AdaWorldAPI · AdaWorldAPI · Aug 18, 2026 · Aug 17, 2026 · Aug 17, 2026 · Aug 17, 2026
diff --git a/.claude/board/EPIPHANIES.md b/.claude/board/EPIPHANIES.md
@@ -4,6 +4,70 @@
 > `**Status:**`/`**Confidence:**` line. A correction gets its own new,
 > dated entry that references the one it corrects — the storno rule.
 
+## 2026-08-18 — E-LGJ-ERGONOMICS-MUST-NOT-LEAK-INTO-CURRENCY-1 (STORNO, operator-ruled, council-ratified)
+
+**Status:** RULED — operator CORRECTION WAVE + RULING CLARIFICATION +
+A1 ARCHITECTURE RULING (all 2026-08-18), ratified by the full 5+3
+council (spec v3: `.claude/plans/mask-native-navigation-correction-v1.md`).
+**Confidence:** Operator-locked.
+
+### What this corrects
+
+`Graph.java:18-19`'s claim that the `long[]` frontier is *"a deliberate
+simplification for this consumer example, not a workaround."* That
+sentence is VALID for the bounded fixture it shipped in — and INVALID
+as target precedent. The entries below in this log (the graph-wave
+entries of 2026-08-18) accurately recorded the decision as made; what
+they could not know is that the currency itself was drift. The
+sharpest evidence that fixtures calcify: `GraphHopTest:459`'s own
+reflective allowlist had begun ENFORCING `long[]` as the sanctioned
+non-scalar egress — the guard test was defending the drift.
+
+### What PR #18 DID prove (preserved, not deleted)
+
+Membrane-surviving edge data (the 10/19/29 pinned hop counts reproduced
+cross-language), falsifiable hop semantics (set-equality caught a +4
+decode corruption a count check missed), measurable crossing cost
+(first-hop 2 / steady-state 1 — measured, then pinned), and a viable
+fluent vocabulary (`from`/`hop`/`minus`/`count`). It remains the SCALAR
+REFERENCE ORACLE — both independent BFS transcriptions stay verbatim.
+
+### What it did NOT prove
+
+That `long[]` is a frontier currency, that `TreeSet<Long>` is a query
+substrate, or that per-row payload reads are a target execution model.
+**Zero serialization does not imply zero-copy semantic navigation — a
+`long[]` of selected row IDs is still a materialised population.**
+
+### The ruling (the law this repo now carries)
+
+The three currencies — ClassView (meaning, late-bound provider) /
+WideFieldMask–FieldMask (facet participation) / Mask (population) —
+come from `lance-graph-contract`, THE semantic law; the contract split
+from the engine is blast-radius containment, NOT semantic optionality.
+The correction is one architectural closure: contract →
+ClassView/FieldMask/WideFieldMask → mask-native population ops →
+lgj-abi → Java ergonomic facade, complete only when Java ergonomics
+navigate the contract-governed substrate without row hydration. The A1
+ARCHITECTURE RULING extends the same law to COMPUTE: the 64K parallel
+SoA / deterministic-landing / batch-version model is inherited as
+architecture (parallel compute PROVEN in-tree via the EXP-KIA
+concurrency probe; the ~125/233 ms figures are operator-measured with
+the receipt currently out-of-tree; the batch-publication sole writer is
+shipped but not yet production-wired) — with the landing-key Rubicon:
+CastId/arrival order never becomes semantic, and the generic parallel
+write API stays BLOCKED until landing identity is arrival-order
+independent. Root `CLAUDE.md` (created with this entry) is the policy
+guard; wave D-LGJ-W8 (STATUS_BOARD) is the execution record.
+
+### Honest note
+
+The drift was self-documented at every step by this repo's own board
+hygiene — the D1 ruling, the javadoc, this log — which is exactly why
+it was correctable by audit instead of archaeology. The failure mode
+was not silence; it was a locally-reasonable decision calcifying into
+precedent because its own guard test enforced it.
+
 ## 2026-08-18 (final for now) — the graph wave's own crossing-cost assumption was wrong, measured and fixed before shipping
 
