program, ladder placement, pre-registered falsifier | **Shipped (plan-only, this PR)** |
+| D-HOUSE-1 | PROBE-HOUSE-DIFFERENTIAL-1: planted-cause recovery on synthetic arenas with shared-predicate distractors; arms RCR-alone / +admit+ASC / +council control / full cycle / shuffle null (fixture direction `cause→feature` + `case→feature` so RCR yields `case→cause`; `quorum_project` NOT on the path — it is a `todo!`); PASS iff Δp@1 ≥ 0.05 above null p95, planted cause never eliminated, periphery fires on 10–90 % | **Queued — next step** |
+| D-HOUSE-1b | resonance prior probe (`res-seed` + `res-gate`): anchoring vs recognition, same arenas, ρ < 1 guard against the D-HOUSE-1 A2 ordering (plan §4b) | Gated on D-HOUSE-1 reported |
+| D-HOUSE-1c | signed-qualia contrast probe: one i4 per top-3 candidate from the ASC outcome into existing `QualiaI4_16D` lanes, post-challenge re-rank only, all-zero-signs must equal A2 (plan §4b) | Gated on D-HOUSE-1 reported |
+| D-HOUSE-1d | aperture hypothesis (the "humor" label): incongruity → `k` across passes; a hypothesis until its operator is defined and measured (plan §4b) | Not scheduled |
+| D-HOUSE-2 | discriminating-evidence readout (clone arena, observe hypothetical premise, report largest rank displacement) | Gated on D-HOUSE-1 PASS |
+| D-HOUSE-3 | elimination predicate over `truth`/`contradiction`/`stamp` + never-re-enter falsifier | Gated on D-HOUSE-1 PASS |
+| D-HOUSE-4 | anchoring alarm (free energy between wisdom-marker prior and evidence) | Gated on `thinking-engine-harvest-closure-v1` D-TEH-2 |
+| D-HOUSE-5 | the style→recipe edge (O2) with House as its first program | Gated on D-HOUSE-1 PASS and the D-TSC-2/3 mint |
+| D-HOUSE-6 | consumer pointer: `DiffRow` READ fields `eliminated` / `discriminator` (MedCare-rs) | pointer only, out of scope here |
+
+## thinking-engine-harvest-closure-v1
+
+| D-id | deliverable | status |
+|---|---|---|
+| D-TEH-0 | census: live footprint measured (one required consumer via `bridge_gate`; one optional via `with-engine`), 51-file fate table, open-row reconciliation, four-wave closure, idea harvest, four rulings asked | **Shipped (plan-only, this PR)** |
+| D-TEH-1 | W1: `bridge_gate` trait → `lance-graph-contract`; callcenter drops the path dep; DTO-ladder home ruling recorded (D-TTV-1); thinking-engine becomes a leaf | Queued — first code wave |
+| D-TEH-2 | W2: ghost prior → planner `nars/ghost_prior.rs` over `WisdomMarker`, per-thought, two-sided falsifiers; crate `ghosts.rs` deleted | Queued — consumer D-HOUSE-4 |
+| D-TEH-3 | W2: `cronbach` → `jc::reliability`; `semantic_chunker` / `spiral_segment` decided by their falsifiers | Queued |
+| D-TEH-4 | W3: ENTROPY M8 engine collapse with dtype parity suite; 5 cascade shapes + 3 lens modules collapse | Queued |
+| D-TEH-5 | W4: residue deleted, crate renamed `thinking-lab` with a `--manifest-path` CI line; §2 rows closed; `TD-THINKING-ENGINE-EXCLUDED-DEBT-1` paid | Queued — closes the chapter |
+
 ## post-teardown-buildup-survey-v1
 
 | D-id | deliverable | status |
@@ -1343,7 +1369,7 @@ Plan path: `.claude/plans/causaledge64-mailbox-rename-soa-v1.md`.
 | D-id | Title | Status | PR / Evidence |
 |---|---|---|---|
 | D-REUNIFY-1 | Acknowledge dual `CausalEdge64` types in TYPE_DUPLICATION_MAP + LATEST_STATE + EPIPHANIES | **Shipped** | this commit (post-merge #372 board-hygiene tail) |
-| D-REUNIFY-2 | 8-channel → SPO transcoder spec at thinking-engine L3 commit boundary | **Backlog** | per Option R-3; sprint-12+ |
+| D-REUNIFY-2 | 8-channel → SPO transcoder spec at thinking-engine L3 commit boundary | **Backlog → shipped as D-CSV-9** (#387 `e042c70`; annotated 2026-09-02, `thinking-engine-harvest-closure-v1` §2) | per Option R-3; sprint-12+ |
 | D-REUNIFY-3 | `Think` carrier struct prototype unifying thinking-engine cascade + cognitive-shader-driver SoA | **Backlog** | per `.claude/knowledge/splat-shader-rayon-struct-method-vision.md` sprint-12 |
 | D-REUNIFY-4 | Splat op fleet (`splat_gaussian`, `score_hole_closure`, `replay_coherence`, `emit_if_epiphany`) as methods on `Think` | **Backlog** | sprint-13+ |
 | D-REUNIFY-5 | rayon work-stealing par_* method variants | **Backlog** | sprint-14+ |
@@ -1440,11 +1466,11 @@ hot/cold/feedback loop onto our contract types + SoA floor
 
 | D-id | Title | Crate(s) | ~LOC | Risk | Status | PR / Evidence |
 |---|---|---|---|---|---|---|
-| D-PERSONA-1 | escalation+epiphany loop = the checklist (`felt_parse` collapse-hint + `InnerCouncil`/`HdrResonance` split + `EpiphanyDetector`; green-flip = Epiphany/Wisdom ghost) — NOT a bespoke verifier | contract + planner | 160 | LOW | **In progress** | branch `claude/splat3d-cpu-simd-renderer-MAOO0` |
+| D-PERSONA-1 | escalation+epiphany loop = the checklist (`felt_parse` collapse-hint + `InnerCouncil`/`HdrResonance` split + `EpiphanyDetector`; green-flip = Epiphany/Wisdom ghost) — NOT a bespoke verifier | contract + planner | 160 | LOW | **Shipped** (regraded 2026-09-02 by `thinking-engine-harvest-closure-v1` §2: `contract::escalation` {CollapseHint, InnerCouncil, EpiphanyDetector, GhostEcho, WisdomMarker, Checklist} + planner `mul/escalation.rs` {`verdict_from`, `boot_checklist`} are on main; the old branch pointer was stale) | — |
 | D-PERSONA-2 | meta-recipe manifest (declarative child-spec, recipe-as-data, macro-evaluable) | contract | 150 | MED | **Queued** | — |
 | D-PERSONA-3 | hot/cold/feedback wiring — anneal + `CrystalCodebook`→wisdom-marker cold path + Preload hydrate | planner + Lance | 240 | MED | **Queued** | — |
 | D-PERSONA-4 | macro-eval harness (scenario→trace→discover→diagnose; suspect-bridge = blasgraph betweenness; 5 rubrics from D-RUNG-MUL) | planner + Lance | 280 | HIGH | **Queued** | — |
-| D-PERSONA-5 | ractor outer-swarm runtime under `OrchestrationBridge` (batons as messages, async only at boundary) | planner | 200 | MED | **Queued** | — |
+| D-PERSONA-5 | ractor outer-swarm runtime under `OrchestrationBridge` (batons as messages, async only at boundary) | planner | 200 | MED | **Queued — proposed RETIRE** (`thinking-engine-harvest-closure-v1` §2/§6: batons-as-messages contradicts E-NOBODY-WAITS-1 and E-ACK-ELIMINATED-1; ruling needed, not assumed) | — |
 | D-PERSONA-6 | `odoo_scanner` + `OdooBridge` — harvest Odoo `l10n_de` → Finance-ns `MappingProposal`s; bind existing `TaxEngine`; GoBD by construction | ontology + contract + planner | 280 | MED | **Queued** | — |
 
 ---
diff --git a/.claude/board/SUPERSESSION-INDEX.md b/.claude/board/SUPERSESSION-INDEX.md
index 9d51051aa..6806fe13c 100644
--- a/.claude/board/SUPERSESSION-INDEX.md
+++ b/.claude/board/SUPERSESSION-INDEX.md
@@ -17,11 +17,11 @@
 most-referenced symbol here: **68 crate files, 44 plans, 41 of them blind.**
 That is a programme, not a cleanup.
