planner: seal NestedBands (D-NXG-1/5) and ship the D-BLW-5 shape_rank payload (D-NXG-4)

.claude/board/AGENT_LOG.md

@@ -1,3 +1,19 @@
+## 2026-09-05 — D-NXG-4 → D-BLW-5 payload: two Sonnet workers on disjoint files, orchestrator-gated
+
+- **Why:** operator: "go ahead with D-NXG-4 into D-BLW-5". A Sonnet census first (`exec-runs/d-blw-5-design-main-thread.md`, doctrine §2, STATUS_BOARD): D-BLW-5 has NO shipped code and is PAUSED (operator 2026-08-05); its designed home is a `cycle-driver`-gated supervisor test where the planner is reachable and `ndarray` is not. That fixed the membrane split: DTO in the zero-dep contract, producer in the planner, transform in jc.
+- **W1 (Sonnet, contract):** `shape_rank.rs` + one `lib.rs` line from a 62-line spec; verified the 144-byte layout by field arithmetic. 6/6, clippy clean — committed first as `75f87646` with its LATEST_STATE delta in-commit.
+- **W2 (Sonnet, jc + planner):** `fisher_2z`/`fisher_2z_inv`/`FISHER_CLAMP_EPS` in `jc::stats` with a transcribed-formula parity test against helix (live cross-crate parity blocked: helix pins a git `ndarray`, jc a path one); `Z2_SCALE`, `quantize_2z`, `calibrate_equal_width`, `shape_rank` + 5 tests in the planner. W2 flagged its one guess (`RemeasureKey::new`) — correct call, it did not exist; fixed to a struct literal at the gate.
+- **Falsification kept:** the can-it-fire arm ("saturated vs quiet first frame land in different buckets") could not fire — both recordings open with silence and both whole-file statistics sit below the prior's support. Restated inside the support (prior min → 0, prior max → 15) and the out-of-support saturation pinned as E-NXG-22.
+- **Gates:** planner 428/428 + clippy clean, contract 1315/1315 + clippy clean, jc 141/141 (via `--manifest-path`, workspace-excluded), fmt clean on all three; jc clippy is red on `origin/main` in eight files none of which this arc touched, CI runs no jc clippy — TD row, not a fix here. Append-only + citation-decay gates local; supersession index last.
+- **Board:** E-NXG-22, STATUS_BOARD D-NXG-4 + D-BLW-5 rows, TECH_DEBT TD-JC-CLIPPY-RED-ON-BASE-1, LATEST_STATE delta (planner + jc half), plan §6.
+
+## 2026-09-05 — D-NXG-1 `NestedBands`: one Sonnet worker from spec, orchestrator-gated
+
+- **Why:** operator "Go ahead autoattended" + "Sonnet agents for grindwork" after the three probes merged. D-NXG-1 is a write-this-file-from-spec unit: bounded input (a 134-line spec naming every type, method and test), known output shape.
+- **Split:** orchestrator wrote the spec (types, signatures verified against `ndarray/src/simd_int_ops.rs` + `bitwise.rs` + `contract/src/thought_atoms.rs`, twelve named tests on the three real recordings); one Sonnet `general-purpose` worker wrote the 711-line file, edit-only, no cargo, and corrected the spec's `include_bytes!` path (three levels up, not four — verified with `ls`, reported). Orchestrator registered the module in `lib.rs` and ran every gate in the shared `target/`.
+- **Gates:** first run 10/12 — both failures were SPEC errors the worker faithfully reproduced (a pinned floor value one row over the target; a sign-assumed σ assertion). Both fixed in the tests to the strict definitions, recorded in E-NXG-21. One clippy `-D warnings` (test-only import) moved into the tests module. Final: `nested_bands` 12/12, planner lib suite green, clippy clean, fmt clean, append-only + citation-decay + plan-dids gates local, supersession index regenerated last.
+- **Board:** E-NXG-21; STATUS_BOARD D-NXG-1/3/5 rows; plan §6 rows.
+
 ## 2026-09-05 — PROBE-NXG-ROLL-1 + PROBE-NXG-FLOOR-1: orchestrator-only, three falsifications kept
 
 - **Why:** operator "fast forward" after #1178 merged. Plan §5 steps 3 and 4, the last two gates before rooms 4–8 may leave PROPOSAL.

