D-RMAP focus probe and native join over a relationship table; board rows for #1618–#1620

.claude/board/LATEST_STATE.md

@@ -1,4 +1,9 @@
-## 2026-10-19 — MORK / PathMap Schnellverschluss probe (in PR)
+## 2026-10-19 — main after #1620 (merged `dca8e7b1`)
+
+- **On main:** #1618 (`dbe0696f`, post-merge recovery for #1601–#1617 plus the remediation), #1619 (`8e004a20`, D-RMAP binding palette generation and the D-RMAP-1A §7 versioning crosswalk) and #1620 (`dca8e7b1`, the MORK/PathMap Schnellverschluss probe, D-MPS-0..7). All three have arc entries.
+- Contract inventory delta: none.
+
+## 2026-10-19 — MORK / PathMap Schnellverschluss probe (#1620, merged `dca8e7b1`)
 
 - New excluded crate `crates/mork-pathmap-probe` (own workspace, git deps on the AdaWorldAPI MORK and PathMap forks, results only) with CI job `mork-pathmap-probe.yml`; plan `mork-pathmap-schnellverschluss-v1.md`; entry `2026-10-19-mork-pathmap-schnellverschluss.md`; issue `ISS-MORK-LIBRARY-GLOBAL-ALLOCATOR`.
 - Contract inventory delta: none.