 
-**`CollapseGateEmission` is the sharpest case: 12 plans name it and *every one* is blind.**
-Its COMPONENT-MAP note reads: already tombstoned (PR #477); comment-only remains
+**`GateState` is the sharpest case: 1 plan names it and *every one* is blind.**
+Its COMPONENT-MAP note reads: intra-cascade SD gate — fine IF intra-mailbox; warden sign-off queued, not assumed
 
 **`ResonanceDto` → `PerturbationDto` gives the rule that needs no map at all:**
-2 crate files against 10 plans. The code moved; the plans did not.
+2 crate files against 12 plans. The code moved; the plans did not.
 **Plan-mentions exceeding crate-mentions is a staleness signal on its own.**
 
 ### The limit of the mechanical route
@@ -49,9 +49,9 @@ coverage instead.
 
 | symbol | verdict | successor | live in crates | named in plans | blind plans |
 |---|---|---|---|---|---|
-| `A2AMessage` | BLOCKED | — | 2 | 0 | 0 |
-| `StepMask` | BLOCKED | — | 3 | 8 | 4 |
-| `commit_to_l4` | BLOCKED | — | 2 | 1 | 0 |
+| `A2AMessage` | BLOCKED | — | 2 | 1 | 0 |
+| `StepMask` | BLOCKED | — | 3 | 9 | 4 |
+| `commit_to_l4` | BLOCKED | — | 2 | 2 | 0 |
 | `dispatch_busdto` | BLOCKED | — | 3 | 6 | 5 |
 | `persist_cycle` | BLOCKED | — | 10 | 6 | 5 |
 | `CognitiveMarkers` | REPURPOSE | `Commit` | 2 | 0 | 0 |
@@ -59,10 +59,10 @@ coverage instead.
 | `GateDecision` | REPURPOSE | — | 25 | 27 | 24 |
 | `GateState` | REPURPOSE | — | 14 | 1 | 1 |
 | `MergeMode` | REPURPOSE | — | 8 | 13 | 12 |
-| `ResonanceDto` | REPURPOSE | `PerturbationDto` | 2 | 10 | 7 |
+| `ResonanceDto` | REPURPOSE | `PerturbationDto` | 2 | 12 | 7 |
 | `BindSpace` | RETIRE | — | 68 | 44 | 41 |
-| `CollapseGateEmission` | RETIRE | — | 5 | 12 | 12 |
-| `ThinkingStyle` | RETIRE-toward-contract | — | 51 | 27 | 24 |
+| `CollapseGateEmission` | RETIRE | — | 5 | 13 | 12 |
+| `ThinkingStyle` | RETIRE-toward-contract | — | 51 | 28 | 24 |
 
 ## Table 2 — plans naming a ruled symbol without citing the ruling (73)
 
@@ -98,8 +98,8 @@ a licence to act on it.
 | **READ** | `self-reasoning-substrate-v1` | `GateDecision` | PROPOSED — doc-only. No code, no contract ch | 5/5 |
 | **READ** | `v3-convergence-wiring-v1` | `GateDecision` | ACTIVE (2026-07-01). Operator: "I'm all in f | 0/0 |
 | **RESCOPE** | `soa-migration-diff-resolution-2026-06-13` | `BindSpace`, `CollapseGateEmission`, `GateDecision`, `MergeMode` … | — | 3/5 |
-| **RESCOPE** | `cognitive-substrate-convergence-v1` | `BindSpace`, `CollapseGateEmission`, `GateDecision`, `MergeMode` … | PROPOSAL (sprint-10 architectural decisions  | 6/13 |
-| **RESCOPE** | `cognitive-substrate-convergence-v2` | `BindSpace`, `CollapseGateEmission`, `GateDecision`, `MergeMode` … | ACTIVE — sprint-11 Phase A/B COMPLETE (pendi | 7/15 |
+| **RESCOPE** | `cognitive-substrate-convergence-v1` | `BindSpace`, `CollapseGateEmission`, `GateDecision`, `MergeMode` … | PROPOSAL (sprint-10 architectural decisions  | 7/13 |
+| **RESCOPE** | `cognitive-substrate-convergence-v2` | `BindSpace`, `CollapseGateEmission`, `GateDecision`, `MergeMode` … | ACTIVE — sprint-11 Phase A/B COMPLETE (pendi | 8/15 |
 | **RESCOPE** | `bindspace-singleton-to-mailbox-soa-v1` | `BindSpace`, `CollapseGateEmission`, `ResonanceDto`, `ThinkingStyle` | CONJECTURE / design (migration spec). NOT ye | 18/19 |
 | **RESCOPE** | `callcenter-membrane-v1` | `BindSpace`, `GateDecision`, `MergeMode`, `ThinkingStyle` | Active | 0/0 |
 | **RESCOPE** | `causaledge64-mailbox-rename-soa-v1` | `BindSpace`, `GateDecision`, `MergeMode`, `ThinkingStyle` | Active (draft, 2026-05-14) | 1/10 |
@@ -110,7 +110,7 @@ a licence to act on it.
 | **RESCOPE** | `alpha-reason-witness-shader-field-archaeology-pass-1` | `BindSpace`, `MergeMode`, `ResonanceDto` | SOURCE AUDIT / PLAN ONLY. No production wiri | 1/1 |
 | **RESCOPE** | `bindspace-mailbox-soa-dependency-map-v1` | `BindSpace`, `dispatch_busdto`, `persist_cycle` | MAP / preflight. No source wired yet. Read-b | 2/2 |
 | **RESCOPE** | `bindspace-mailbox-soa-w3-w4a-impl-v1` | `BindSpace`, `dispatch_busdto`, `persist_cycle` | v2 — 5-consolidation + 3-brutal-critic pass  | 1/1 |
-| **RESCOPE** | `cognitive-substrate-convergence-v3` | `BindSpace`, `CollapseGateEmission`, `MergeMode` | ACTIVE — sprint-12 Wave F + Wave G complete  | 7/16 |
+| **RESCOPE** | `cognitive-substrate-convergence-v3` | `BindSpace`, `CollapseGateEmission`, `MergeMode` | ACTIVE — sprint-12 Wave F + Wave G complete  | 8/16 |
 | **RESCOPE** | `mailbox-cycle-aware-write-contract-v1` | `BindSpace`, `dispatch_busdto`, `persist_cycle` | CONJECTURE / design. 5+3-gated before code. | 1/1 |
 | **RESCOPE** | `unified-integration-v1` | `BindSpace`, `MergeMode`, `ThinkingStyle` | Active — brainstorm phase complete; delivera | 0/0 |
 | **RESCOPE** | `2026-05-06-splat-osint-ingestion-v1` | `BindSpace`, `MergeMode` | Active — PR 1+2 of 6 in flight on `claude/sp | 1/7 |
diff --git a/.claude/board/TECH_DEBT.md b/.claude/board/TECH_DEBT.md
index fb5275a72..ac81df884 100644
--- a/.claude/board/TECH_DEBT.md
+++ b/.claude/board/TECH_DEBT.md
@@ -1,3 +1,19 @@
+## TD-GHOST-TIER-NAME-COLLISION-1 (2026-09-02) — OPEN, doc-only
+
+`crates/lance-graph-contract/src/counterfactual.rs` calls the −6 minority-pole
+lane a "ghost-tier mailbox" in seven places (lines 16, 111, 229, 243, 247, 340,
+402). The tree already has a "ghost" family with a different meaning: the
+lingering cognitive trace (`thinking_engine::ghosts::GhostType` / `GhostField`,
+contract `escalation::GhostEcho` / `WisdomMarker`). A reader meeting
+`GhostType::Wisdom` beside a "ghost-tier mailbox" can conclude the trace field
+is the counterfactual carrier; it is not (board:
+`E-A-GHOST-TRACE-IS-NOT-THE-COUNTERFACTUAL-LANE-1`). **Pay by:** rename the
+wording to "counterfactual-tier" in the same PR that next touches
+`counterfactual.rs` (D-ATOM-4 v3 / the mailbox arm) — doc comments only, no
+identifier is named "ghost" in that file, so no API changes. Never as a
+drive-by on an unrelated PR. Surfaced by an external review of #1137,
+verified in source.
+
 ## TD-DEEPNSM-V2-SESSION-RESIDUE (2026-08-22) — OPEN
 
 Three leftovers from the corpus-addressing session, surfaced by its own review
diff --git a/.claude/plans/house-differential-style-v1.md b/.claude/plans/house-differential-style-v1.md
new file mode 100644
index 000000000..a1a0996e0
--- /dev/null
+++ b/.claude/plans/house-differential-style-v1.md
@@ -0,0 +1,321 @@
+# House differential style v1 — a differential-diagnosis style program over shipped rung-3 tactics
+
+> **Status:** PROPOSAL, plan-only (no code, no opcode, no tenant, no ClassView,
+> no axis vocabulary, no new struct). **Baseline:** lance-graph `94543a5`
+> (the #1136 merge). **Date:** 2026-09-02. **Working label:** "House" is a
+> working name for the differential-diagnosis style pattern (rank hypotheses,
+> attack the leader, keep the minority alive, order the test nobody ordered,
+> eliminate only on independent evidence). It names a composition; it never
+> becomes an identifier in code, a classid, or a vocabulary row.