 **Status:** FINDING, caught during orchestrator review of dispatched

diff --git a/.claude/board/LATEST_STATE.md b/.claude/board/LATEST_STATE.md
@@ -1,3 +1,33 @@
+## 2026-08-18 — D-LGJ-W8 A3 FREEZE: the mask-native correction is ruled, specced, council-ratified; root CLAUDE.md created (PR-0)
+
+The operator's CORRECTION WAVE + RULING CLARIFICATION + A1 ARCHITECTURE
+RULING landed and went through the full supervision ladder: A0 six-lens
+audit (drift CONFIRMED — `Graph`'s `long[]`/`TreeSet` currency is
+precedent-drift; `View.where` exonerated; trades/bricks clean), A1
+combined spec (mask correction + the 64K parallel-SoA compute
+architecture, grounded in a 44-finding savant pass over the spec and a
+37-finding audit of the lance-graph spine), operator A2 verdict, three
+brutal reviewers (1 P0 — model identifiers in the to-be-committed spec
+— resolved by de-naming to roles; 6 P1 applied, including the §9
+axis-split at the arrival-order leak and the G2/G11 gate repairs),
+v3 RATIFIED. This commit freezes: spec v3 + root `CLAUDE.md` (the
+mask-native policy guard — first root CLAUDE.md this repo has) +
+`E-LGJ-ERGONOMICS-MUST-NOT-LEAK-INTO-CURRENCY-1` storno + the
+D-LGJ-W8 STATUS_BOARD row.
+
+**Correction of record for the entry below:** Graph #18 remains a VALID
+correctness fixture and becomes the preserved scalar oracle; its
+`long[]`/`TreeSet` implementation is scaffolding pending D-LGJ-W8's
+migration (Mask frontier, contract-governed ClassView/WideFieldMask,
+`lgj_hop`/`lgj_mask_andnot` at ABI minor 4). Trades/Bricks remain valid
+— their execution is already lazy/fused/mask-oriented. Implementation
+order: ndarray PR-N first (mask_andnot + simd.rs re-export), then
+PR-W8a (contract dep + provider + 2 ABI symbols), then PR-W8b (Java
+facade + Graph migration), each with same-commit board artifacts per
+the spec's §3.10 gate columns. The GridLake/compute surface is
+specified, NOT implemented, and hard-gated on arrival-order-independent
+landing identity.
+
 ## 2026-08-18 (final for now) — graph consumer wave landed: 2 workers, 43/43, one real crossing-cost finding caught before shipping
 
 Dispatched G1 (`Graph`/`Edge`, the traversal facade) and G2 (`GraphHopTest`,

diff --git a/.claude/board/STATUS_BOARD.md b/.claude/board/STATUS_BOARD.md
@@ -1,4 +1,15 @@
-## lgj-vertical-slice-v1 — the first Panama×Valhalla×ndarray::simd proof (PRE-REGISTERED 2026-08-17)
+## mask-native-navigation-correction-v1 — D-LGJ-W8 (operator CORRECTION WAVE, 2026-08-18)
+
+Plan: `.claude/plans/mask-native-navigation-correction-v1.md` (spec v3,
+council-ratified). Storno: `EPIPHANIES.md`
+`E-LGJ-ERGONOMICS-MUST-NOT-LEAK-INTO-CURRENCY-1`. Root `CLAUDE.md` =
+the policy guard, created at the A3 freeze.
+
+| D-id | Deliverable | Status |
+|---|---|---|
+| D-LGJ-W8 | Mask-native navigation correction + 64K-compute architecture inheritance. Gate ladder: AUDIT → SPEC → COUNCIL → FREEZE → SUBSTRATE (ndarray PR-N + lgj-abi PR-W8a) → FACADE (PR-W8b) → GRAPH MIGRATION → FALSIFIERS → POLICY | **AUDIT ✓** (A0 6-lens, drift CONFIRMED) → **SPEC ✓** (combined mask + compute, two audit fleets: 44 savant findings + 37 spine findings) → **COUNCIL ✓** (operator A2 verdict + 3 reviewers: 1 P0 resolved — model identifiers de-named from the spec; 6 P1 + ~20 P2 applied; v3 ratified 2026-08-18) → **FREEZE: this commit (PR-0)** → SUBSTRATE/FACADE/MIGRATION queued per spec §3.10 |
+
+
 
 Plan: `.claude/plans/lgj-vertical-slice-v1.md`. Every D-id below maps 1:1 to
 a phase in that plan and to a Phase-tracking task in this session's task