.claude/board/EPIPHANIES.md

@@ -1,3 +1,78 @@
+## 2026-09-05 — E-NXG-22 — `shape × rank` is bounded by the prior's support: an out-of-support statistic saturates at the edge bucket
+
+**Status:** FINDING (pinned by `nested_bands::tests::shape_rank_from_real_lag1_autocorrelations`); D-NXG-4 SHIPPED into the D-BLW-5 payload.
+**Confidence:** High on the mechanism; the consequence for the D-BLW-5 loop is a design question left open on purpose.
+
+D-NXG-4 now has its first consumer: `NestedBands::shape_rank(observed, V₀)`
+produces `lance_graph_contract::shape_rank::ShapeRankPayload` — the doctrine's
+`shape₀ × rank₀`, frozen at V₀, with no `f64` field by construction. The
+transform lives in `jc::stats::fisher_2z` (ruled home of calibrated math), the
+ladder is `calibrate_equal_width` (the D-BLW-5 design's "equal-width in 2z"),
+and the remeasure guard is `RemeasureLedger::seal` (second write ERRORS).
+
+**The finding.** The pooled prior in the test is real: the lag-1 Pearson
+autocorrelation of 92 frames of speech, `r ∈ [0.9205, 0.9917]`, i.e.
+`2z ∈ [3.18, 5.48]`. The two other recordings' whole-file statistics are
+`2z ≈ 1.46` (saturated) and `1.67` (quiet) — far BELOW the prior's support.
+Both rank 0, and both are therefore indistinguishable from the prior's own
+minimum, which also ranks 0. The first version of the test asked the two
+out-of-support values to land in DIFFERENT buckets; they cannot, and the
+test was restated to what is true: inside the support the ladder
+discriminates (prior min → rank 0, prior max → rank 15), and outside it
+saturates. **`rank` carries no "how far out".** That is partly by design —
+the payload law forbids the raw scalar precisely so awareness cannot parrot
+a number — but the D-BLW-5 loop's "T-silence is a reportable null" and bloom
+criterion may need to know that an observation left the prior's support
+entirely. The candidate fix is one bit (`out_of_support`), not a scalar; it is
+a decision for the loop, which stays PAUSED, and is not made here.
+
+**Also measured, not claimed:** first frames of both recordings are silence
+(that is why the spec's "first frame" observed statistic was wrong — it
+measured leading silence, not the recording); `quantize_2z` at ×1024 keeps
+three decimals of 2z and the 16 equal-width boundaries over the speech prior
+differ in width by at most one unit. Gates: planner 428/428, contract
+1315/1315, jc 141/141, planner + contract clippy clean; jc clippy red on
+base and untouched by this arc (TD-JC-CLIPPY-RED-ON-BASE-1).
+
+## 2026-09-05 — E-NXG-21 — `NestedBands` sealed: the three probes became one type and twelve tests, and two of the probes' numbers were off by one row and by one sign
+
+**Status:** SHIPPED (`crates/lance-graph-planner/src/nested_bands.rs`, D-NXG-1; merge arm of D-NXG-5; room 3 closed by moments).
+**Confidence:** High — 12/12 tests on the same three real recordings the probes used, planner lib suite green.
+
+`NestedBandsBuilder::{new, budget_factor, calibrate, with_boundaries}` seals a
+`NestedBands` (version, boundaries, band masks, bucket masks, popcounts).
+Every method is `&self`; `split` and `merge` return NEW values under a new
+version and the original is asserted unchanged (data-flow rule, one writer at
+build). The top band is the universe by construction (E-NXG-18 is now a type
+invariant, not a probe caveat). `overflow()` is the budget test (E-NXG-19),
+`collapsed()` is the merge-on-collapse arm — it fires when speech boundaries
+meet the quiet recording and stays silent on the calibration epoch; `merge`
+removes one boundary and the merged popcount equals the pair's sum.
+`best_achievable_floor` is E-NXG-20's floor by bisection over
+`gt_i32_to_mask` + popcount, no sort.
+
+**Two corrections the type forced on the probes' numbers.** (1) The strict
+floor — the SMALLEST value whose exceedance is ≤ the rate — lands at 15 077 on
+speech, not the probe's 15 072: the probe's rank floor sat one row OVER the
+target (473/94 572 = 0.0050015) inside its ±1/n tolerance. A floor with a row of
+slack is not "never exceeds"; the type has no slack and the test pins both the
+literal and a sorted-copy reference. (2) The midpoint-σ estimator is not even
+sign-stable: with the stale top boundary it over-read by 12 % (E-NXG-17); with
+the last boundary as the top bucket's midpoint it under-reads by 7 %
+(3 525.8 vs 3 785.8). An open-ended top bucket has no midpoint, so the
+estimator has no principled value at all. The regression guard is now
+direction-agnostic (|Δ|/σ > 5 %), and room 3 closes the other way: σ is exact
+from the seal iff the seal stores two accumulators per bucket (`BucketMoment
+{count, sum, sumsq}`, 16 B/bucket) — `sigma_exact` matches the direct σ to
+1e-9. "Recoverable for free" was wrong; "recoverable for 256 bytes" is right.
+
+**Not done, on purpose.** D-NXG-3 (`bisect_column_by_mask`) lives inside
+`NestedBands::split`, not as a named `ndarray::simd` primitive: it is a
+memory-bound gather-and-popcount loop, not a lane op, and the W1a contract's
+all-backends parity requirement would be ceremony for a scalar. Promotion, if
+ever, is its own deliberate PR. No consumer is wired yet — the type exists, the
+first caller (D-NXG-4's `shape × rank` payload) is the next unit.
+
 ## 2026-09-05 — E-NXG-18 — the ladder's top band must BE the universe, or the histogram loses rows silently
 