+> **Predecessors:** `post-teardown-buildup-survey-v1` (atoms / recipes /
+> styles framing; PROBE-POP-READOUT-1 KILL), `persona-vs-rung-ladder.md`
+> (rung-content ladder; styles are programs applicable at any rung),
+> `rung-persona-orchestration-v1` (D-PERSONA-1 shipped the council +
+> epiphany loop this plan composes). **Sibling:**
+> `thinking-engine-harvest-closure-v1` (the ghost-prior harvest this plan's
+> D-HOUSE-4 waits on). **Domain fence:** the medical meaning of a differential
+> lives in MedCare-rs (private). This plan carries only the reasoning
+> mechanics; no clinical vocabulary, code, or data enters lance-graph.
+
+## 0. Why a plan and not a struct
+
+A grep for the pattern's name over `.claude/`, `crates/thinking-engine`,
+`crates/lance-graph-contract` and `crates/lance-graph-planner` returns
+nothing. The mechanics are nevertheless shipped, in pieces, across four
+surfaces. Naming the composition is therefore the whole deliverable; the
+first code this plan licenses is a falsifier, not a feature.
+
+```text
+families = state / evidence carriers        (unchanged; six, see survey §1)
+atoms    = readouts over the BeliefArena     (expectation, contradiction, gaps)
+recipes  = the rung-3 tactics RCR / ASC / CR (+ TR, CAS as needed)
+style    = the House program: WHICH recipes fire, in WHAT order, with WHICH
+           periphery budget — a policy, applicable at any rung
+```
+
+## 1. What is already shipped (the parts list, file-cited)
+
+| House move | Shipped surface | Where |
+|---|---|---|
+| The whiteboard: ranked hypotheses from an observation | RCR abduction over a shared predicate → `Frontier { candidates, gaps }`, `Candidate { stmt, truth, premises, rung, tactic }`, `ReasoningGap` | planner `nars/tactics.rs:177 rcr_abduce`, `:71 Candidate`, `:88 ReasoningGap` |
+| Attack the leading hypothesis | ASC: `challenge_target = ⟨1−f, c⟩`; counter-evidence admitted ONLY with a disjoint stamp; self-refutation `BlockedSelfReference` | `tactics.rs:457 challenge_target`, `:479 asc_challenge`, `:463 AscOutcome` |
+| Thesis vs antithesis on one claim | CR: NARS revision, the absolute frequency difference kept as contradiction depth, never averaged | `tactics.rs:503 cr_synthesize` |
+| The team argues; the minority view stays on the board | `InnerCouncil::deliberate`, `is_split(0.7, 0.5)` amplified ×1.2; a split quorum is a Contradiction, never averaged; the minority pole is forked as a −6 counterfactual nibble | contract `escalation.rs:137 InnerCouncil`, `:116 is_split`; `quorum.rs:215 quorum_project`; `counterfactual.rs:140 deposit_counterfactual` |
+| The test nobody ordered (the contrarian) | excluded tactics run as watchers; a watcher that moves the score may force rung elevation; it never decides | contract `recipes.rs:557 peripheral_recipes`, `:573 peripheral_sample`, `:601 peripheral_sample_rotating`, `:641 peripheral_sample_where`; planner `strategy/style_strategy.rs:121 peripheral_dissent`, `:184 cross_family_dissent` |
+| Cheap passes first, the counterfactual leap last | `Recipe::min_rung` / `admissible_at`, `RungLevel::for_pass`; admissible set 4 → 11 → 24 → 34; `ExtremelyHard` gated to `Counterfactual` | `recipes.rs:505–545` |
+| Pattern memory and its failure mode (anchoring) | `GhostField::{imprint, bias, prediction, free_energy}` (excluded crate); contract twin `GhostEcho` + `WisdomMarker` (floor 0.1, decay 0.85) | `thinking-engine/src/ghosts.rs:75–200`; contract `escalation.rs:289 GhostEcho`, `:312 WisdomMarker` |
+| Saturation: when everything is admissible, stop picking a winner | passive `quorum_mantissa` + `meta_basin` outlier SUGGESTIONS | `E-SATURATION-SWITCHES-TO-PASSIVE-QUORUM-1` |
+| The one selection edge that exists today | `tactic_for_bias(GraphBias) -> TacticChoice` — a pure LUT driven by graph health, not by style | planner `nars/tactic_select.rs:77` |
+| The consumer shape | `differential() -> Vec<DiffRow { disease, expectation, frequency, confidence, contradiction }>`, `Frontier::gaps` | MedCare-rs `medcare-cohorts/src/reasoning.rs:363`, `:499` (private repo; cited by path only) |
+
+Everything in the table is consumed as-is. Nothing in it is modified by
+this plan.
+
+## 2. The composition (the style program, recipe level)
+
+One House cycle over a `BeliefArena`, written as the order in which shipped
+tactics fire. Each step names its output atom so the falsifier in §4 can
+measure it.
+
+1. **Board.** `rcr_abduce(arena, throttle)`. Abduction direction matters
+   (Codex P1 on #1137, verified in `tactics.rs:177`): RCR pairs two `Inh`
+   beliefs sharing a PREDICATE, `{P→M (rule), S→M (obs)} ⊢ S→P`. So the
+   arena must carry cause knowledge as `cause→feature` rules and the case as
+   `case→feature` observations; the shared predicate is the FEATURE, and the
+   candidates come out as `case→cause`. Beliefs written the other way round
+   (`feature→cause`) make RCR relate features to each other and the cause
+   never appears as a candidate. The board = the candidates with
+   `stmt.s == case`, ranked by `truth.expectation()` (the ranking key
+   MedCare's `DiffRow` already uses); the throttle's `hub_indegree` must
+   exceed the number of causes sharing a feature, or shared features are
+   barred as hubs. `Frontier.gaps` records why candidates were NOT formed
+   (`NoSharedMiddle`, `HubExcluded`, `BudgetExhausted`, …) — the raw
+   material for step 5.
+1b. **Admit.** `rcr_abduce` does not mutate the arena; a `Candidate` is a
+   proposal. Before any tactic can act on the leader it must be admitted:
+   `BeliefArena::admit_derived` (`belief.rs:226`) for the top-k. Without
+   this step `asc_challenge` returns `AscOutcome::NoTarget` for every
+   leader and the challenge arm is vacuous (Codex P1, verified).
+2. **Periphery.** `RungLevel::peripheral_sample_where(k, eligible)` picks the
+   watchers the current rung excludes; `StyleStrategy::peripheral_dissent`
+   runs them as observers. If one moves the score beyond `tol`, the rung
+   elevates and step 1 re-runs with the wider admissible set. The periphery
+   never votes (E-PERIPHERAL-DISSENT-GUARDS-THE-STRATIFICATION-1).
+3. **Challenge the leader.** For the top candidate, `asc_challenge` with the
+   best available counter-evidence. Only a DISJOINT stamp is admitted; a
+   `BlockedSelfReference` outcome is recorded, not retried with the same
+   sources. For a contested claim carried by two candidates,
+   `cr_synthesize` keeps both frequencies' distance as contradiction depth.
+4. **Council.** `InnerCouncil::from_signals(trust, humility, flow, load)`
+   (planner `mul/escalation.rs::verdict_from` already derives these from a
+   `MulAssessment`) → `CouncilVerdict`; on `verdict.split`,
+   `deposit_counterfactual(&verdict, &mut edge)` tags the minority
+   diagnosis's edge with the −6 mantissa (the planner already has
+   `impl EpisodicEdge for CausalEdge64`, D-DCR-3). **Not on the path:**
+   `quorum.rs::quorum_project` and `quorum_project_blackboard` are
+   unconditional `todo!` scaffolds blocked on D-ATOM-1/3 (Codex P1,
+   verified at `quorum.rs:216, 253`), and `CounterfactualMailbox` /
+   `revise_if_minority_wins` are the v3 `todo!` arm. The probe uses the
+   council verdict and the deposit only; the minority stays addressable
+   under the −6 lane and may win later when the mailbox arm exists.