.claude/board/PR_ARC_INVENTORY.md

@@ -1,3 +1,24 @@
+## 2026-10-19 — lance-graph PR #1620 (merged `dca8e7b1`, branch `ccr-08f48cec-lbv9l4-mork-pathmap`, head `d514396c`, 5 commits) — MORK / PathMap Schnellverschluss: source switch, typed backend switch, differential probe
+
+- **Added:** excluded crate `crates/mork-pathmap-probe`. It has its own workspace and git deps on the AdaWorldAPI MORK (`de9312e3`) and PathMap (`ef02591`) forks; one `[patch]` line keeps a single `pathmap` package in the graph. `boundary.rs` is the only module naming either library. It holds `Backend::{Mork, PathMapZipper, Native}` and typed refusals: `SetOnly`, `NoAdmittedNativeJoin` and `Stale`. The crate also has 8 differential tests plus 1 unit test against a nested-loop oracle, the cost example, CI job `mork-pathmap-probe.yml` (default and `--features leapfrog`), plan `mork-pathmap-schnellverschluss-v1.md` (D-MPS-0..7; §7a ties `RelationView.generation` to the D-RMAP-1A §7 witness), entry `2026-10-19-mork-pathmap-schnellverschluss.md`, and `ISS-MORK-LIBRARY-GLOBAL-ALLOCATOR`.
+- **Locked:** none. Results-only probe: no production executor, no CE64 change, nothing wired into Quack/R2IL/Mask-RISC or Lance storage, no upstream file modified, no MORK source vendored (no licence declared, D-MPS-7). The native arm refuses under the 2026-10-16 no-gather ruling.
+- **Deferred / open:** D-MPS-2..5 upstream proposals (PathMap as a git/version dep; stable-build commit; no `jemalloc` global allocator in the library; cursor-provider generic `UnifyJoin`), none opened. D-MPS-6 BLOCKED until a sealed `RelationOrder` exists. ABA on a reused generation number is pinned as not detected; the D-RMAP-1A §7 `DurableRef` witness is the replacement.
+- **Confidence:** measured. The differential tests pass in both builds, and four disable runs each turned their guard's test red. Cost is from one release run per cell. A Codex review found that the first cost table compared a per-thought shared cost with a 64-thought rebuilt total and reported about 110×. The probe was fixed and re-run in `d514396c`: sharing a PathMap index over 64 thoughts saves 1.6× at n = 50 000 (1 556 vs 2 487 ms), and only the build is saved. The MORK query time includes the probe's text decode.
+
+## 2026-10-19 — lance-graph PR #1619 (merged `8e004a20`, branch `ccr-08f48cec-lbv9l4`, head `79baa045`, 3 commits) — D-RMAP palette generation in the binding; versioning derived from Thought Git and #1605
+
+- **Added:** D-RMAP-1A §7 (`.claude/plans/d-rmap-separate-tables-thought-palettes-v1.md`). It is an ownership crosswalk: map generation is a reading of the sealed RelationshipMap/RelationFacts `DurableRef` only; palette generation is the palette table's `DurableRef` once sealed, or the `ThoughtCommitId` of a speculative palette edit; and the thought-palette binding is `(ThoughtCommitId, palette generation)`. Focus generation and the index seal are retained witnesses, and authorization scope is a separate key. §7 also specifies falsifiers X1–X9, a cost/entropy comparison, and the Schnellverschluss note. The PR also adds gap notes in `thought-git-api-v1.md` and the causal-pickaxe plan, STATUS_BOARD row D-RMAP-1A-V, ENTROPY-MILESTONES M28, and the board entry `2026-10-19-d-rmap-thought-git-versioning.md`. `a7e075d3`: `ThoughtMapPaletteBinding` stores the palette generation it bound; `act-r.yml` also runs on `rust-toolchain.toml` changes.
+- **Locked:** none. Status PROPOSED. CE64 is unchanged. There is no new executor, store, type or per-fold commit, and the no-gather ruling stands.
+- **Deferred / open:** `ThoughtCommitId` must be never-reused, which #1605 does not say. A thought needs a dirty bit. A multi-dataset write needs one `DurableRef` per dataset, or a refusal (the measured instance is `versioned.rs::commit_encounter_round`). `DurableRef` has no type until CPG-00 runs. X1–X9 are not run.
+- **Confidence:** doc only. The code facts in §7.2 were read in source. CI caught a line-number citation (`citation-decay`); it was fixed in `79baa045` by citing symbols.
+
+## 2026-10-19 — lance-graph PR #1618 (merged `dbe0696f`, branch `ccr-08f48cec-lbv9l4`, head `b2c5453c`, 5 commits) — post-merge recovery for #1601–#1617
+
+- **Added:** board rows for #1601–#1617 (arc entries, STATUS_BOARD, plan index, entry `2026-10-19-post-merge-recovery-1601-1617.md`). Remediation: the ACT-R crate renamed to `lance-graph-act-r`, with `Refusal::StaleCandidate` and its own CI job; #1616's `OutgoingHop` quarantined behind the off-by-default `quarantined-gather-hop` feature (`ISS-GATHER-HOP-1616`); verified review findings applied across #1603–#1613; the D-RMAP `RelationTile` key carries the map generation (or index seal) and `MapPalette` gained its own palette generation; the thought-palette binding did not yet carry it.
+- **Locked:** the gather hop stays quarantined while the 2026-10-16 no-gather ruling holds.
+- **Deferred / open:** the palette generation of a thought's binding was left to the follow-up, which landed in #1619.
+- **Confidence:** the ACT-R refusal and the quarantine are pinned by disable-verified tests. The rest is documentation.
+
 ## 2026-10-19 — lance-graph PR #1617 (merged `47697c77`, branch `design/relationship-map-implicit-register-v1`, head `55be2fed`, 7 commits) — D-RMAP: separate RelationshipMap tables and thought-selected palettes
 
 - **Added:** plans `.claude/plans/d-rmap-implicit-relationship-register-v1.md` (D-RMAP-1), `d-rmap-separate-tables-thought-palettes-v1.md` (D-RMAP-1A, the controlling contract for ownership and CE64 opacity) and `d-rmap-2-context-tile-algebra-probe-v1.md` (D-RMAP-2, a preregistered GraphBLAS vs BF16 tile comparison); entry `2026-10-19-d-rmap-implicit-relationship-register.md`; entries README row. Nothing implemented.

