planner: residue fixed point + Maslov Pyramid of Cognition probe (D-RESIDUE-FP-0, D-MATURATION3-0)

.claude/board/entries/2026-10-11-maturation3-ladder-probe.md

@@ -0,0 +1,100 @@
+# 2026-10-11 — Novelty → Wisdom: eight-stage 3-bit maturation ladder (D-MATURATION3-0)
+
+**Status:** EXPERIMENTAL probe, test-local. No production change.
+`crates/lance-graph-planner/tests/maturation_ladder.rs`
+
+```text
+cargo test -p lance-graph-planner --test maturation_ladder
+cargo test --release -p lance-graph-planner --test maturation_ladder measure_ -- --nocapture
+```
+
+## Codebook (bits 50..52 under an explicitly declared experimental reading)
+
+| code | stage | earned by (executed result) |
+|---|---|---|
+| 7 | Staunen | violation behind one observation root |
+| 6 | Anomaly | violation behind ≥ 2 independent roots |
+| 5 | DogEar | `hydrate` → ProposedOnly (gap localized) |
+| 4 | Candidates | `hydrate` → Partial |
+| 3 | Mechanism | `hydrate` → Hydrated |
+| 2 | Falsification | discriminating test: survivor passed AND rival ruled out |
+| 1 | Integration | applied as a premise; ≥ 1 new inference resolved through it |
+| 0 | Wisdom | from Integration, an ablation in an isolated copy has impact > 0 |
+
+Forward results never move a more mature state backward; only a violation
+reopens it. Repeating a completed result never advances it. Unsupported
+results abstain.
+
+## Findings
+
+- **VERIFIED-IN-CODE — production hydration has no feedback seam.**
+  `pearl::hydrate` reads only the sealed chain; a resolved relation is never a
+  premise. Measured on the A → B → C chain: baseline residue 2, test-local
+  feedback (production `hydrate` unchanged, over a premise chain) residue 0.
+- **MEASURED (release, this container):** `mature` median 1.99–2.83 ns,
+  p95 3.0–4.6 ns, 0 allocations over 1e6 transitions. Fixed point, 3 relations
+  / 4 observations: baseline ~165 ns, 10 allocs, residue 2; feedback ~157 ns,
+  8 allocs, residue 0. The probe loop allocates (premise pool `Vec`); that is
+  test scaffolding, not a production cost claim.
+- **EXPERIMENTAL — anchor ablation:** cutting derived A in an isolated copy
+  raises residue by 2 (B and C lose their bridge); the original population is
+  untouched.
+- **EXPERIMENTAL — lineage:** a derived premise carries the union of its
+  parents' observation roots and never adds one (C: 4 roots from 4
+  observations). Two paths from the same roots count 2 supports, not 4.
+- **EXPERIMENTAL — self-support:** only resolved relations are premises; a
+  two-relation cycle stays open. Disabled, it grounds itself.
+- **EXPERIMENTAL — deeper Staunen:** "1 → 4 does not hold" contradicts
+  nothing without the anchor chain and is a violation with it.
+
+Disable runs, all red: forward result moving backward; derived dropping a
+parent's roots; write leaking into W; write not clearing the old code;
+integration without new inferences; Wisdom without ablation impact; existing
+provenance accepted; discrimination without ruling out; feedback off; open
+relations as premises. (The first T1 fixture hid the leak/clear disables; it
+was fixed before recording.)
+
+## CONTRACT DECISION — migration (not taken)
+
+Bits 50..52 are plasticity: `CausalEdge64::learn` gates on them,
+`MooreNars16` packs them at bit 50, `CausalEdgeV3` byte 3 carries them, and
+D-CE64-STAUNEN-0 measured that another meaning there changes `learn`. The
+probe therefore refuses `V2Stamped`, `V3Register`, `V1Legacy` and `Unknown`.
+Smallest viable path: a new declared reading (a provenance/generation such
+that historic rows keep plasticity), a decision where plasticity lives under
+that reading, and a `learn` that does not gate on maturation bits. Not done.
+
+## OPEN
+
+- Feedback in production requires D2 provenance ids (roots here are a
+  test-local u64 bitset).
+- `Executed` results are constructed by the test from real `hydrate`
+  outcomes; mapping recipe outputs onto them is undesigned.
+- T5 uses hydration outcomes as the discriminating test, a weak stand-in for
+  a Pearl trial; certification is untouched by design.
+
+## Addendum, same day — Maslov Pyramid of Cognition
+
+Named by the operator; the model is documented in
+`.claude/knowledge/maslov-pyramid-of-cognition.md`. Stage 7 renamed
+Astonishment.
+
+- **Corrected:** the first T9 applied the new violation to the anchor itself.
+  It now lands on the contradicted relation (B), which starts at 7, and the
+  anchor (A) stays at 0. Direct counterevidence against A's own relation
+  still reopens A.
+- **Added:** an indirect 3-cycle (A → B → C → A) stays open; disabled, it
+  grounds itself. A grounded 3-cycle resolves once and 64 further passes add
+  no root.
+- **Disable runs, red:** a derived relation manufacturing its own root;
+  violation routed by subject only (hits the anchor).
+- The root union matches the existing `lance_graph_contract::revision` mask
+  algebra (`independent_roots`, difference against ancestry).
+
+## Addendum — lineage guard restored (review on #1494)
+
+I had removed the "descendants are not premises" guard as unreachable. That
+holds only while relations never reopen; T9 reopens an anchor. Reproduced:
+ground the 3-cycle, remove the ground, reopen R0, and R2 (derived through R0)
+re-grounds it. The guard is restored. A test pins that R0 stays open, and a
+disable arm shows that without the guard R0 is re-grounded.