+5. **Discriminating evidence (OPEN — the first readout this plan asks for).**
+   For the top two candidates, name the single observation whose presence
+   would swap their order. Definition (a READOUT, no state): clone the
+   arena; for each candidate premise `observe` it with the candidate's own
+   truth and re-rank; keep ONLY premises that actually swap the top two;
+   among those report the one that leaves the largest post-swap margin,
+   ties broken by ascending premise id. If no single premise swaps the
+   order the readout returns an explicit `None` carrying the current
+   top-two margin ("no single observation discriminates") — never the
+   largest non-swapping displacement (CodeRabbit on #1137: a largest
+   displacement can preserve the order, and the no-swap case must be
+   defined).
+   `Frontier.gaps` gives the candidate premises; the clone-and-observe is
+   what `dismech_counterfactual` already does for a cut edge, run here for
+   an added one. Not built; D-HOUSE-2.
+6. **Elimination (OPEN — the second readout).** "Ruled out" is NOT "ranked
+   low" (MedCare's own plan says so). Definition: a candidate is eliminated
+   when an ASC challenge with DISJOINT counter-evidence drove its
+   `expectation()` below a pre-registered floor AND the contradiction depth
+   is recorded on the belief. Both facts are already on the arena entry
+   (`truth`, `contradiction`, `stamp`); elimination is a predicate over
+   them, never a new enum. Falsifier: an eliminated candidate must not
+   re-enter the top-k without a NEW disjoint stamp. D-HOUSE-3.
+7. **Memory and the anchoring alarm (gated on the sibling plan).** On a
+   green-flip (`CollapseHint::Flow` plus an `Epiphany`) a `WisdomMarker`
+   is laid; on the next case the marker biases the initial ranking. The
+   alarm is `free_energy`: prior says X, evidence says Y, surprise rises →
+   the prior is demoted, not the evidence. The contract has the marker but
+   no field (no per-atom `bias`/`prediction`/`free_energy`); the excluded
+   crate has the field but no consumer. D-HOUSE-4 waits for
+   `thinking-engine-harvest-closure-v1` D-TEH-2 (ghost prior harvested with
+   a falsifier) and is NOT built by porting `GhostField` as a singleton.
+
+The list above reads linearly; the execution is a LOOP. On a shallow pass
+steps 3–4 do not exist yet (Control-bucket tactics open at `Analogical`), so
+the real shape is: board → periphery → (elevate) → board again → challenge →
+council, and only after the loop settles do the two readouts (5–6) and the
+memory step (7) run. §4b's aperture is definable only because of this loop.
+
+What the style ADDS beyond the parts: the order (1→7), the periphery budget
+`k`, the elimination floor, and the rule that step 3 runs before any
+commit. It adds no arithmetic of its own.
+
+## 3. Persona, rung, style — where House sits on the ladder
+
+- **Rung content:** House is a rung-3 composition (the 34 recipes ARE the
+  runbooks: RCR #4, TR #6, ASC #7, CAS #8, CR #11). Its ADMISSION follows
+  `Recipe::min_rung`: on a shallow pass only the gate/datapath tactics fire,
+  and the House cycle degenerates to "board + periphery"; ASC/CR (Control
+  bucket) open at `Analogical`. That degeneration is the pyramid working as
+  designed, not a defect.
+- **Style application is not confined to rung 4** (scope correction of
+  2026-08-30 in `persona-vs-rung-ladder.md`). House is a program over the
+  atom/mask substrate; the landing zone named for such programs is
+  `.claude/v3/FUTURE-DESIGN.md` E-THINKING-STYLES-ARE-CLASSES-1 (style = StepMask ×
+  WideFieldMask + rung set + KausalSpec). House would be ONE such class.
+  This plan does not mint it (D-TSC-2/3 are the mint vehicle, still
+  queued).
+- **Persona is out of scope by ruling O3.** `PersonaProfile` /
+  `CognitiveBaseline` (12 dims incl. `precision_drive`,
+  `epistemic_humility`) and the adjective-36 are the unwired
+  persona-modelling storyline. A "House persona" would be a data card there;
+  nothing in §2 needs it. `StyleFamily` (12) is enough to address the
+  program: the natural default is `Analytical` with a raised periphery
+  budget; measured, not assumed (§4).
+- **The missing edge is O2**, unchanged: nothing selects recipes FROM a
+  style. `tactic_for_bias` selects from graph health. A House program needs
+  the style→recipe edge; §4's probe drives the cycle by hand so the edge's
+  absence does not block the measurement, and the probe's result is the
+  evidence for building the edge.
+
+## 4. Falsifier first — PROBE-HOUSE-DIFFERENTIAL-1 (D-HOUSE-1)
+
+**Question.** Does the House cycle (§2 steps 1–4, run by hand) recover a
+planted cause more often than RCR alone, above a size-preserving shuffle
+null, on synthetic arenas whose distractors share predicates with the
+planted cause?
+
+**Why synthetic.** No clinical data may enter lance-graph (MedCare-rs
+commitment #9; the private repo is the only place that corpus exists), and
+the KJV SPO stream has no observation→cause structure. Planted-cause
+recovery is the standard falsifier for an abductive ranker and needs no
+domain vocabulary.
+
+**Fixture (pre-registered).**
+- N = 200 arenas, SplitMix64 seed `0x9E3779B97F4A7C15 ^ i`.
+- One planted cause `C*` with 4–6 features `O_j`, written as RULES
+  `C* → O_j` (`Copula::Inh`, frequency 0.9, confidence 0.6, disjoint
+  stamps), and one `case` with observations `case → O_j` for every planted
+  feature (the direction RCR needs; see §2 step 1).
+- 5 distractor causes `D_i`, each with rules `D_i → O_j` sharing 1–3
+  features with `C*` (the shared-predicate trap RCR abduces over) plus 2–3
+  private features the case does not show. `Throttle::hub_indegree` is set
+  above 7 so a feature shared by all six causes and the case is not barred
+  as a hub.
+- One disjoint counter-evidence stamp per distractor (frequency ≤ 0.3) that
+  is ONLY visible to step 3.
+- Half the arenas additionally carry a "far" fact reachable only by a tactic
+  the starting rung excludes (the periphery's job).
+
+**Arms.**
+- A0: RCR alone (`rcr_abduce`, rank by expectation).
+- A1: RCR + admit + ASC on the leader (steps 1, 1b, 3).
+- A1c: A1 + council (steps 1, 1b, 3, 4; no periphery) — the control that
+  isolates the periphery's contribution.
+- A2: full cycle (steps 1, 1b, 2, 3, 4; `k = 4` periphery, `tol = 0.02`).
+- AN: A2 on 25 permutations of the observation→cause links per arena
+  (size-preserving null).
+
+**Metrics.** p@1 and p@3 for `C*`; elimination false-positive rate (the
+planted cause must never be eliminated); periphery fire rate.
+
+**PASS iff** (a) A2 p@1 − A0 p@1 ≥ 0.05 AND A2 p@1 > AN p95, (b) planted
+cause eliminated in 0 arenas, (c) periphery fires on ≥ 10 % and ≤ 90 % of
+arenas (can-fire AND can-stay-silent). **KILL** otherwise, and the plan
+stops at §1 as a parts list.
+
+**Anti-vacuity guards (each disable-verified before the run is trusted).**
+- Removing the disjoint-stamp gate in the fixture must make A1 worse than
+  A0 (otherwise the guard was never binding).
+- Zeroing the far facts must drop the periphery fire rate to ~0.
+- A2 with `k = 0` must equal A1c to the bit (A1 lacks the council, so
+  equality with A1 was the wrong control — CodeRabbit on #1137).
+**Deliverable shape.** One example in `lance-graph-planner/examples/`
+(no library surface), plan §4 RESULT (added as a dated subsection when the
+run reports), one board entry, STATUS row.
+Nothing else lands from a PASS except a licence to design the O2 edge.
+
+## 4b. Three follow-on probes — kept OUT of D-HOUSE-1 (external review, 2026-09-02, adopted)
+
+An earlier revision of this plan folded three extra arms into D-HOUSE-1. An
+external review (GPT-class, adjudicated against the code per
+`E-EXTERNAL-REVIEW-ADJUDICATED-1`) objected, and the objection holds: D-HOUSE-1
+asks ONE causal question — does attacking the leader and preserving dissent
+improve planted-cause recovery — and five hypotheses in one harness destroy
+attribution. #1136's value came from isolating that one arm was rank-inert.
+So the three become separate, gated probes, each run on the SAME arenas and
+null AFTER D-HOUSE-1 has reported, each graded against the D-HOUSE-1 A2
+ordering, each with its own rank-inertness guard (Spearman ρ < 1 against
+A2 before any p@1 is read) and its own inertness falsifier. Re-checking the
+processing order also found three defects in the folded version; they are
+fixed here.