.claude/board/STATUS_BOARD.md

@@ -24,6 +24,8 @@ These PRs merged with no board rows. Every row below is a PROPOSAL (documents an
 | D-RMAP-1 | implicit RelationshipMap and focus-relative registers (`d-rmap-implicit-relationship-register-v1.md`) | PROPOSAL (superseded on ownership by D-RMAP-1A) | falsifier matrix F0–F10 in the plan |
 | D-RMAP-1A | separate map tables, borrowed projections, thought-selected palettes (`d-rmap-separate-tables-thought-palettes-v1.md`) | PROPOSAL; generation witnesses corrected 2026-10-19 (§6a) | refusal on map regeneration, palette-slot reassignment, focus-slot reuse, incarnation reuse, stale tile, cross-tenant change |
 | D-RMAP-1A-V | versioning crosswalk: map/palette/binding generations derived from D-TGA `ThoughtCommitId` + #1605 `DurableRef`; focus generation and index seal retained (`d-rmap-separate-tables-thought-palettes-v1.md` §7) | PROPOSED; no code, no type; falsifiers specified, none run | X1–X9 in §7.4 (cross-dataset, compaction, palette ABA, slot reuse, uncommitted change, rebase, authorization, squash+replay, deletion/DRed) |
+| D-RMAP-1A-P | focus-bound RelationshipMap probe over existing sorted/trie indexes: CSR, 256×256 tile via mask-risc, incremental rebind, thought deltas, place masks, postings (`d-rmap-separate-tables-thought-palettes-v1.md` §8; `crates/mork-pathmap-probe/src/rmap.rs`) | MEASURED, probe only; nothing adopted | 12 tests vs scan oracle; 11 guards disable-verified; break-even 76/11/1.8 reuses (sparse d2 / d8 / dense) |
+| D-MPS-8 | native join over a versioned relationship table: R and S laid out by `y` slot, leapfrog intersection as one mask-risc `And`, per-side generations and reseal, per-thought filter planes (`mork-pathmap-schnellverschluss-v1.md` §7b; `crates/mork-pathmap-probe/src/join.rs`) | MEASURED, probe only; not admitted (D-RMAP-1 BIND membrane review) | 8 join tests + native arms in the differential suite vs the oracle; 7 guards disable-verified; cyclic bodies open |
 | D-RMAP-2 | context-declaration reuse and GraphBLAS/BF16 tile probe (`d-rmap-2-context-tile-algebra-probe-v1.md`) | PROPOSAL; not run | arms 1–5 in the plan; 297× (D-GTM-0j) is not evidence |
 | D-SRAM-0 | inventory Soufflé AST/ast2ram/RAM passes against D-ASM/D-RA/D-TGA owners | Queued | evidence per mapped pass; no code imported |
 | D-SRAM-1 | bound AST → RAM-like schedule for pure conjunctive selection | Queued | same answers and read sets on hand-built fixtures; unsupported ops typed-refused |

.claude/board/entries/2026-10-19-d-rmap-focus-probe.md

@@ -0,0 +1,17 @@
+# D-RMAP focus probe: a focus-bound RelationshipMap over existing indexes (2026-10-19)
+
+**Status:** MEASURED (D-RMAP-1A-P), probe only. Record: `.claude/plans/d-rmap-separate-tables-thought-palettes-v1.md` §8. Code: `crates/mork-pathmap-probe/src/rmap.rs`, `tests/rmap.rs`, `examples/rmap_cost.rs`.
+
+## Measured
+
+- A 256-slot focus over 65 536 identities, prepared into a slot-space CSR (and a 256×256 tile per predicate), answers 1–3-hop scoped reach through the existing mask-risc executor with no gather and no copied fact. 12 tests equal a full-slice scan; 11 guards disable-verified.
+- Break-even reuse count (arithmetic from measured medians): 76 (sparse, degree 2), 11 (degree 8), 1.8 (dense). The tile beats CSR only on the dense relation.
+- Reporting escapes as source slots instead of identities cut the degree-8 CSR eval from 50.4 to 5.6 µs.
+- Incremental rebind beats full re-preparation for small k and loses at k = 256 (dense: from k ≥ 16).
+- 1/8/64 thoughts over one shared tile: flat per-eval cost; 32 B per thought as sparse edits, 16 KiB as executor planes.
+- PathMap: prefix seek beats the sorted run on sparse relations; trie build is 4–6× slower and keys are ≥ 3× the bytes; `restrict` is exact but slower than CSR preparation. MORK matching not exercised.
+- Place focus: zero-copy counting over canonical lanes is 20–24 µs (word-granular gate); a 3 072 B register copy counts in ~1–2 µs.
+
+## Open
+
+- Lance-backed `DurableRef` and stable-row-id seals; cold-cache measurement; bit-granular `under` gating; any gather admission.

