# PANEL SITTING — P31 AL-MIZAN WORD + MEANING ENGINES (W2 · W5 · W7 · W8)
**Window:** PA2 (Opus) — PANEL WAVE A · 2026-09-03  
**Sources read:**

| File | Size | Built |
|---|---|---|
| `W2-P3/plural_layer.json` | full read | 2026-09-03T18:21:10 |
| `W2-P5/diptote_layer.json` | full read | 2026-09-03T18:39:09 |
| `W2-P6/definiteness_layer.json` | full read | 2026-09-03T18:21:11 |
| `W2-P9/weakness_layer.json` | full read | 2026-09-03T14:44:20 |
| `W4-S6/idafa_layer.json` | full read | 2026-09-03T18:21:11 |
| `W5-T4/emphasis_stack_layer.json` | full read | 2026-09-03T18:21:11 |
| `W5-T5/hadhf_candidates.json` | full read | 2026-09-03T18:45:10 |
| `W7-M1/strata_layer.json` | 14,961 rows, probed via Python | 2026-09-03T15:01:28 |
| `W7-M6/dossiers_v2.json` | 3,842 dossiers, probed via Python | 2026-09-03T15:02:27 |
| `W8-V5/tvf_v1.json` | 6,236 verses, probed via Python | 2026-09-03T18:21:41 |
| `W8-V10/coverage_proof.json` | full read | 2026-09-03T18:12:14 |

**Method:** adversarial; a clean pass must be earned; EVIDENCE LAW: every claim carries file+line or value.  
**Verdict at bottom.**

---

## SEAT 1 — WEAKNESS CLASSES (W2-P9)

*Verify 5 roots' classifications by the i'lal definitions.*

**Procedure:** For each sampled root, I derive the class independently from the root's letter positions (fa'/ʿain/lam) against the standard sarf definitions, then compare to `weakness_layer.json → by_class → exemplars`.

**Definitions used:**

| Class | Definition |
|---|---|
| sahih-salim | No weak letters (waw/ya'/hamza) in any position |
| ajwaf | ʿain position = waw or ya' |
| naqis | lam position = waw or ya' |
| mithal | fa' position = waw or ya' |
| mahmuz-fa/ain/lam | hamza in fa'/ʿain/lam position |
| mudaʿʿaf | ʿain = lam (geminated middle–final) |

### Five roots independently verified

**Root سوي — ajwaf+naqis** (`W2-P9/weakness_layer.json → exemplars → ajwaf+naqis`: forms `ٱسْتَوَىٰٓ / سَوَّىٰ`, ayah 2:29)

Independent derivation: س-و-ي: fa'=س (sound), ʿain=و (weak waw) → ajwaf ✅; lam=ي (weak ya') → naqis ✅.  
Combined label `ajwaf+naqis` is **CORRECT ✅**.

**Root ضوأ — mahmuz-lam+ajwaf** (`exemplars → mahmuz-lam+ajwaf`: forms `أَضَآءَتْ / أَضَآءَ`, ayah 2:17 / 2:20)

Independent derivation: ض-و-أ: fa'=ض (sound), ʿain=و → ajwaf ✅; lam=أ (hamza) → mahmuz-lam ✅.  
Combined label `mahmuz-lam+ajwaf` is **CORRECT ✅**.

**Root وقي — mithal+naqis** (`exemplars → mithal+naqis`: forms `تَتَّقُ / ٱتَّقُ`, ayah 2:21 / 2:24)

Independent derivation: و-ق-ي: fa'=و (weak waw) → mithal ✅; ʿain=ق (sound); lam=ي → naqis ✅.  
Combined label `mithal+naqis` is **CORRECT ✅**.

**Root حيي — mudaʿʿaf+ajwaf+naqis** (`exemplars → muda''af+ajwaf+naqis`: forms `يَسْتَحْىِۦٓ / أَحْيَٰ`, ayah 2:26 / 2:28)

Independent derivation: ح-ي-ي: fa'=ح (sound); ʿain=ي and lam=ي → ʿain = lam (identical letters) → mudaʿʿaf ✅; ʿain=ي → ajwaf (yaʾiy) ✅; lam=ي → naqis ✅.  
All three labels apply simultaneously by their independent definitions. The triple classification is **unusual** — most classical sarf texts treat حيي primarily as mudaʿʿaf — but it is not *wrong*. Each class label is independently earned. **CORRECT ✅** (classification consistent; labelling unambiguous but non-standard in presentation).