+
+- **D-HOUSE-1b — resonance prior: anchoring vs recognition.** Ingredient:
+  the perspectival `awareness_dto::ResonanceDto` / `HdrResonance` shape, the
+  pyramid base. TWO sub-arms, because the pyramid names a different shape
+  than "seed the board": `res-seed` (resonance orders the board, RCR runs
+  regardless) and `res-gate` (a unanimous, epiphany-grade resonance commits
+  WITHOUT RCR; otherwise RCR runs — the faithful base-first form, and the
+  more dangerous one). Falsifiers: neither sub-arm may beat the planted
+  cause's rank on no-distractor arenas (a prior overriding clean evidence
+  is anchoring, measured); `res-gate` must abstain on ≥ 10 % of arenas
+  (a gate that always commits is not a gate).
+- **D-HOUSE-1c — signed qualia as contrast agent.** Per candidate ONE signed
+  i4 from the ASC outcome (+ `Revised`, − `BlockedSelfReference`,
+  0 `NoTarget`), read into the existing `QualiaI4_16D` lanes (tenant 1) and
+  consulted by the POST-challenge re-rank only — the dye cannot inform the
+  initial board because ASC has not run yet. Order fix: ASC runs on the
+  top-k (k = 3), not the leader alone; a single dyed candidate is trivially
+  inert or trivially decisive and would trip the ρ guard for the wrong
+  reason. No lane is minted: if no existing lane fits, the probe reports
+  that and stops (OQ-CSV-1 is answered by the probe's need, never
+  pre-empted). Falsifier: all signs forced to 0 must equal A2 to the bit.
+- **D-HOUSE-1d — aperture (the "humor" hypothesis, held as a STYLE
+  hypothesis).** The cycle has one knob, `k` in `peripheral_sample_where(k)`.
+  The candidate operator is incongruity → aperture, `k(pass n) =
+  f(incongruity at pass n−1)` with fixed `k` on pass 1 — the order fix: `k`
+  is spent at step 2 but CR contradiction depth exists only after steps 3–4,
+  so the knob can only read the PREVIOUS pass of the fixpoint loop. The word
+  "humor" does not substitute for that operator; until the operator is
+  defined and measured, this row is a hypothesis, not a scheduled probe.
+  Inertness falsifier when it runs: raising `f` must admit a watcher on some
+  arena, lowering it must silence one.
+
+Ghost-family fence for all three (banked as
+`E-A-GHOST-TRACE-IS-NOT-THE-COUNTERFACTUAL-LANE-1`): `GhostType` /
+`GhostField` / `GhostEcho` / `WisdomMarker` are LINGERING TRACES (Staunen =
+persistent wonder, Wisdom = harvested knowing); the −6 counterfactual lane
+(`deposit_counterfactual`, `CounterfactualMailbox`) is a NON-AUTHORITATIVE
+RUNG. The second may consume the first as a starting prior; it is never one
+of them, and none of these probes reads `WisdomMarker` as counterfactual
+state.
+
+## 5. Deliverables
+
+| D-id | title | scope | status |
+|---|---|---|---|
+| D-HOUSE-0 | this plan: parts list, composition, ladder placement, falsifier | plan-only | Shipped (this PR) |
+| D-HOUSE-1 | PROBE-HOUSE-DIFFERENTIAL-1 (§4) | planner example | Queued — next step |
+| D-HOUSE-1b | resonance prior probe, `res-seed` + `res-gate`, anchoring vs recognition (§4b) | planner example, same arenas | Gated on D-HOUSE-1 reported |
+| D-HOUSE-1c | signed-qualia contrast probe over top-k ASC, post-challenge re-rank only (§4b) | planner example, same arenas | Gated on D-HOUSE-1 reported |
+| D-HOUSE-1d | aperture hypothesis: incongruity → `k` across passes; held until the operator is defined (§4b) | hypothesis | Not scheduled |
+| D-HOUSE-2 | discriminating-evidence readout (§2 step 5) | planner `nars` readout, no state | Gated on D-HOUSE-1 PASS |
+| D-HOUSE-3 | elimination predicate (§2 step 6) + never-re-enter falsifier | planner `nars` readout | Gated on D-HOUSE-1 PASS |
+| D-HOUSE-4 | anchoring alarm over a harvested ghost prior (§2 step 7) | planner, after D-TEH-2 | Gated on `thinking-engine-harvest-closure-v1` D-TEH-2 |
+| D-HOUSE-5 | the style→recipe edge (O2) with House as its first program — two faces after PASS: (a) persona modelling as a LENS (a readout policy over the same arena: admitted recipes, periphery budget, elimination floor, council weights; philosophical labels are further lenses, and two lenses with identical rankings are one lens), (b) the program materialised over the loco palette (`ogar-loco` / `ogar-r2il`, the ruled convergence for style application; no new atom — the 226-atom palette's 30 reserved slots stay untouched; any opcode rides the batched OGAR mint, never solo) | contract + planner (+ OGAR mint) | Gated on D-HOUSE-1 PASS and the D-TSC-2/3 mint |
+| D-HOUSE-6 | consumer pointer: `DiffRow` gains READ fields `eliminated` / `discriminator` | MedCare-rs (out of scope here) | pointer only |
+
+## 6. What this plan does NOT do
+
+- No opcode, tenant, ClassView, axis set, or lane; no `ENVELOPE_LAYOUT_VERSION`
+  change.
+- No new struct for the differential: `Frontier` + `BeliefArena` entries are
+  the state; §2 steps 5–6 are readouts.
+- No port of `GhostField` into the contract or planner as a singleton; the
+  ghost prior arrives per-mailbox through the sibling plan or not at all.
+- No persona modelling, no adjective-36 wiring (O3).
+- No clinical vocabulary, fixtures or data in lance-graph.
+- No name "House" in any identifier, test name, or vocabulary row.
+
+## 7. Provenance
+
+Read-only survey on 2026-09-02 over `94543a5`: `persona-vs-rung-ladder.md`
+(full), `recipes.rs` (rung admission + periphery), `escalation.rs`,
+`quorum.rs`, `counterfactual.rs`, planner `nars/{tactics,tactic_select,
+stance}.rs`, `mul/escalation.rs`, `thinking-engine/src/{ghosts,persona,
+meaning_axes}.rs`, board entries `E-PERIPHERAL-DISSENT-GUARDS-THE-
+STRATIFICATION-1`, `E-A-WATCHER-THAT-CANNOT-DISSENT-IS-NOT-A-WATCHER-1`,
+`E-SATURATION-SWITCHES-TO-PASSIVE-QUORUM-1`, `E-DIALECTIC-V1-TACTICS-IN-
+PLANNER-1`; MedCare-rs `docs/REASONING_CHAIN_INTEGRATION_PLAN.md` §A rows
+(private; cited by path).
diff --git a/.claude/plans/thinking-engine-harvest-closure-v1.md b/.claude/plans/thinking-engine-harvest-closure-v1.md
new file mode 100644
index 000000000..c458b40fc
--- /dev/null
+++ b/.claude/plans/thinking-engine-harvest-closure-v1.md
@@ -0,0 +1,210 @@
+# thinking-engine harvest & closure v1 — harvest every gem into the spine, retire the residue, close the chapter
+
+> **Status:** PROPOSAL, plan-only (W0 = this census; no code moves in this
+> PR). **Baseline:** lance-graph `94543a5`. **Date:** 2026-09-02.
+> **Purpose:** the crate `crates/thinking-engine` (51 files,
+> workspace-EXCLUDED, off CI, ~40 clippy lints + a stale test-compile break
+> per `TD-THINKING-ENGINE-EXCLUDED-DEBT-1`) is the ladybug-rs migration's
+> landing zone. Most of what was good in it has already been lifted into the
+> spine; what remains is a mix of hot-path engines reached only through one
+> optional feature, unwired gems with named landing zones, a calibration
+> battery, and self-declared residue. This plan fixes the FATE of every file,
+> lists the open plans and rows that still point at the crate, and sequences
+> the closure so that nothing is lost and nothing is ported without a
+> consumer and a falsifier.
+> **Inputs (read, not re-derived):** `.claude/v3/MODULE-TABLE.md` ancestry
+> census (51 rows, gem-status column), `.claude/v3/COMPONENT-MAP.md` §3,
+> `.claude/v3/FUTURE-DESIGN.md` "the migration arc", `TECH_DEBT.md`
+> (`TD-THINKING-ENGINE-EXCLUDED-DEBT-1`, `TD-GHOST-ECHO-DUP-1`,
+> `TD-RESONANCEDTO-DUP-1`), STATUS_BOARD rows D-PERSONA-1..6, D-TRI-1..6,
+> D-TSC-1b, D-TTV-1, D-REUNIFY-2..6, ENTROPY-MILESTONES M8, board entries
+> `E-RUNG-ASCENT-WIRED-1`, `E-CE64-NAME-COLLISION-DEDUP`,
+> `E-MORTON-CASCADE-V3-1` (downgraded), `E-ACK-ELIMINATED-1`,
+> `E-NOBODY-WAITS-1`. **Sibling:** `house-differential-style-v1` (consumer
+> of the ghost-prior harvest, D-TEH-2).