.claude/board/entries/2026-10-19-native-join-relationship-table.md

@@ -0,0 +1,15 @@
+# Native join over a versioned relationship table (2026-10-19)
+
+**Status:** MEASURED (D-MPS-8), probe only; not admitted. Record: `.claude/plans/mork-pathmap-schnellverschluss-v1.md` §7b. Code: `crates/mork-pathmap-probe/src/join.rs`.
+
+## Measured
+
+- `Q(x,y,z) :- R(x,y), S(y,z)` with R and S laid out by the slot of `y` in one shared domain. The leapfrog `y`-level intersection is one mask-risc `And` over two presence planes; enumeration reads both sides at the surviving slot only. No gather. Native now answers in the differential suite, for set and bag semantics.
+- n = 50 000: prepare 2.7 ms (PathMap 14.7, MORK 22.9). Set answers into a map 7.4 ms (PathMap 26.1, MORK 144). Streamed answers 0.33 ms; bag count 36 µs; joint plane 0.15 µs.
+- One dataset resealed: 0.21 ms to re-prepare its side vs 2.47 ms for the table. 64 thoughts with their own `y` filter planes: 5.2 µs each over one shared table.
+- The presence-plane `And` is an answer only as the existence mask; for enumeration and counts it prunes. Seven guards disable-verified.
+
+## Open
+
+- Cyclic bodies (triangle) need two variables co-addressed at once; chain joins over large domains would need a forbidden scatter.
+- Admission of the global-to-local binding as a membrane (D-RMAP-1 BIND review).

.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.
 
-372 entries, 2026-08-06 .. 2026-10-19.
+374 entries, 2026-08-06 .. 2026-10-19.
 
 | date | entry id | finding | file |
 |---|---|---|---|
 | 2026-10-19 | `post-merge-recovery-1601-1617` |  | [2026-10-19-post-merge-recovery-1601-1617.md](2026-10-19-post-merge-recovery-1601-1617.md) |
+| 2026-10-19 | `native-join-relationship-table` |  | [2026-10-19-native-join-relationship-table.md](2026-10-19-native-join-relationship-table.md) |
 | 2026-10-19 | `mork-pathmap-schnellverschluss` |  | [2026-10-19-mork-pathmap-schnellverschluss.md](2026-10-19-mork-pathmap-schnellverschluss.md) |
 | 2026-10-19 | `D-RMAP-1A` |  | [2026-10-19-d-rmap-thought-git-versioning.md](2026-10-19-d-rmap-thought-git-versioning.md) |
 | 2026-10-19 | `D-RMAP-1` |  | [2026-10-19-d-rmap-implicit-relationship-register.md](2026-10-19-d-rmap-implicit-relationship-register.md) |
+| 2026-10-19 | `D-RMAP` |  | [2026-10-19-d-rmap-focus-probe.md](2026-10-19-d-rmap-focus-probe.md) |
 | 2026-10-17 | `thinking-tables-wiring` |  | [2026-10-17-thinking-tables-wiring.md](2026-10-17-thinking-tables-wiring.md) |
 | 2026-10-17 | `D-ASM-11` |  | [2026-10-17-pgring-idea-catalogue.md](2026-10-17-pgring-idea-catalogue.md) |
 | 2026-10-17 | `epistemic-frontier-clicks` |  | [2026-10-17-epistemic-frontier-clicks.md](2026-10-17-epistemic-frontier-clicks.md) |

.claude/plans/d-rmap-2-context-tile-algebra-probe-v1.md

@@ -5,6 +5,8 @@
 
 > **PRIOR-ART CORRECTION (2026-10-19, precedence over earlier wording):** The operator already reverse-engineered BlasGraph, MKL-like backends, BF16 16x16, AMX, VNNI and mask-GEMM interop in AdaWorldAPI/ndarray and AdaWorldAPI/lance-graph. **Do not start a new GraphBLAS/MKL implementation, import production FFI or redo settled kernel comparisons.** The unproven subject is the separate RelationshipMap tables, zero-copy index binding and thought-selected map palettes. Use existing native kernels and probes as references; any external GraphBLAS implementation is optional independent oracle only.
 
+> **Partial answer (2026-10-19):** the first "updated test priority" item (bind a focus, compare direct indexed seek with a repeated local Boolean fold, focus churn, external-neighbour escapes, 1/8/64 thoughts) is measured in D-RMAP-1A §8 for in-memory warm fixtures. The GraphBLAS and BF16/INT8 arms here remain unrun.
+
 ## 0. Existing implementation and measurements: mandatory reuse
 
 | Precedent | Source | What is already present |