.claude/board/entries/2026-10-11-residue-fixed-point.md

@@ -0,0 +1,25 @@
+# 2026-10-11 — Residue fixed point: only new evidence lowers it (D-RESIDUE-FP-0)
+
+**Status:** MEASURED. Test only, no production change.
+`crates/lance-graph-planner/tests/residue_fixed_point.rs`.
+
+```text
+cargo test -p lance-graph-planner --test residue_fixed_point
+```
+
+Claim: the reasoning loop is cheap (`pearl::hydrate` over the sealed chain),
+so it runs to a fixed point; what remains open is residue. Residue here is
+hydration debt in bits 59..63 (`IndirectUnknown` edges).
+
+Pinned:
+- extra passes over the same evidence move no bit;
+- re-sealing evidence already present does not lower the residue;
+- unrelated evidence changes nothing; the missing binding lowers the residue
+  exactly where it lands;
+- the loop moves only bits 59..60, never certification.
+
+Disable runs, both red: promoting on partial evidence (3 of 4 tests fail);
+moving certification inside `hydrate` (the bit-isolation test fails).
+
+OPEN: the surprise level proposed for bits 50..52 is not in the layout (still
+plasticity); its residue is untested until O14/O4 decide the v3 layout.

.claude/board/entries/README.md

@@ -25,14 +25,16 @@ index row, (3) no duplicate entry id. Checks 1 and 2 are deliberately
 opposite directions; the stranding this convention prevents shows up in
 exactly one of them, never both.
 
-306 entries, 2026-08-06 .. 2026-10-11.
+308 entries, 2026-08-06 .. 2026-10-11.
 
 | date | entry id | finding | file |
 |---|---|---|---|
 | 2026-10-11 | `D-TELEPORT-0` |  | [2026-10-11-teleport-probe-source-restriction.md](2026-10-11-teleport-probe-source-restriction.md) |
 | 2026-10-11 | `synergy-ledger-classid-hotplug` |  | [2026-10-11-synergy-ledger-classid-hotplug.md](2026-10-11-synergy-ledger-classid-hotplug.md) |
+| 2026-10-11 | `D-RESIDUE-FP-0` |  | [2026-10-11-residue-fixed-point.md](2026-10-11-residue-fixed-point.md) |
 | 2026-10-11 | `research-recovery-reports` |  | [2026-10-11-research-recovery-reports.md](2026-10-11-research-recovery-reports.md) |
 | 2026-10-11 | `D-PHOTO-0` |  | [2026-10-11-photo-probe-64-populations.md](2026-10-11-photo-probe-64-populations.md) |
+| 2026-10-11 | `D-MATURATION3-0` |  | [2026-10-11-maturation3-ladder-probe.md](2026-10-11-maturation3-ladder-probe.md) |
 | 2026-10-11 | `D-TELEPORT-2` |  | [2026-10-11-lance-source-probe.md](2026-10-11-lance-source-probe.md) |
 | 2026-10-11 | `coresearch-labels-address-lifetime` |  | [2026-10-11-coresearch-labels-address-lifetime.md](2026-10-11-coresearch-labels-address-lifetime.md) |
 | 2026-10-11 | `coresearch-instance-authz` |  | [2026-10-11-coresearch-instance-authz.md](2026-10-11-coresearch-instance-authz.md) |