+
+## 0. The measured live footprint
+
+Two crates depend on `thinking-engine` (`grep thinking-engine crates/*/Cargo.toml`):
+
+| consumer | dependency | symbols actually used |
+|---|---|---|
+| `cognitive-shader-driver` | OPTIONAL, feature `with-engine`, default off | `dto::BusDto`; `cognitive_stack::{SD_FLOW_THRESHOLD, …}`; lens lookups (`jina_lookup`, `bge_m3_lookup`, `reranker_lookup`) in doc/test paths |
+| `lance-graph-callcenter` | required, `default-features = false` | production: `bridge_gate::{CognitiveBridgeGate, CognitiveAuthResult, CognitiveOpKind}` (`cognitive_bridge_gate.rs:45`; callcenter IMPLEMENTS the trait; direction callcenter → thinking-engine is the only allowed one). **Tests also** call `jina_lens::{jina_lookup, jina_distance}`, `bge_m3_lens::bge_m3_lookup`, `reranker_lens::reranker_lookup` (`cognitive_bridge_gate.rs:474–477`, test `pure_ops_emit_zero_audit_events`) — a Codex finding on #1137; the first draft of this row said "only `bridge_gate`", which was true of production code and false of the crate |
+
+So the spine's hard PRODUCTION dependency on a 51-file crate is ONE trait
+module (`bridge_gate.rs`), plus one unit test that touches three lens
+modules; the optional dependency is the DTO ladder plus two thresholds. Everything else is internal to the crate or reached only by its
+own examples. That is the finding this plan is built on: closing the chapter
+costs one trait move and one DTO-home decision, not a rewrite.
+
+## 1. Fate table — every file, one verdict
+
+Verdict vocabulary: **HARVESTED** (already in the spine; the crate copy is a
+dup to delete or a re-export), **HOT-VIA-FEATURE** (real compute, reached
+only through `with-engine`), **GEM→landing** (unwired; a named spine home
+and a consumer), **LAB** (calibration battery; keep in a lab crate),
+**RESIDUE** (self-declared legacy; delete).
+
+### 1a. HARVESTED — the spine already has it
+
+| file / symbol | spine home | action |
+|---|---|---|
+| `cognitive_stack::RungLevel` | `contract::cognitive_shader::RungLevel` (`pub use` since E-RUNG-ASCENT-WIRED-1) | already a re-export; nothing to do |
+| `cognitive_stack::ThinkingStyle` (12) | `contract::style_family::StyleFamily` (D-TSC-1) | deprecated alias; delete with the crate |
+| `contract_bridge::contract_style_to_engine` | `StyleFamily::family()` | routed; delete with the crate |
+| `ghosts::GhostType` (8) | `contract::escalation::GhostEcho` (identical 8, same order; `TD-GHOST-ECHO-DUP-1`) | the ENUM is harvested; the FIELD is not (see 1c) |
+| `meaning_axes::Archetype` + council vote | `contract::escalation::{Archetype, InnerCouncil, is_split}` (D-PERSONA-1) | harvested; `CouncilWeights::modulate/shift_toward` has no spine twin — see 1c |
+| `layered::CascadeChannels8` (ex-`CausalEdge64`, E-CE64-NAME-COLLISION-DEDUP) | 8-channel ↔ SPO transcoder D-CSV-9 (PR #387); channel = CE64 signed mantissa per FUTURE-DESIGN | transcoder shipped; the cascade itself is HOT-VIA-FEATURE (1b) |
+| `dto::ResonanceDto` → `PerturbationDto` | D-PERT-1 (#630) | done; `awareness_dto::ResonanceDto` keeps its name (perspectival) |
+| `cronbach::cronbach_alpha`, `QuorumLevel` | `jc::reliability::{cronbach_alpha, icc}` | dup; retire into jc (W2) |
+| `superposition::ThinkingStyle = DetectedStyle` alias | renamed `DetectedStyle` (D-TSC-1) | the alias name still collides with `cognitive_stack`'s alias in the same crate; delete both with the crate |
+
+### 1b. HOT-VIA-FEATURE — real compute, one caller behind `with-engine`
+
+`engine.rs` (u8 MatVec, `cycle_vnni/cycle_auto`), `bf16_engine.rs`,
+`signed_engine.rs`, `f32_engine.rs`, `builder.rs`, `lookup.rs`
+(`TextToThought`), `codebook_index.rs`, `pooling.rs`, `sensor.rs`,
+`dto.rs` (Φ/Ψ/B/Γ ladder), `awareness_dto.rs`, `qualia.rs` (`Qualia17D`),
+`think.rs` (the "Thinking is a struct" carrier, minimum scope),
+`superposition.rs`, `layered.rs`, and two lens modules (`bge_m3_lens`,
+`reranker_lens`). (`role_tables.rs` is LAB, 1d; `jina_lens.rs` is RESIDUE,
+1e — each file has exactly one fate; the two callers that still touch
+`jina_lens` — callcenter's test and the driver's doc/test lookups — are
+rewritten in W1, not carried.)
+
+Two decisions own this group and both already have rows:
+- **M8** (ENTROPY-MILESTONES, QUEUED): four near-duplicate engines
+  (u8/BF16/i8/f32, same 7-method API) → one enum-dispatched engine with a
+  parity suite across dtypes. `signed_domino`/`domino`/`composite_engine`/
+  `dual_engine`/`branching` (1c) collapse into the same slice.
+- **D-TTV-1** (Queued): thinking-related tenants → V3 substrate. This is
+  the vehicle for the DTO ladder's home: `BusDto.{converged, cycle_count}`
+  is the D-MBX-A6 Outcome signal (COMPONENT-MAP §3), so the ladder lands
+  where the outcome is consumed — the driver side — never in the zero-dep
+  contract. Ruling needed on WHICH crate (driver vs a small `thinking-dto`
+  crate); this plan does not decide it.