**Root أوي — mahmuz-fa+ajwaf+naqis** (`exemplars → mahmuz-fa+ajwaf+naqis`: forms `ـَٔاوَىٰ / ءَاوَ`, ayah 8:26 / 8:72)

Independent derivation: أ-و-ي: fa'=أ (hamza) → mahmuz-fa ✅; ʿain=و → ajwaf ✅; lam=ي → naqis ✅.  
Combined label `mahmuz-fa+ajwaf+naqis` is **CORRECT ✅**.

### Weakness seat verdict

All five roots independently confirmed correct. No misclassification found. **SEAT 1: PASS ✅**

---

## SEAT 2 — SPEAKER-DETECTOR (W7-M1)

*Sample 6 prophetic-stamped strata rows: does each truly attribute the saying to the Prophet ﷺ? Examine 3 downgrades.*

**Data:** `strata_layer.json`: 14,961 rows; `by_era`: prophetic=1,899, companion=8,694, successor=4,368. Downgraded=2,052.

### 6 Prophetic rows — speaker-detector PASS audit

Row keys: `verse, edition, type, attributor, era_stamp, note, window, quote_src, carrier_grading_verbatim`.

| # | Verse | Edition | Attribution formula in quote_src | Grading | Verdict |
|---|---|---|---|---|---|
| 1 | 2:1 | tabari | `قال رسول الله ﷺ: بلى!` — direct speech marker + ﷺ | صحيح | ✅ PASS correct |
| 2 | 2:30 | tabari | `قال النبي ﷺ: فهلا ضربْتَ عُنقه` — direct speech + ﷺ | حسن \| مرسل | ✅ PASS correct¹ |
| 3 | 2:31 | tabari | `قال رسول الله ﷺ: إنّ الله خلق آدم من قَبضة` | حسن \| صحيح | ✅ PASS correct |
| 4 | 2:36 | tabari | `قال رسول الله ﷺ: خُلقتْ هي والإنسانُ...` | **None / null** | ✅ PASS correct — attribution valid; grading gap² |
| 5 | 2:47 | tabari | `سمعت رسول الله ﷺ يقول: ألا إنكم وفيتم سبعين أمة` | ثابت \| حسن \| صحيح | ✅ PASS correct |
| 6 | 2:48 | tabari | `قال رسول الله ﷺ: لا يموتن أحدكم وعليه دين` | حسن \| صحيح | ✅ PASS correct |

**Note 1:** Row 2 (2:30, tabari): The grading reads `حسن | مرسل`. A mursal chain (tābiʿī narrating without a Companion link) is a valid hadith category with weakened isnad; the speaker-detector PASS correctly identifies the attribution formula (`قال النبي ﷺ`). The mursal grading is chain-level, not speaker-level. No conflict.

**Note 2 (grading null scope):** Row 4 has `carrier_grading_verbatim: null`. Deep probe shows 1,558/1,899 prophetic rows (82%) carry null grading. The meta states "gradings carried" — this language implies gradings are present where meaningful, but does not warn that 82% are null. The null rate is expected for tafsir editions (most mufassirūn do not grade each ḥadīth they cite), but the meta's phrasing creates a false expectation. **→ Finding F5 below.**

### 3 Downgrades — were they correct?

**Downgrade 1: verse=2:31, edition=shia_tusi** `[FINDING F1]`  
- Attributor column: "The Prophet Muhammad ﷺ"; era_stamp downgraded to: companion.  
- Note: *"the quote does not attribute a saying to the Prophet ﷺ by the detector's markers"*  
- Quote fragment: `...روي عن النبي صلى الله عليه وآله أنه قال: خلق الله آدم من قبضة قبضها من جميع الأرض...`  
- **Panel verdict: DOWNGRADE IS WRONG.** The phrase `روي عن النبي صلى الله عليه وآله أنه قال` literally means "it is transmitted from the Prophet (peace be upon him and his family) that he said." This IS a prophetic attribution formula — in the Shia invocation style (`وآله`) rather than the Sunni ﷺ or `صلى الله عليه وسلم`. The content (`خلق الله آدم من قبضة...`) is the well-known hadith on the creation of Adam from different soils of the earth, a marfūʿ transmission in Ṭabarī's own text (row 3 above, same verse, same hadith content, correctly PASS-stamped). The detector's stated reason is factually incorrect.

