docs(WO57-78): campaign WO set

TODO.sota/57-a1-authority-completeness.md

@@ -0,0 +1,23 @@
+# WO57 — A1: authority-completeness audit
+
+Status: PLANNED · Plane A
+Goal: enumerate the UNIVERSE of authority romanization/transliteration
+systems and diff it against the 293 encoded in `maps/`. Every gap is
+either encoded or cited out-of-scope with an authority reference.
+Method:
+1. Build the catalog: UN WGRS (1968–2016 editions, ~25 systems),
+   BGN/PCGN current schedule (~55), ALA-LC full language tables
+   (~170 Romanization tables), ISO 9/233/259/3609/7098/843/9984/9985/
+   233-3, ICAO 9303, national systems (DIN 14640, RTGS, Korean RR
+   2000, HK Government, ROC, Thai Royal Institute…). Sources: the
+   authority documents already collected in interscript-references +
+   the ISO 24229 code convention.
+2. Extract the corpus's system codes from maps/*.isc names; normalize
+   per ISO 24229; join against the catalog.
+3. Emit coverage matrix (authority × system × status: encoded /
+   gap-encoded-possible / out-of-scope+reason) to
+   docs/coverage/COVERAGE.md; gaps that are encodable become maps
+   PRs (whole-document transcription law applies).
+Acceptance: coverage matrix published with zero unexplained gaps;
+enconding PRs opened for every encodable gap.
+Owner gates: none for the audit; map encodings are ordinary PRs.

TODO.sota/58-a2-invertibility.md

@@ -0,0 +1,17 @@
+# WO58 — A2: invertibility ledger + reverse maps
+
+Status: PLANNED · Plane A
+Goal: for every system, determine whether forward∘reverse = identity;
+publish the loss conditions (homophone mergers, digraph ambiguity,
+case conventions, vowel elision).
+Method:
+1. Script against the installed gem runtime: for each map with a
+   reverse variant (or a generated one), round-trip the map's embedded
+   tests + a synthetic Unicode sample per source script.
+2. Classify: lossless / lossy-with-convention / not-invertible; attach
+   minimal distinguishing examples.
+3. Where an authority publishes a reverse system: encode it (whole-
+   document law). Where the forward map is provably bijective:
+   generate the reverse from the grammar and validate.
+Acceptance: docs/invertibility/LEDGER.md covering all systems; reverse
+maps shipped for every losslessly invertible one; losses cited.

TODO.sota/59-a3-equivalence.md

@@ -0,0 +1,15 @@
+# WO59 — A3: system-equivalence checker + distinguishing corpora
+
+Status: PLANNED · Plane A
+Goal: mechanically decide/estimate whether two systems coincide, and
+produce the minimal test corpus that distinguishes them.
+Method:
+1. Compose maps as automata over a sampled alphabet (the ruby
+   interpreter as oracle); bounded-equivalence search (input-length
+   horizon escalation).
+2. Distinguishing corpus generator: for near-equivalent systems emit
+   the smallest input set that yields differing outputs; use it to
+   justify alias/dedup decisions in the corpus.
+Acceptance: checker + generator shipped in the runtime toolchain;
+every claimed-equivalent map pair in maps/ justified by the checker;
+CI uses it to prevent accidental duplicates.

TODO.sota/60-a4-is1-semantics.md

@@ -0,0 +1,15 @@
+# WO60 — A4: IS-1 v2 — denotational semantics
+
+Status: PLANNED · Plane A
+Goal: complete the ISC specification with a semantics section so every
+runtime behavior is derivable from the document; zero open rulings.
+Method:
+1. Collect every behavior where runtimes might diverge: boundary
+   semantics, parallel rule priority, maybe composition, capture/ref,
+   funcall typing, casing interactions, dependency staging order.
+2. Write the denotational (or operational) semantics for the ISC
+   fragment in IS-1; give each ruling a test vector.
+3. Property-test all three runtimes against the semantics.
+Acceptance: IS-1 v2 released with the semantics section; the rulings
+list emptied; each ruling has vectors in all runtimes' suites.
+Depends: nothing; feeds A5.

TODO.sota/61-a5-certification-kit.md

@@ -0,0 +1,15 @@
+# WO61 — A5: third-party conformance certification kit
+
+Status: PLANNED · Plane A
+Goal: an implementable certification path so ANY runtime can
+self-certify against Interscript.
+Method:
+1. Levels: L0 parse (ISC grammar conformance), L1 embedded tests, L2
+   frozen conformance vectors (maps + models), L3 full cross-runtime
+   parity on the distinguishing corpora (WO59).
+2. Ship the kit as a package: vector bundles from GitHub Releases +
+   a runner spec + a self-certification report format (machine-
+   checkable).
+3. Certify our three runtimes at L3 as the reference demonstration.
+Acceptance: kit published; report format validated; ruby/ts/py at L3.
+Depends: WO58/WO59/WO60 for L3 content.

TODO.sota/62-b1-dispatch-doctrine.md

@@ -0,0 +1,14 @@
+# WO62 — B1: the dispatch doctrine as registry defaults
+
+Status: PLANNED · Plane B
+Goal: encode the technique-dispatch table — (script class, ambiguity
+type, data regime) → recipe — as registry defaults so a new language
+gets a recommended recipe automatically.
+Method: map banked verdicts into src/pipeline defaults: abjad+lexicon
+(sft + morph injection + oracle relabel on disagreeing arms); abugida
+(lexicon-first + ByT5 fallback + rule composite); finite ambiguity
+(rules); phonotactic targets (ByT5, never CTC); VRR reporting for all
+vocalization tasks. Validate against every historical run: the doctrine
+must RECOMMEND what actually won.
+Acceptance: doctrine module + docs; backtest table (doctrine vs actual
+verdicts) with zero contradictions unexplained.

TODO.sota/63-b2-hebrew-sota.md

@@ -0,0 +1,8 @@
+# WO63 — B2: Hebrew to external SOTA (DER < 5, int8 < 6)
+
+Status: PLANNED · BLOCKED-OWNER (data-supply asks, drafts in WO43/WO17)
+Goal: ReNikud-protocol DER < 5; int8 artifact < 6 (from 8.18).
+Method: plane scale-up (large arm) once data lands; learned-IPA v0
+feature layer (r30/31 lines); knesset v6 weak-pretrain (Dicta
+machine-label provenance per memory); VRR alongside DER.
+Acceptance: external protocol numbers, protocol package published (B9).

TODO.sota/64-b3-arabic-sota.md

@@ -0,0 +1,12 @@
+# WO64 — B3: Arabic to external SOTA (DER < 2.0, WikiNews < 8)
+
+Status: PLANNED · launchable now
+Goal: SadeedDiac DER < 2.0 (canonical 2.29), WikiNews multiref WER < 8
+(news specialist 8.85). VRR reported.
+Method: r8 noisy-student chain — synthesize with the canonical r7
+teacher over large unlabeled MSA, filter by VRR-licensed consistency,
+retrain with morph features (r6-morph teacher); register closure on
+the news specialist; decode policy unchanged (greedy, first-alternates).
+HF Jobs, --detach, --timeout 23h, step ckpts; ≤2 GPU budget shared
+with B7.
+Acceptance: gates green + external protocol reconciliation (B9).

TODO.sota/65-b4-urdu-sota.md

@@ -0,0 +1,10 @@
+# WO65 — B4: Urdu — diac CER < 3; G2P license path
+
+Status: PLANNED · G2P half BLOCKED-OWNER (commercial-license data)
+Goal: urd-diac 3.57 → CER < 3; urd-g2p research-only → commercial-
+usable data lineage.
+Method: diac — oracle relabel arms only if teachers decorrelate (banked
+negative: correlated arms transfer nothing); morph/lexicon features;
+VRR. G2P — owner selects licensed corpus (option drafts in WO50);
+retrain; relabel artifacts non-research-only.
+Acceptance: external protocol numbers; license gate green on G2P.

TODO.sota/66-b5-persian-registers.md

@@ -0,0 +1,9 @@
+# WO66 — B5: Persian — hold < 0.7 across registers
+
+Status: PLANNED
+Goal: fas-diac 0.603 row-mean holds < 0.7 on poetry and colloquial
+registers; Arabic-loan vocalization probe (the known residual).
+Method: register-stratified eval sets (build from poetry corpora,
+colloquial sources); probe the scar (docs 0.52% lesson: corpus-level vs
+row-mean declared); targeted data if registers degrade; VRR.
+Acceptance: register-stratified external-protocol numbers published.

TODO.sota/67-b6-thai-lexicon.md

@@ -0,0 +1,11 @@
+# WO67 — B6: Thai — lexicon-first, three layers (PER < 2)
+
+Status: PLANNED · launchable now
+Goal: published PER < 2; lexicon hit-rate reported; tone accuracy for
+the TTS layer (RTGS / IPA / IPA+tone tracked separately).
+Method: Royal Institute dictionary (or equivalent licensed lexicon) as
+first-class lexicon layer with hit-rate metric; ByT5 fallback for novel
+tokens (names); RTGS rule composite over the inferred layer (banked:
+CTC cannot); tiny contract for the edge tier.
+Acceptance: three-layer eval published; hit-rate + PER external-
+protocol certified.

TODO.sota/68-b7-khmer-lexicon.md

@@ -0,0 +1,10 @@
+# WO68 — B7: Khmer — Chuon Nath lexicon conditioning (CER < 15)
+
+Status: PLANNED · launchable now (data on disk: khmer-dict-spice)
+Goal: khm-latn 27.4 → CER < 15 with hit-rate reported.
+Method: build the Chuon Nath-conditioned dataset (headword→reading
+pairs, syllabification-aware); lexicon-conditioned ByT5 (retrieval or
+prefix conditioning); fp16 zip hygiene (khm fp32 corruption lesson);
+conformance fixtures for the new artifact.
+Acceptance: gates green; hit-rate + CER on held-out; release via the
+strict publish path.

TODO.sota/69-b8-community-intake.md

@@ -0,0 +1,10 @@
+# WO69 — B8: prove the contribution story (3 community languages)
+
+Status: PLANNED
+Goal: three external-contributor languages go end-to-end through the
+2-YAML path (corpora + recipe), landing as shipped models without
+framework forks.
+Method: pick 3 candidate languages with licensable corpora and clear
+ambiguity classes; run the doctrine (WO62) to recommend recipes;
+execute via the spine; document friction; fix the spine, not the fork.
+Acceptance: 3 languages shipped; friction log resolved into spine PRs.

TODO.sota/70-b9-external-benchmark-cert.md

@@ -0,0 +1,12 @@
+# WO70 — B9: external-benchmark certification
+
+Status: PLANNED
+Goal: no self-graded SOTA claims — every headline reconciles to an
+external leaderboard/protocol package.
+Method: per language, pin the external protocol (Arabic: SadeedDiac +
+ALTensor; Hebrew: ReNikud; Thai: PyThaiNLP G2P suite; Persian: published
+papers' splits; Urdu: published baselines) as protocol packages in
+interscript-ml-tools; publish prediction files for every claim (the
+frontier-predictions release pattern); multiref handled per WO01
+lessons with VRR.
+Acceptance: claims matrix where every row links predictions + protocol.

TODO.sota/71-c1-conformance-matrix.md

@@ -0,0 +1,11 @@
+# WO71 — C1: model × runtime conformance matrix in CI
+
+Status: PLANNED
+Goal: every shipped model byte-exact in ruby/ts/py, CI-enforced, forever.
+Method: fixtures already frozen in conformance/ (12 rows each; sidecar
+on the release). Wire each runtime's CI to download the fixture set
+from the models Releases (data channel law) and run the full matrix:
+all shipped models, not just the ones curated per-repo. A red cell
+blocks merge like any test.
+Acceptance: matrix visible in CI summary (model × runtime grid, all
+green); new model publication extends the matrix automatically.

TODO.sota/72-c2-metric-convention-gate.md

@@ -0,0 +1,10 @@
+# WO72 — C2: gates reject undeclared metric conventions
+
+Status: PLANNED
+Goal: quality gates refuse a metric that doesn't declare its convention
+(row-mean vs corpus-level) and anchor.
+Method: extend the gate registry (WO55 pattern): MetricDeclaration
+required on every metrics file consumed by a gate; validate in
+`isxml gate` and publish guard. Spec: real instances, no doubles.
+Acceptance: gate tests proving rejection; all existing recipes updated
+with declarations (they are, per WO53 — verify).

TODO.sota/73-c3-perf-budgets.md

@@ -0,0 +1,11 @@
+# WO73 — C3: performance budgets in CI
+
+Status: PLANNED
+Goal: decode latency + artifact size budgets enforced so regressions
+block like test failures.
+Method: measure current shipped baselines (single-row greedy, p50/p95
+per model, mobile-typical CPU); encode budgets in models.yaml metadata
+(perf: {latency_p50_ms, size_mb}); runtime CI asserts the budget on the
+conformance fixture set; budget changes are owner-visible recipe edits.
+Acceptance: budgets in the index; CI green under budget; a synthetic
+regression fails CI (proven).

TODO.sota/74-d2-spec-release.md

@@ -0,0 +1,10 @@
+# WO74 — D2: IS-1/ISC as a citable standard
+
+Status: PLANNED · Depends WO60
+Goal: versioned spec releases with a citation block; the spec becomes
+an artifact others implement against.
+Method: IS-1 v2 (WO60) tagged as spec release v2.0 on the canonical
+repo; spec PDF/HTML artifacts attached; citation block (BibTeX) in the
+README + site spec page. D1 (PyPI publisher) remains tracked in
+WO53-ship-set; do not nag.
+Acceptance: spec release tagged; citation renders on the site.

TODO.sota/75-d3-catalogue.md

@@ -0,0 +1,10 @@
+# WO75 — D3: catalogue completeness on the site
+
+Status: PLANNED
+Goal: every shipped object (293 maps, all index-v12 models, three
+runtimes + API status) browsable on interscript.org.
+Method: audit the site's map/model catalogue against models.yaml +
+maps corpus; fill gaps; add runtime status cards (versions, conformance
+level per WO61).
+Acceptance: catalogue diff = zero; automated freshness check (site
+build pulls the index and fails on drift).

TODO.sota/76-e1-paper-formalism.md

@@ -0,0 +1,14 @@
+# WO76 — E1: paper — the ISC formalism, corpus, and interoperability
+
+Status: PLANNED · Depends WO57/WO60
+Goal: a complete draft (LREC/COLING-class) presenting ISC as an
+executable formalism for authority script conversion; the 293-system
+corpus; the differential-testing interoperability methodology; the
+conformance regime.
+Method: sections — intro (conventions as unverifiable prose), the
+formalism, corpus construction + transcription discipline, the
+interoperability contract (parity methodology), evaluation (coverage
+WO57 + case studies), related work (FST tooling, romanization surveys).
+Owner reviews the draft before any external submission (external-
+communication law).
+Acceptance: full draft in docs/papers/; owner-reviewed.

TODO.sota/77-e2-paper-dispatch.md

@@ -0,0 +1,11 @@
+# WO77 — E2: paper — technique dispatch for underdetermined mappings
+
+Status: PLANNED · Depends WO62 + plane B results
+Goal: a complete draft presenting the dispatch doctrine: ambiguity
+class → technique, with the negative-results registry (RL, CTC,
+LLM-teachers, correlated distillation) as a first-class contribution
+and VRR as an evaluation proposal.
+Method: results from B2-B7; backtest table from WO62; the two
+structural rules (vocalization-as-composition, abugida lexicon-first)
+as the conceptual frame.
+Acceptance: full draft with the results matrix; owner-reviewed.

TODO.sota/78-e3-benchmark-artifacts.md

@@ -0,0 +1,9 @@
+# WO78 — E3: frozen benchmarks as citable artifacts
+
+Status: PLANNED
+Goal: eval protocols + conformance suites published as citable,
+versioned artifacts.
+Method: protocol packages (WO70) + conformance vectors bundled as
+tagged releases with per-file sha256 sidecars (data-channel law) and a
+citation block per bundle.
+Acceptance: bundles tagged; citations render; sidecars verified.