+
+### 1c. GEM → landing — unwired, with a named home AND a consumer
+
+| file / symbol | what it is | landing | consumer that justifies the port | gate |
+|---|---|---|---|---|
+| `ghosts::GhostField::{imprint,bias,prediction,free_energy,prune}` | Friston prior as a per-atom decaying field; the anchoring alarm. **Family fence:** this is the LINGERING-TRACE family (Staunen = persistent wonder, Wisdom = harvested knowing) — NOT the non-authoritative counterfactual rung (the −6 lane, `deposit_counterfactual` / `CounterfactualMailbox`), whose docs call themselves "ghost-tier" (`TD-GHOST-TIER-NAME-COLLISION-1`). The rung may consume a trace as a starting prior; it is never one. `E-A-GHOST-TRACE-IS-NOT-THE-COUNTERFACTUAL-LANE-1` | planner `nars/ghost_prior.rs` over `contract::escalation::WisdomMarker`; per-thought/per-mailbox, NEVER a singleton field. **Intentional semantic change, not source parity** (Codex on #1137, verified): the decay rate agrees (0.85) but the FLOORS do not — `GhostField` drops contributions below 0.001 and `prune` deletes them, `WisdomMarker` clamps at 0.1 forever. Porting the field over the marker raises long-lived bias by up to two orders of magnitude; the port declares the floor it adopts and a calibration gate (free-energy response on a recurrence fixture under both floors) decides it | `house-differential-style-v1` D-HOUSE-4 (anchoring alarm) | falsifiers: bias decays monotonically to the DECLARED floor (whichever the calibration gate picks); `free_energy` RISES on a context shift and FALLS on a recurrence (two-sided); disable-run on the decay constant; the two-floor calibration comparison is reported, not assumed |
+| `meaning_axes::CouncilWeights::{modulate, shift_toward}` | free-energy-weighted three-archetype modulation with renormalisation | `contract::escalation` beside `InnerCouncil` — only if a caller needs weighted (not majority) council output | none today | do NOT port until a caller exists; record the formula in the plan (done here) |
+| `meaning_axes::AXES_48` + `AxisActivation` | 48 bipolar semantic axes (r = 0.9913 vs Jina cosine, per census) | `quorum.rs` names an `AxisId` BLOCKED on D-ATOM-1; AXES_48 is the natural candidate vocabulary | `quorum_project` per-axis projection | falsifier-first (the September ruling forbids a fixed axis basis minted by assertion): a size-preserving shuffle of axis assignments must NOT reproduce the 0.9913; until measured, the table stays where it is |
+| `persona::{PersonaProfile, CognitiveBaseline}` | 12 personality constants + mode/temperature/rung bounds | a Layer-2 role catalogue DATA card (≤32 identities, I-VSA-IDENTITIES) in the persona storyline | none (O3: persona unwired by design) | park; no port |
+| `persona::{A2AMessage, A2APayload, AgentDto}` | identity-carrying inter-agent messages | none | none | **RETIRE** — inter-mailbox handoff types were eliminated (`E-ACK-ELIMINATED-1`, no CollapseGateEmission/baton); COMPONENT-MAP already marks A2AMessage BLOCKED for exactly this |
+| `world_model::{SelfState,UserState,FieldState,GestaltState,WorldModelDto}` | situational-awareness DTO | driver side with the DTO ladder (D-TTV-1) after the `GestaltState` dedup (defined twice: here and `awareness_dto.rs`) | driver | dedup first, then move with 1b |
+| `cognitive_trace::{SpoTriple,CognitiveTrace,LensTrace}` | provenance chain text → thought | `temporal.rs` witness rows already carry provenance; a trace is a READ over versions | debug tooling only | port only if a probe needs it; otherwise delete with the crate |
+| `domino.rs`, `signed_domino.rs`, `composite_engine.rs`, `dual_engine.rs`, `branching.rs` | 5 cascade shapes (top-K NARS-context cascade, signed variant, multi-lens superposition, u8-vs-BF16 disagreement, 4×4 spawning cascade) | M8 collapse surface | the one engine (1b) | parity suite; `branching.rs`'s "spawn, don't filter" shape is the idea worth keeping (§4) |
+| `l4.rs`, `l4_bridge.rs`, `tensor_bridge.rs` | L4 "particle" personality layer + L3→L4 learning bridge (self-declared LOSSY) + embedding-output type | the LEARNED tenant of the triangle (`triangle-tenants-gestalt-separation-v1` W4b) | D-TRI-1 learned lane | BLOCKED on the orphan-write (`commit_to_l4` writes `&mut L4Experience` with no owner stamp — COMPONENT-MAP §3); stays until the learned lane wires, then ports as a ClassView reading |
+| `prime_fingerprint.rs` | prime-DFT bit fingerprint, orthogonal by construction | VSA niche (I-VSA-IDENTITIES four tests) | none today | re-scope before any wiring; keep in LAB |
+| `spiral_segment.rs` | 51× generative row compression (4×BF16 per row) | codec home: `bgz-tensor` / `ndarray::hpc` (encoding-ecosystem.md governs) | codec certification | `certification-officer` battery vs ground truth before promotion |
+| `semantic_chunker.rs` | chunk boundaries = convergence jumps, no forward pass | `deepnsm-v2` (text side) or the paperless sentence assembler | tesseract-paperless `assemble_sentences` is the only live sentence producer | falsifier: boundaries vs a gold sentence split; else stays LAB |
+| `contrastive_learner.rs`, `osint_bridge.rs` | online EMA table update; crawl → engine pipeline | `lance-graph-osint` owns the arc | osint arc | move with the osint arc or delete when it does |
+| `inference_backend.rs` | 7-backend runtime registry ("nothing is killed, deprecation is data-driven") | LAB | R&D bench only | keep in the lab crate |
+| `bridge_gate::{CognitiveOpKind, CognitiveAuthResult, CognitiveBridgeError, auth_to_result, CognitiveBridgeGate, PassthroughGate, DenyAllGate}` — the WHOLE public surface of the module, seven items | cross-tenant authorization injection point (trait + result/error enums + two reference gates) | `lance-graph-contract` (zero-dep, trait-only — it already says it lives low so callcenter can implement it), as one module, all seven items together; owner = contract | `lance-graph-callcenter` (live: `CognitiveBridgeGate`, `CognitiveAuthResult`, `CognitiveOpKind` at `cognitive_bridge_gate.rs:45`) | W1 compatibility contract: (a) all seven items move in one PR, identical shapes; (b) thinking-engine keeps `pub use lance_graph_contract::<module>::*` re-exports for that PR so callcenter's imports flip in the same change without a window where either path is missing; (c) callcenter's `pure_ops_emit_zero_audit_events` test is rewritten to exercise the gate without the three lens lookups (or moved to the driver's `with-engine` tests) BEFORE the dep line is deleted; (d) callcenter builds and its tests pass with no thinking-engine path dep — only then is the claim made |
+
+### 1d. LAB — the calibration battery (keep, rename, gate)
+
+`auto_detect.rs`, `ground_truth.rs` (candle-gated), `reencode_safety.rs`
+(x256 re-encode proof), `silu_correction.rs`, `tokenizer_registry.rs`,
+`centroid_labels.rs`, `bridge.rs` (spiral → table index), `role_tables.rs`
+(BF16 per-role tables). These feed D-MTS-2's certification gate and are
+calibration-only by design. Fate: they survive as a lab crate
+(`thinking-lab`, excluded, `calibration` feature) with a `--manifest-path`
+CI line so the formatting/clippy debt cannot re-accumulate unseen.
+
+### 1e. RESIDUE — delete
+
+`splat_ops.rs` (self-declared deprecated, superseded by `Think::*`, removal
+scheduled sprint-15+), `jina_lens.rs` (self-declared legacy; Jina v5 is
+ground truth), the second `GestaltState`, the two `ThinkingStyle` aliases,
+`persona::A2A*` (see 1c).
+
+## 2. Open plans and rows that still point at the crate — and their fate
+
+| item | today | fate under this plan |
+|---|---|---|
+| `rung-persona-orchestration-v1` D-PERSONA-1 | STATUS row says "In progress, branch claude/splat3d-…" | **STALE → Shipped**: `contract::escalation` (CollapseHint, InnerCouncil, EpiphanyDetector, GhostEcho, WisdomMarker, Checklist) and planner `mul/escalation.rs` (`verdict_from`, `boot_checklist`) are on main. Row regraded in this PR |
+| D-PERSONA-2 meta-recipe manifest | Queued | absorbed by `house-differential-style-v1` §2 (a style program IS a declarative recipe composition) — first program = House; the manifest FORM waits for D-TSC-3 |
+| D-PERSONA-3 hot/cold/feedback, `CrystalCodebook` → wisdom markers | Queued | absorbed by D-TEH-2 (ghost prior harvest) + D-HOUSE-4; the cold path is a WisdomMarker store keyed by fingerprint (I-VSA-IDENTITIES), not a history dump (rung-persona §8 already says so) |
+| D-PERSONA-4 macro-eval harness | Queued | its first instance is PROBE-HOUSE-DIFFERENTIAL-1 (scenario → trace → diagnose); no separate harness before a second consumer |
+| D-PERSONA-5 ractor outer-swarm, batons as messages | Queued | **proposed RETIRE** — contradicts `E-NOBODY-WAITS-1` (ractor = compile-time ownership only; no messages, no actors) and `E-ACK-ELIMINATED-1`; ruling needed, noted on the row |
+| D-PERSONA-6 Odoo `l10n_de` harvest | Queued | not thinking-engine work; belongs to the odoo blueprint plans, unchanged |
+| D-TRI-1..6 (triangle tenants) | value-tenant half merged (#717); D-TRI-6 ascent wired | unchanged; the LEARNED lane is the landing for `l4*` (1c) |
+| D-TSC-1b → TD (5 missing planner `default_modulation` arms) | In PR / TD open | small pay, independent of this plan; listed so it is not forgotten |
+| D-TTV-1 thinking tenants → V3 | Queued | the vehicle for 1b's DTO ladder home; ruling on crate placement needed |
+| D-REUNIFY-2 (8ch → SPO transcoder) | Backlog | **already shipped as D-CSV-9** (#387); row to regrade |
+| D-REUNIFY-3 (`Think` carrier unification) | Backlog | `think.rs` exists at minimum scope; moves with 1b under D-TTV-1 |
+| D-REUNIFY-4 (splat ops as `Think` methods) | Backlog | done in `think.rs`; `splat_ops.rs` is the residue (1e) |
+| D-REUNIFY-5 (rayon par_*), D-REUNIFY-6 (DOLCE filter wiring) | Backlog | superseded by V3 (SoA sweep, ontology at the membrane); propose CLOSE-superseded, ruling needed |
+| ENTROPY M8 four-engine collapse | QUEUED | W3 of this plan |
+| `TD-THINKING-ENGINE-EXCLUDED-DEBT-1` | open | paid by W1+W4 (the crate shrinks to a lab crate with a CI line) |
+| `TD-GHOST-ECHO-DUP-1` | open | paid by W2 (field ported over `GhostEcho`; crate enum deleted) |
+| `TD-RESONANCEDTO-DUP-1` | Deferred | resolved by D-PERT-1's split (two names now); row to close |
+| `E-MORTON-CASCADE-V3-1` | ⊘ CONJECTURE | the "reimagine the compute as a V3 Morton read" path; independent of closure — closure must not delete the legacy arm it compares against until that probe runs or is dropped |
+
+## 3. Closure sequence
+
+Each wave lands only with its gate green; no wave ports without a consumer.