 **Status:** FINDING (PROBE-NXG-ROLL-1 falsified its own first run).

.claude/board/LATEST_STATE.md

@@ -1,3 +1,17 @@
+## 2026-09-05 — branch (D-NXG-4 → D-BLW-5, producer half): the payload has a producer and a transform — INVENTORY DELTA
+
+- ADDED `lance_graph_planner::nested_bands::{Z2_SCALE, quantize_2z}`, `NestedBandsBuilder::calibrate_equal_width` (the D-BLW-5 design's equal-width-in-2z ladder, beside the equal-mass `calibrate`), `NestedBands::shape_rank(observed, V₀) -> ShapeRankPayload` (asserts 16 bands). Five tests, one on a REAL pooled prior (92 speech frames' lag-1 autocorrelation).
+- ADDED `jc::stats::{FISHER_CLAMP_EPS, fisher_2z, fisher_2z_inv}` — the "2z" of plan §Stage A, forced copy of `helix::Similarity::hyperbolic_depth` with a transcribed-formula parity test (live cross-crate parity blocked by helix's git-pinned `ndarray`; recorded in the doc comment).
+- FINDING E-NXG-22: `rank` saturates at the edge bucket for out-of-support statistics; the loop-level fix (one bit) is left to D-BLW-5, which stays PAUSED.
+- TD-JC-CLIPPY-RED-ON-BASE-1 opened (pre-existing, untouched files, no CI clippy on jc).
+
+## 2026-09-05 — branch (D-NXG-4 → D-BLW-5, contract half): the payload-law DTO exists — INVENTORY DELTA
+
+- ADDED `lance_graph_contract::shape_rank` (zero-dep, std only): `SHAPE_BUCKETS = 16`; `ShapeRankPayload { shape: [u64; 16], rank: u8, version: u64 }` with `new` (asserts `rank < 16`), `mass`, `mass_below`, `prozentrang`, `rank_fraction`, `is_frozen_at`, and a `const _` guard `size_of <= 144`; `RemeasureKey { stat_id, arm, cohort, metric, dataset_version }` (`Ord`); `RemeasureError::AlreadySealed { key, sealed_version }`; `RemeasureLedger { seal (write-once, second write ERRORS, never overwrites), get, len, is_empty }`.
+- WHY here and not in the planner: the doctrine's `shape₀ × rank₀` is what crosses into awareness; the type has NO `f64` field by construction, which is the Goodhart guard the payload law asks for, enforced by shape rather than discipline. `ndarray` stays unreachable supervisor-side (D-BLW-5 design §c), so the DTO is plain numbers and the producer (`NestedBands::shape_rank`, planner, next commit) does the mask work.
+- STATUS: D-BLW-5 remains **PAUSED** as a loop (operator 2026-08-05); the operator relaunched ONLY the payload half on 2026-09-05 ("go ahead with D-NXG-4 into D-BLW-5"). The four arms, bloom criterion and supervisor test file are untouched.
+- Gates: contract `shape_rank` 6/6, clippy `--lib --tests -D warnings` clean, fmt clean. Written by one Sonnet worker from a 62-line spec, edit-only; orchestrator-gated.
+
 ## 2026-09-04 — two mechanical board gates land, and they were shipped WITHOUT their own hygiene — INVENTORY DELTA
 
 - ADDED `.claude/tools/append_only_gate.py` + `.github/workflows/append-only-gate.yml` (#1167) — 8 protected board files, compared against `git merge-base` (not the raw base ref, which reports a shrink whenever the base moved ahead). Absent-at-base = new, silent; deleted = FAIL; unresolvable ref = fail closed with a `fetch-depth: 0` hint. Verified at four levels: 7 self-test cases (3 fire / 4 silent), the real git path against `b67f1958`, an END-TO-END fire test (`ISSUES.md` shortened 2591 → 2541 lines → exit 1), and the fail-closed arm (`origin/main~200` → exit 1).