.claude/knowledge/maslov-pyramid-of-cognition.md

@@ -0,0 +1,94 @@
+# Maslov Pyramid of Cognition — from Astonishment to generative Wisdom, and back
+
+> READ BY: truth-architect, contradiction-cartographer, trajectory-cartographer,
+> v3-envelope-auditor, and any session touching CausalEdge64 bits 50..52,
+> novelty / Staunen / wisdom scheduling, `pearl::hydrate` feedback, or anchor
+> ablation.
+> STATUS: WORKING-MODEL. The probe is EXPERIMENTAL and test-local
+> (`crates/lance-graph-planner/tests/maturation_ladder.rs`, D-MATURATION3-0).
+> Nothing below is wired in production.
+
+## Name and scope
+
+The operator named this model after Maslow's hierarchy. It is a proposed
+computational model, not a claim that Maslow defined these stages.
+
+It is **not** Wang's NAL-1..9. NAL levels describe what a reasoner is able to
+do. The pyramid describes how far **one discovery** has matured.
+
+## The eight stages (bits 50..52, 3 bits, direction 7 → 0)
+
+| code | stage | earned by an executed result |
+|---|---|---|
+| 7 | Astonishment | an observation contradicts an expectation (one root) |
+| 6 | Anomaly | the contradiction has ≥ 2 independent roots |
+| 5 | DogEar | hydration localizes the gap (ProposedOnly) |
+| 4 | Candidates | hydration finds one binding (Partial) |
+| 3 | Mechanism | hydration finds both bindings (Hydrated) |
+| 2 | Falsification | a discriminating test passed the survivor AND ruled out a rival |
+| 1 | Integration | used as a premise, it resolved ≥ 1 other relation |
+| 0 | Wisdom | from Integration, ablating it in an isolated copy reopens ≥ 1 relation |
+
+Rules:
+
+- Stages may be skipped when an executed result justifies it.
+- A forward result never moves a more mature state backward.
+- Repeating a completed result (another hydration pass) never advances it.
+- Only a contradiction reopens a stage.
+- Wisdom is not "zero uncertainty". It is measured reuse: an ablation must
+  show that other relations depend on the anchor.
+
+## The recursion
+
+An anchor at 0 generates expectations through the relations derived from it.
+An observation that contradicted nothing before can now contradict one of
+those expectations. That contradiction lands on **the contradicted
+relation**, which starts its own pyramid at 7. **The anchor stays at 0.**
+
+Only counterevidence against the anchor's **own** relation reopens the anchor.
+
+So wisdom does not end surprise. It makes new surprises recognizable. This is
+recursive scaffolding, not circular confirmation.
+
+## Separation of contracts
+
+| bits | contract | role here |
+|---|---|---|
+| 50..52 | maturation (this model) | where the discovery stands |
+| 53..58 | witness / breadcrumb | handle to the transition record |
+| 59..63 | `EpistemicState5` | what is licensed; the input to ablation |
+| 24..39 | F/C | statistical evidence under its declared reading |
+| — | thought / recipe | which operation ran |
+
+Maturation never writes F/C, W or 59..63. Certification never rises because a
+rival was ruled out.
+
+## Generative reuse is not independent evidence
+
+A resolved relation may become a premise for further inference. It carries
+the **union of its parents' observation roots** and never adds one of its
+own. This is the same mask algebra as the existing
+`lance_graph_contract::revision` (`independent_roots`, `difference` against
+the ancestry). Two guards keep reuse from becoming self-support:
+
+1. Only resolved relations may serve as premises, so neither a direct cycle
+   nor an indirect one (A → B → C → A) can ground itself.
+2. A relation's own descendants are never its premises. Without this, a
+   relation reopened by counterevidence would be re-grounded by what was
+   derived from it, and circular support would survive the loss of its
+   external ground.
+
+Going round a grounded cycle adds nothing.
+
+## Status
+
+| claim | status |
+|---|---|
+| `pearl::hydrate` reads only the sealed chain; resolved relations never feed back | VERIFIED-IN-CODE |
+| test-local feedback makes A → B → C resolvable without a new observation | EXPERIMENTAL (probe) |
+| anchor ablation measures dependence (impact 2 on the chain) | EXPERIMENTAL (probe) |
+| an anchor turns a neutral observation into new Astonishment; the anchor stays at 0 | EXPERIMENTAL (probe) |
+| `mature`: ~2–3 ns median, 0 allocations (release, one container) | MEASURED |
+| bits 50..52 carry the code in production | BLOCKED — they are plasticity (`learn`, `MooreNars16`, `CausalEdgeV3`) |
+| production feedback | BLOCKED — needs D2 provenance ids |
+| bit migration and where plasticity lives | CONTRACT DECISION (operator) |