+
+| wave | content | gate |
+|---|---|---|
+| **W0 (this PR)** | census, fate table, row regrades, idea harvest (§4) | boards consistent; SUPERSESSION-INDEX regenerated |
+| **W1 — cut the hard dependency** | move `bridge_gate` trait + gates to `lance-graph-contract` (trait-only, zero-dep); callcenter re-imports; decide the DTO-ladder home (D-TTV-1 ruling) so the driver's `with-engine` feature can point at it | `lance-graph-callcenter` builds with no thinking-engine path dep; driver `with-engine` still green; thinking-engine becomes a LEAF |
+| **W2 — harvest gems with consumers** | ghost prior → planner `nars/ghost_prior.rs` (D-TEH-2, consumer D-HOUSE-4); `cronbach` → retire onto `jc::reliability`; `semantic_chunker` → deepnsm-v2 only if its falsifier passes; `spiral_segment` → codec home via certification battery | each port has a two-sided falsifier + a disable run; each source file deleted in the same PR |
+| **W3 — M8** | one enum-dispatched engine; the 5 cascade shapes and 3 lens modules collapse; parity suite across u8/BF16/i8/f32 | NOT bit-parity across dtypes — u8 / BF16 / i8 / f32 differ in encoding by design, and `dual_engine.rs` exists to MEASURE that disagreement (Codex on #1137). Gate: per-dtype output tolerances plus dtype-invariant ranking/convergence invariants (top-k order, `converged`, `cycle_count` bounds) on real engine fixtures that instantiate all four engines (the driver's fixtures do not — they round-trip `BusDto` only); the pre-collapse `DualResult` disagreement is the baseline the collapsed engine must not exceed; the `branching` spawn shape kept as a mode, not lost |
+| **W4 — retire and rename** | delete RESIDUE (1e) + retired persona A2A; rename what is left `thinking-lab` (calibration feature, `--manifest-path` CI line); regrade every row in §2; pay `TD-THINKING-ENGINE-EXCLUDED-DEBT-1` | the name `thinking-engine` no longer appears in any `Cargo.toml` dependency; board rows closed or re-owned |
+
+Stop rules (non-negotiable inside this plan): nothing ports without a
+consumer named in §1b (the DTO ladder, D-TTV-1) or §1c (the gems); nothing ports as a singleton field or a new lane
+(ClassView reading or per-mailbox module only); `A2AMessage` and any
+inter-mailbox handoff type never revive; the `E-MORTON-CASCADE-V3-1`
+legacy arm is not deleted while that probe is open; no port of a fixed
+axis vocabulary by assertion (AXES_48 goes falsifier-first or stays LAB).
+
+## 4. The good ideas — what the chapter leaves behind even where the code dies
+
+Kept as doctrine with a named present or future home, so the closure is a
+harvest and not an amputation.
+
+| idea (from the crate's own docs) | keep as | home |
+|---|---|---|
+| "Models are sensors. The matrix is the brain. DTOs are the bus. One MatVec per cycle." | the Zone-1 cost model (200–500 ns cycle) that sizes every rung budget | `VISION.md`, three-zone hot-path model (#372) |
+| Ghosts = Friston priors with an asymptotic floor; anchoring is MEASURABLE as free energy between prior and evidence | the anchoring alarm | D-TEH-2 / D-HOUSE-4 |
+| Staunen / Wisdom as a DUALITY: wonder keeps a question alive, wisdom is what was harvested from it. The shipped `curiosity() = (1/(q+1))·(1−c)` is a later operational SLICE of Staunen, not the original pair | the two ends of the epistemic cycle, kept as two names | `exploration.rs` (curiosity), `escalation.rs` (`GhostEcho::{Staunen, Wisdom}`, `promote_to_wisdom`) |
+| A council SPLIT is the learning signal (one archetype sees what the others do not) — amplify it, never average it | shipped invariant | `escalation.rs`, `quorum.rs` ("split quorums are Contradictions — NEVER averaged") |
+| 48 bipolar meaning axes with a measured r = 0.9913 against a real embedding | the candidate for the per-axis quorum vocabulary, falsifier-first | `quorum.rs` D-ATOM-1 pointer (1c) |
+| Branching cascades SPAWN, they do not filter (parallel vectors, not survivors) | a mode of the one engine after M8 | W3 |
+| "Nothing is killed; deprecation is data-driven" (the backend registry doctrine) | the standing rule of this very plan (every retirement here cites a measurement or a ruling) | this plan §1 |
+| Chunk boundaries are convergence jumps — segmentation without a forward pass | a falsifiable chunker for the sentence seam | W2 (`semantic_chunker`) |
+| An L4 "particle" layer that learns OUTSIDE the wave cascade, XOR-bound, save/load | the LEARNED tenant of the autopoiesis triangle | D-TRI-1 learned lane |
+| Provenance text → thought as an appendable trace | already the temporal witness's job; keep as a read | `temporal.rs` |
+| The 8-channel cascade edge (BECOMES / CAUSES / SUPPORTS / REFINES / GROUNDS / ABSTRACTS / RELATES / CONTRADICTS) | transcoded to the SPO palette, kept as the comma-level mantissa carrier | D-CSV-9, FUTURE-DESIGN "first wiring target" |
+| Persona as 12 constants + a mode with rung bounds and a collapse bias | a Layer-2 data card for the persona storyline, when it opens | O3 (parked) |
+
+## 5. Deliverables
+
+| D-id | title | scope | status |
+|---|---|---|---|
+| D-TEH-0 | census + fate table + open-row reconciliation + idea harvest (this plan) | plan + board rows | Shipped (this PR) |
+| D-TEH-1 | W1: `bridge_gate` trait → contract; DTO-ladder home ruling recorded; thinking-engine becomes a leaf | contract + callcenter (+ driver feature pointer) | Queued — first code wave |
+| D-TEH-2 | W2: ghost prior harvested as planner `nars/ghost_prior.rs` over `WisdomMarker`, per-thought, with two-sided falsifiers; crate `ghosts.rs` deleted | planner | Queued — consumer = D-HOUSE-4 |
+| D-TEH-3 | W2: `cronbach` → jc, `semantic_chunker` / `spiral_segment` decided by their falsifiers | jc / deepnsm-v2 / codec home | Queued |
+| D-TEH-4 | W3: M8 engine collapse with parity suite; cascade shapes and lens modules collapse | thinking-engine → the one engine | Queued (owns ENTROPY M8) |
+| D-TEH-5 | W4: residue deleted, crate renamed `thinking-lab` with a CI line; §2 rows closed; TD paid | workspace | Queued — closes the chapter |
+
+## 6. Rulings this plan asks for (not assumed)
+
+1. D-PERSONA-5 RETIRE (contradicts E-NOBODY-WAITS-1 / E-ACK-ELIMINATED-1).
+2. D-REUNIFY-5/6 CLOSE-superseded by V3.
+3. The DTO ladder's crate home under D-TTV-1 (driver vs a small DTO crate).
+4. Whether `thinking-lab` survives as a named excluded crate or the
+   calibration battery moves into `jc` outright.
+
+## 7. What this PR does NOT do
+
+No file moves, no deletions, no Cargo changes, no feature flips, no new
+types. Row regrades touch only status cells and cite their evidence. The
+crate builds exactly as before.
diff --git a/.claude/tools/supersession_index.py b/.claude/tools/supersession_index.py
index e4ebe197c..69441a7e7 100644
--- a/.claude/tools/supersession_index.py
+++ b/.claude/tools/supersession_index.py
@@ -119,7 +119,8 @@ def plans_for(sym):
     f"That is a programme, not a cleanup.\n\n")
 if allblind:
     s0 = max(allblind, key=lambda s: stat[s][1])
-    out(f"**`{s0}` is the sharpest case: {stat[s0][1]} plans name it and *every one* is blind.**\n"
+    n0 = stat[s0][1]
+    out(f"**`{s0}` is the sharpest case: {n0} plan{'' if n0 == 1 else 's'} name{'s' if n0 == 1 else ''} it and *every one* is blind.**\n"
         f"Its COMPONENT-MAP note reads: {syms[s0][2][:120]}\n\n")
 used = {s0} if allblind else set()
 if [s for s in inverted if s not in used]: