D-PUZZLE-0 step 2: crosswords on the same population algebra as Sudoku

.claude/board/STATUS_BOARD.md

@@ -48,7 +48,7 @@ Plan: `.claude/plans/2026-10-06-global-sudoku-replayable-orchestration-v1.md`. O
 | **D-EPI-CANON-0** | Canonical joint `EpistemicState5` reading of bits 59..63 in the contract | Shipped (#1381) | `contract::epistemic_state5`: Cartesian `Topology2 × Certification3`, `raw5 = t \| c << 2`, 24 meaningful / 8 reserved; per-factor facts; declarations + `project_state5`; `CausalEdge64::{epistemic_raw5, with_epistemic_raw5}`; non-declared-reading writers deprecated |
 | **D-EPI-POP-0** | Facts → 32-state population: the missing operator between one edge's coordinate (u5) and a semantic condition over all coordinates (u32) | Shipped (#1381) | `contract::epistemic_state5::{Population, facts_population, EpistemicState5::bit}`: `const fn`, no DSL / query object / planner. Tests pin exact sets (`CAUSES` = {20..23}, `IND_UNKNOWN \| RELATED` = {10,14,18,22}, `IND_KNOWN \| SUPPORTS` = {13,17,21}, `RELATED` = 8..23), operator == state enumeration over all 2048 requirements with no reserved code, conjunction == intersection; 2 disable runs red (any-of semantics; reserved read as fact-less) |
 | **D-EPI-LEGACY-DEPROJECT-0** | Deproject the remaining `ReasoningBand` / CE64-`TrustTexture` consumers: each goes to an exact coordinate, a `facts_population(..)`, MUL, or another subsystem (e.g. counterfactual context) — never a 1:1 replacement enum | Shipped (#1381) | `entries/2026-10-07-epi-legacy-deprojection.md`: 6 ReasoningBand caller sites (1 certification question, 1 non-interference witness → both exact coordinates; 2 context/obsolete → compat; 2 refusal witnesses), 0 production CE64 `TrustTexture` consumers, 4 `TrustTexture` types with no bridge; `EdgeRole::band` removed; hinge + V3 lift migrated; 4 code loops → `facts_population` |
-| **D-PUZZLE-0** | Sudoku / crossword / chess on the same population algebra under domain laws | In progress | step 1 Sudoku at scale (`cognitive-shader-driver/examples/sudoku_population_fold_probe.rs`, `entries/2026-10-07-puzzle-0-sudoku-population-fold.md`): 1,000,066 claim edges, counts equal 3 ways + solver counters; filter fold 0.35–0.39 ns/edge, 32-bin histogram 0.93–1.07, decode oracle 3.8–4.0 (one machine). Next: crossword (lance-graph), chess (`stockfish-rs`, GPL) |
+| **D-PUZZLE-0** | Sudoku / crossword / chess on the same population algebra under domain laws | In progress | step 1 Sudoku shipped (#1382). Step 2 crossword (`crossword_population_fold_probe.rs`, `entries/2026-10-07-puzzle-0-crossword-same-algebra.md`): domain-free half moved to `examples/shared/population_fold.rs` (fence test, Sudoku counts unchanged); 1,000,052 crossword edges, counts equal 3 ways; filter 0.30–0.43 ns/edge in both domains. Step 2b real words (`crossword_real_words_probe.rs`, `entries/2026-10-07-puzzle-0-crossword-real-words.md`): English = DeepNSM v1 4096 + academic_20k (committed), German = DeReKo top 20k only from `DEREKO_PATH` (CC BY-NC, nothing committed); ~1.0M edges each, counts equal 3 ways. Next: chess in `stockfish-rs` (GPL) — how it reaches the shared half is open |
 | **D-EPI-MIG-0** | Migrate readers/writers to the canonical reading; #1378 → conformance | Shipped (#1381) | `entries/2026-10-07-epi-mig-0.md`: affordance law (per-factor compile) + P7a (identity packing) + 6 probes + dismech + V3 lift; conformance census 24/0/8 (was 3/5/16); 3 historical-band probes cannot migrate without invention |
 | **D-EPI-CONFLICT-0** | (superseded by D-EPI-MIG-0: probe converted to conformance) Pin the disagreement between the D-ACR-7-declared P7a band and the affordance law on CE64 bits 59..63 | Superseded | **Current:** the same probe is the conformance test, census 24 agree / 0 disagree / 8 reserved under the Cartesian V1 (D-EPI-MIG-0). **HISTORICAL — the pre-Cartesian pins below no longer describe the code:** `cognitive-shader-driver/examples/epistemic_reading_conflict_probe.rs`: contract-projected `Causes` reads as affordance code 20 `Related` (topology 0) or refuses (1..3), `COUNTERFACTUAL_PROBE` never eligible; `CAUSES` codes 1 and 17 sit on P7a `Open` / `CausalCandidate`; codes 25, 30 on P7a reserved bands; census over 24 cells = 3 agree, 5 disagree, 16 refuse; the affordance law measures where the contract refuses (undeclared / band `Absent`). Both readings run from shared code (`shared/affordance_law.rs`, and P7a's reading moved unchanged into `shared/certification_reading.rs`, used by `relational_certification_probe` too), never copied. Pins of current state, to become agreement tests when the issue closes (`ISS-EPISTEMIC-READINGS-DISAGREE-ON-BITS-61-63`); 5 tests, 6 disable runs red (3 on the affordance codebook, 3 on P7a's: `Contributes` entailment, a reserved code admitted, `Causes` recoded) |
 | **D-MBX-CUTOVER-0** | ShaderDriver on MailboxSoA without BindSpace | In PR | `entries/2026-10-06-shaderdriver-bindspace-cutover.md`: every ShaderDriver → BindSpace edge classified (production state / shared resource / fallback / test oracle / fossil / dead). In `run()` the only mailbox-mode BindSpace read was the ontology; outside it, metrics and the builder read the singleton and `backing()` silently fell back for any mailbox not under id 0. Cutover: optional BindSpace, explicit mailbox selection with no fallback, ontology on the driver, metrics from the selected substrate; no temporal change. `tests/mailbox_cutover.rs` 8 tests, 8 disable runs red |

.claude/board/entries/2026-10-07-puzzle-0-crossword-real-words.md

@@ -0,0 +1,76 @@
+# 2026-10-07 — D-PUZZLE-0-XWORD-REAL (D-PUZZLE-0 step 2b): real-word crosswords (DeepNSM vocabularies, optional DeReKo)
+
+**Probe:** `cognitive-shader-driver/examples/crossword_real_words_probe.rs`,
+over `examples/shared/population_fold.rs` (unchanged).
+**Status:** MEASURED (this machine), TEST-PINNED (11 tests, 6 disable runs red).
+
+## What changed from step 2
+
+The step-2 crossword drew its words from a synthetic 3-letter alphabet. This
+probe fills the same 5×5 template from a real vocabulary. The dictionary is
+every word of a slot's length in that vocabulary.
+
+- **English:** exactly the two vocabularies DeepNSM reads, both committed
+  under `crates/deepnsm/word_frequency/`. These are v1's 4096-word table
+  (`word_rank_lookup.csv`, ranks 1..=4096) and v2's academic list
+  (`academic_20k.csv`), taken as their union. Alphabetic ASCII only, scored by
+  the larger COCA frequency. `word_forms.csv` is not a source. Dictionary:
+  1,250 four-letter, 1,697 five-letter words.
+- **German:** the 20,000 most frequent DeReKo-2014 forms, after excluding
+  proper nouns (`NE`) and non-alphabetic forms and summing frequencies per
+  lowercase form. Read only when `DEREKO_PATH` names the file at run time and
+  skipped otherwise. DeReKo is © IDS Mannheim, CC BY-NC 3.0. No German word,
+  list or derived artifact is committed; this entry carries counts only.
+  Dictionary: 802 four-letter, 1,466 five-letter words.
+
+The first version of this probe (`cc887e3`) used the top 1000 `word_forms.csv`
+surfaces per length. It was replaced on review because that list is not a
+vocabulary DeepNSM uses.
+
+An instance is a random fill of the template from the dictionary. Given slots
+are then added in random order until the dictionary admits exactly one fill
+(backtracking). Law, claims, questions, fold and oracle are the step-2 ones.
+
+## Measured (release, `avx2=true avx512f=false`, median of 7, one run)
+
+| | English (v1 4096 + academic) | German (DeReKo top 20k) |
+|---|---|---|
+| edges | 1,005,146 | 1,000,074 |
+| instances | 1,209 | 1,390 |
+| build time | 7.4 s | 6.7 s |
+| given | 5,370 | 6,118 |
+| forced | 1,278 | 1,400 |
+| entailed, not yet forced | 5,442 | 6,382 |
+| live candidates | 993,056 | 986,174 |
+| filter fold, ns/edge | 0.35 | 0.36 |
+| histogram, ns/edge | 2.23 | 2.10 |
+| decode oracle, ns/edge | 4.01 | 3.79 |
+
+Every count is equal three ways (filter, histogram, decode) and to the
+instances' own counters. Every instance holds exactly 10 asserted claims.
+
+Givens needed for uniqueness (givens: instances), EN:
+`1:23 2:246 3:208 4:204 5:161 6:129 7:109 8:88 9:41`. DE:
+`1:53 2:250 3:263 4:208 5:200 6:152 7:132 8:80 9:52`. These are minimal
+along a random order, not globally minimal.
+
+The filter and the oracle cost the same as for Sudoku and the synthetic
+crossword. The histogram is slower again (2.1–2.2 ns, against 1.7 synthetic
+and 1.0 Sudoku). That fits the step-2 hypothesis: here 98.6–98.8 % of edges
+share one code, against 77 % there. It fits, but nothing here measures the
+cause.
+
+## Falsifiers (disable-verified red)
+
+- Uniqueness check returning 1 always → `instances_are_unique_and_givens_are_needed`.
+- Index ignoring fixed letters → 4 tests, including `the_index_agrees_with_a_scan`.
+- `NE` rows kept → `dereko_rows_are_parsed_by_tag_and_summed`.
+- 4096 rank cut removed → `the_english_dictionary_is_deepnsm_v1_plus_academic`
+  (words ranked past 4096 and absent from the academic list leak in).
+- Academic list dropped → the same test (academic-only words missing).
+- German 20k truncation removed → `german_keeps_the_top_20k_forms`.
+
+## Open
+
+- German runs only where the DeReKo file is present; CI covers English only.
+- Chess in `stockfish-rs` and how it reaches the shared fold: still open.

.claude/board/entries/2026-10-07-puzzle-0-crossword-same-algebra.md

@@ -0,0 +1,74 @@
+# 2026-10-07 — D-PUZZLE-0-XWORD (D-PUZZLE-0 step 2): crosswords on the same algebra as Sudoku
+
+**Probes:** `cognitive-shader-driver/examples/crossword_population_fold_probe.rs`,
+`sudoku_population_fold_probe.rs`, both over
+`examples/shared/population_fold.rs`.
+**Status:** MEASURED (this machine), TEST-PINNED (fence + 6 domain tests, 3
+disable runs red).
+
+## What "same algebra" means here, and how it is checked
+
+The domain-free half moved out of the Sudoku probe into
+`shared/population_fold.rs`: the propagation reading (given / forced /
+entailed-not-forced / candidate), the five questions as `facts_population`
+masks, the filter / histogram / decode folds, the group fold, the declaration
+gate. Each domain file supplies only its claims, its law and its counters.
+
+- **Fence:** a test fails if `sudoku`, `crossword`, `chess`, `digit`, `slot`,
+  `cell`, `letter`, `grid`, `board` or `move` appears in the shared half.
+  Disable run: one planted word → red.
+- **Refactor checked by output:** the Sudoku probe's five counts are
+  identical before and after the move (1,000,066 edges).
+
+## The crossword domain
+
+A 5×5 grid, two black squares, ten slots, every white square in two slots.
+Fills are random over a 3-letter alphabet; the dictionary is the solution
+words plus 10 distractors per length; two slots are given. An instance is kept
+only with exactly one solution (backtracking). The law is the forced single: a
+slot whose consistent dictionary words narrow to one is placed.
+
+## Measured (1,000,052 edges, 22,455 instances)
+
+| question | edges |
+|---|---|
+| every asserted claim | 224,550 |
+| given | 44,910 |
+| forced, chain known | 13,745 |
+| entailed, not yet forced | 165,895 |
+| live candidates | 775,502 |
+
+Every count is equal three ways and to the instances' own counters. Every
+instance holds exactly 10 asserted claims. The instance build takes 7.3 s
+(rejection sampling with backtracking).
+
+Timing, release, `avx2=true avx512f=false`, same 4-core shared Xeon, median of
+7, three runs:
+
+| path | crossword ns/edge | Sudoku ns/edge |
+|---|---|---|
+| one population, filter | 0.30 – 0.43 | 0.35 – 0.39 |
+| 32-bin histogram | 1.67 – 1.75 | 0.93 – 1.07 |
+| per-edge decode (oracle) | 3.9 – 4.1 | 3.8 – 4.0 |
+
+The filter and the oracle cost the same in both domains: they read five bits
+and do not care what the claim is about. The histogram costs more on the
+crossword lane, where 77 % of edges share one code. Repeated increments of
+one bin are the likely cause; this is a hypothesis, not measured.
+
+## What differs, and it is the law, not the algebra
+
+Forcing is rare in crosswords: 13,745 of 177,640 non-given slots, against
+Sudoku's naked singles, which solve the base puzzle completely. A slot rarely
+narrows to one word without stronger propagation (arc consistency over
+crossings), so most true words stay `IndirectUnknown × Causes`. That is a
+statement about the law. The questions and folds did not change.
+
+## Next
+
+Chess in `stockfish-rs` (GPL; shakmaty supplies legality), using the same
+shared half. The open point is how it gets there. The module is an
+example-local file in lance-graph and is not part of `lance-graph-contract`.
+Copying it into stockfish-rs would make a second copy (a fence violation in
+spirit); promoting it into the contract is a contract change. The decision is
+open.

.claude/board/entries/README.md

@@ -25,13 +25,15 @@ 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.
 
-247 entries, 2026-08-06 .. 2026-10-07.
+249 entries, 2026-08-06 .. 2026-10-07.
 
 | date | entry id | finding | file |
 |---|---|---|---|
 | 2026-10-07 | `tinker-janus-fold-harvest` | TinkerPop bulk = K (GroupReduce), ONE_BULK = S, GValue pinning = bundle invalidation; JanusGraph slice = OrderedLaneWitness→Range (30–38×); no new V4 op | [2026-10-07-tinker-janus-fold-harvest.md](2026-10-07-tinker-janus-fold-harvest.md) |
 | 2026-10-07 | `selector-seq-aware-coverage` |  | [2026-10-07-selector-seq-aware-coverage.md](2026-10-07-selector-seq-aware-coverage.md) |
 | 2026-10-07 | `D-PUZZLE-0` |  | [2026-10-07-puzzle-0-sudoku-population-fold.md](2026-10-07-puzzle-0-sudoku-population-fold.md) |
+| 2026-10-07 | `D-PUZZLE-0-XWORD` |  | [2026-10-07-puzzle-0-crossword-same-algebra.md](2026-10-07-puzzle-0-crossword-same-algebra.md) |
+| 2026-10-07 | `D-PUZZLE-0-XWORD-REAL` |  | [2026-10-07-puzzle-0-crossword-real-words.md](2026-10-07-puzzle-0-crossword-real-words.md) |
 | 2026-10-07 | `gated-gather-bit-schedule` | Production gather is 3.8× slow from a per-row branch; bit-gated visitation wins or ties at every density; ordinal vectors never win (R1 holds) | [2026-10-07-gated-gather-bit-schedule.md](2026-10-07-gated-gather-bit-schedule.md) |
 | 2026-10-07 | `D-EPI-MIG-INVENTORY` |  | [2026-10-07-epi-mig-inventory.md](2026-10-07-epi-mig-inventory.md) |
 | 2026-10-07 | `D-EPI-CANON-0` |  | [2026-10-07-epi-mig-0.md](2026-10-07-epi-mig-0.md) |

crates/cognitive-shader-driver/Cargo.toml

@@ -153,6 +153,16 @@ test = true
 name = "sudoku_population_fold_probe"
 test = true
 
+# D-PUZZLE-0 step 2: crosswords on the same population algebra (shared fold
+# with the Sudoku probe); tests run under `cargo test`.
+[[example]]
+name = "crossword_population_fold_probe"
+test = true
+
+[[example]]
+name = "crossword_real_words_probe"
+test = true
+
 [dependencies]
 lance-graph-contract = { path = "../lance-graph-contract" }
 # Read-only ontology registry (Phase 7, v4 plan). default-features = false so the

crates/cognitive-shader-driver/examples/README.md

@@ -413,3 +413,32 @@ checked once per lane. Every count agrees three ways — population filter,
 ```bash
 cargo run --release -p cognitive-shader-driver --example sudoku_population_fold_probe
 ```
+
+## crossword_population_fold_probe.rs
+
+D-PUZZLE-0, step 2. Crosswords on the same population algebra as the Sudoku
+probe: ten-slot 5×5 grids with generated fills, distractor words, two given
+slots and a forced-single law, each instance checked to have exactly one
+solution. The probe holds only the domain; the reading, questions, folds and
+oracle come unchanged from `shared/population_fold.rs`, whose fence test
+fails if domain vocabulary enters the shared half.
+
+```bash
+cargo run --release -p cognitive-shader-driver --example crossword_population_fold_probe
+```
+
+## crossword_real_words_probe.rs
+
+D-PUZZLE-0 step 2b. The step-2 crossword with a real dictionary: every 4- and
+5-letter word of DeepNSM's English vocabularies (v1's 4096-word
+`word_rank_lookup.csv` plus v2's `academic_20k.csv`, both committed), and of
+the 20,000 most frequent DeReKo-2014 German forms, read only when
+`DEREKO_PATH` names the file (CC BY-NC 3.0; nothing derived is committed).
+Instances are random fills, givens added until the fill is unique. About 1.0M
+edges per language; every count agrees three ways. Board:
+`.claude/board/entries/2026-10-07-puzzle-0-crossword-real-words.md`.
+
+```
+cargo run --release -p cognitive-shader-driver --example crossword_real_words_probe
+DEREKO_PATH=/path/DeReKo-2014-II-MainArchive-STT.100000.freq cargo run --release -p cognitive-shader-driver --example crossword_real_words_probe
+```