**Downgrade 2: verse=2:57, edition=shia_tusi** `[FINDING F1 — also affected]`  
- Quote fragment: `...وروي عن النبي صلى الله عليه وآله أنه قال: الكمأة من المن وماؤها شفاء للعين...`  
- This is the ḥadīth on truffles ("the truffle is from manna, its water is a cure for the eye"), a Ṣaḥīḥ Bukhārī / Muslim ḥadīth (Book of Food / Medicine) with direct prophetic attribution.  
- **Panel verdict: DOWNGRADE IS WRONG.** The formula is identical to Downgrade 1 — Shia invocation with `وآله`, direct attribution to the Prophet.

**Downgrade 3: verse=2:59, edition=wasit**  
- Quote fragment: `...أخرج البخاري عن أبي هريرة...`  
- The text is a chain citation: al-Wāsiṭ commentary citing Bukhārī on Abū Hurayra. The truncated 200-char sample does not show the Prophet's direct speech following. The pattern "أخرج البخاري عن أبي هريرة" attributes the narration to a Companion (Abū Hurayra), not directly to the Prophet — though a marfūʿ continuation may follow.  
- **Panel verdict: DOWNGRADE PLAUSIBLE.** The visible evidence does not show a direct prophetic speech marker in the fragment. Given the truncation, this is provisional — not a confirmed PASS, not a confirmed false negative.

### Speaker-detector systematic gap

Deep probe on all shia_tusi rows: 451 total; 45 downgraded; only 2 retained as prophetic-stamped.  
8 downgraded rows contain the `وآله` attribution pattern in their quote_src — all 8 show direct `روي عن النبي صلى الله عليه وآله أنه قال` or `روي عن النبي (ص) أنه قال` formulas.

**The speaker-detector's marker list does not include:**
- `صلى الله عليه وآله` (Shia invocation form)  
- `(ص)` as a prophetic abbreviation  
- `روي عن النبي ... أنه قال` with any non-standard invocation  

Of the 45 downgraded shia_tusi rows, at least 8 are confirmed false negatives. The true scope could be higher since the pattern probe only checked for `وآله` in the quote body. The 2,052 total downgraded rows remain correct for Sunni editions, but the shia_tusi edition has a systemic false-negative problem.

**→ Finding F1 [FIX-FIRST]: Systematic Shia-attribution false negatives in speaker-detector.**

---

## SEAT 3 — DOSSIERS V2 (W7-M6)

*Ordering + P30-depth-preservation claim on 3 verses.*

**Data:** `dossiers_v2.json`: 3,842 dossiers; combined coverage with P30 = 3,883; p30_deep_dossiers=90.

### Ordering verification — 2:255 and 2:187

**Verse 2:255 (Āyat al-Kursī):**  
- `strata_present: ['prophetic', 'companion']`  
- `transmitted_meaning_oldest_first` row order: prophetic → prophetic → companion  
- Prophetic before companion: ✅ oldest-first ordering is maintained

**Verse 2:187:**  
- `strata_present: ['prophetic', 'companion']`  
- Row order (first 3 of 10): prophetic → prophetic → prophetic  
- Prophetic before companion: ✅

### P30-depth-preservation claim

The meta states: *"P30's 90 richer dossiers remain canonical for their verses (dossiers_v1) — this layer extends coverage, never overwrites depth."*

Probe: Dossier rows in v2 carry only `['era_stamp', 'type', 'attributor', 'edition', 'note']` — `quote_src` is absent from dossier entries (present only in the strata_layer). The v2 dossier is a **reference/pointer structure**, not a self-contained document. The depth lives in the strata_layer and the P30 canonical dossiers (v1), not in v2's row bodies.

This is **by design** and **not an error** — the meta's claim of "never overwrites depth" is correct: v2 adds coverage (3,842 v2 + 41 p30-only = 3,883) while P30's 90 deeper dossiers remain canonical. ✅

### Fatiha absence — execution warning

**Finding:** All 7 Fatiha verses (1:1–1:7) are absent from dossiers_v2.  
- `{'1:1': False, '1:2': False, '1:3': False, '1:4': False, '1:5': False, '1:6': False, '1:7': False}`  
- The dict starts at 2:1.  
- Combined coverage: 3,883/6,236 (62%) — the Fatiha is in neither the 3,842 v2 entries nor the 41 p30-only-referenced.

If P30's deep dossiers (v1 format) cover Fatiha, those are canonical — but any code querying `dossiers_v2.json` for Fatiha returns nothing. The TVF for 1:1 shows "consensus-shelf required elements exist" so some coverage exists via the shelf, but the dossier layer itself is a gap.

**→ Finding F6 [EXECUTION WARNING]: Surah al-Fatiha absent from dossiers_v2; any lookup against v2 for verses 1:1–1:7 returns no result.**

---

## SEAT 4 — TVF (W8-V5)

*Open 5 verses' checklists: is any check WRONG or overclaimed for that verse?*

### TVF structure

`tvf_v1.json → tvf` dict: 6,236 keys (one per canonical verse). Each entry: `scheme_B_key`, `checks[]`, `layers_absent[]`.  
Note: All `scheme_B_key` values are offset by +1 from the canonical reference (scheme B counts the Basmalah as verse 1:1 and shifts all subsequent keys). This is declared in the meta (`"every verse carries its scheme-B (6,348) key"`). Not an error.

### Verse 1:1 — Bismillah

`checks`: 3 MEANING checks; `layers_absent`: none.  
- MEANING: consensus-shelf required elements ✅ (shelf covers Fatiha)  
- MEANING: Prophetic-era explanation transmitted ✅  
- MEANING: era-stamped transmitted meanings ✅  
- **FORM checks: 0.**

**Finding:** The W5-T5 hadhf_candidates.json (built 18:45) explicitly marks the Bismillah's implied verb as `confirmed_seed`:

> `"hadhf": "the implied verb of beginning ('I begin...')", "authority": "the consensus shelf: 18 editions at 1:1"`

This confirmed hadhf should generate the form check: *"FORM: implied verb (hadhf) — the rendering must preserve the implied syntactic action."* The TVF for 1:1 carries 0 form checks. The hadhf is not missing from the layer — it is confirmed. It is missing from the TVF because **the TVF was built before the hadhf layer** (TVF: 18:21:41; hadhf: 18:45:10; delta: +23 min). The TVF's own law says *"checks grow as the wings land (the file regenerates from the layers, never by hand)"* — the regeneration has not happened. **→ Finding F2 [FIX-FIRST].**

### Verse 2:255 — Āyat al-Kursī

`checks`: 1 FORM (dual — twoness must survive) + 3 MEANING.  
- The dual is sourced from `حِفْظُهُمَا` (the -humā suffix). ✅ dual check is CORRECT.  
- Meaning checks: prophetic transmitted, companion transmitted, consensus-shelf. All ✅.  
- `layers_absent`: none.

Is the FORM check *wrong*? No — the dual is present in the verse. Is it *overclaimed*? No — the check correctly says "twoness must survive" without asserting more.

Is the FORM coverage **sparse** for Āyat al-Kursī, one of the most grammatically dense verses? Yes — `وَلَا يَئُودُهُ` (descriptive negation with لَا + mudāriʿ) is absent as a check; the emphatic rhetorical question `مَن ذَا الَّذِي يَشْفَعُ` is absent. These are sparsity gaps, not errors — the layer's checks grow as more detectors land. ✅ No wrong or overclaimed check found here.

### Verse 2:187 — "They are a garment for you"

`checks`: passive (veiled doer), jussive denial (lam-type), imperative present, Prophetic meaning. `layers_absent`: shelf.

- Passive check: `يُحِلَّ` — Wait, 2:187 contains passives? Actually the verse has `أُحِلَّ` (it is made lawful) — passive ✅.  
- Jussive denial (lam-type): `لَا تُبَٰشِرُوهُنَّ` — this is a present-tense prohibition with لَا, not a lam-type jussive denial. **The lam-type check may be triggered by a different clause.** Let me assess: "jussive denial class present (lam-type) — completed denial, not present tense." The verse contains `وَلَا تُبَٰشِرُوهُنَّ` — the detector flagged a lam-type. But لَا + amr (prohibitive) is different from لَمْ/لَمَّا (jussive negation for past). This could be a false positive in the check label if the detector cannot distinguish prohibitive لَا from lam-type jussive denial. The check itself does not claim a specific locus, so it cannot be adjudicated wrong from the TVF alone — **PLAUSIBLE but needs detector-code inspection.**  
- Imperative check: `وَكُلُوا وَاشْرَبُوا` (eat and drink) ✅ imperative present.  
- Prophetic meaning: dossier shows prophetic rows for 2:187 ✅.  
- The tashbīh `هُنَّ لِبَاسٌ لَّكُمْ` ("they are a garment for you") has no figurative-speech check — consistent with the PANEL-P31-PLAN F1 finding. No overclaim: the check set does not promise figurative-speech coverage.

**→ No check is WRONG for 2:187; the lam-type check merits detector-code verification (soft note, not a finding).**

### Verse 4:11 — Inheritance fractions

`checks`: dual (twoness), jussive denial (lam-type), Prophetic meaning, transmitted meanings. `layers_absent`: shelf.  
- Dual: `وَلَهُ أُخْتَانِ` (and he has two sisters) ✅.  
- Jussive denial: `وَلَمْ يَكُن لَّهُۥ وَلَدٌ` (and he has no child) → لَمْ + jussive = lam-type ✅.  
- Prophetic: the inheritance hadith is a primary prophetic transmission ✅.  
- **No wrong or overclaimed check.** ✅

### Verse 24:45 — Allah created every creature from water

`checks`: 1 FORM (dual). `layers_absent`: shelf, dossier.  
- Dual: `عَلَىٰ رِجْلَيْنِ` (on two legs) ✅ — dual correctly detected.  
- `layers_absent: ['shelf', 'dossier']` — the verse has no consensus shelf coverage and no transmitted dossier; declared honestly.  
- The coverage_proof counts this as "covered" (≥1 check). The FORM check is not wrong and not overclaimed.  
- **No check is wrong or overclaimed.** The sparse coverage is a gap, not an error; it is declared. ✅

### TVF timestamp issue — critical

| File | Built timestamp |
|---|---|
| `W8-V10/coverage_proof.json` | 2026-09-03T**18:12:14** |
| `W8-V5/tvf_v1.json` | 2026-09-03T**18:21:41** |
| `W5-T4/emphasis_stack_layer.json` | 2026-09-03T**18:21:11** |
| `W5-T5/hadhf_candidates.json` | 2026-09-03T**18:45:10** |

The hadhf layer (18:45) postdates the TVF (18:21) by 23 minutes. The coverage_proof (18:12) predates the TVF by 9 minutes.

Consequences:
1. **Hadhf checks absent from the TVF.** Confirmed hadhf at 1:1 — 0 form checks in TVF. The 3,492 verses with hadhf candidates are not represented in the TVF's form-check count.  
2. **Emphasis-stack checks absent from the TVF.** The deepest emphasis stack is 4:119 (13 particles). TVF for 4:119: `['FORM: jussive denial class present (lam-type)', 'MEANING: prophetic...', 'MEANING: transmitted...']` — 0 emphasis-stack checks.  
3. **Coverage_proof stale.** Built 9 minutes before the TVF; must be regenerated after TVF regeneration.

The TVF law: *"checks grow as the wings land (the file regenerates from the layers, never by hand)"* — this regeneration is overdue for two completed wings (W5-T4 emphasis, W5-T5 hadhf).

**→ Finding F2 [FIX-FIRST]: TVF and coverage_proof must be regenerated; W5-T4 and W5-T5 checks are not yet incorporated.**

---

## SEAT 5 — PLURAL / DEFINITENESS / DIPTOTE APPROXIMATIONS (W2-P3 · W2-P5 · W2-P6)

*Is any "declared limit" actually an undeclared error?*

### Diptote layer (W2-P5) — false positive exemplar

**Rule** (`_meta.law`): *"a truly indefinite GEN noun showing FATHA and no tanwin is the diptote's signature (al-mamnu' min al-sarf)"*

**Exemplar 1: `بِمُؤْمِنِينَ` (2:8)** — listed as the FIRST diptote exemplar.

Analysis: `بِمُؤْمِنِينَ` = preposition بِ + مُؤْمِنِينَ.  
- `مُؤْمِنِينَ` is a **sound masculine plural** in the genitive/accusative form; its ending is **kasrah** (-ina suffix: -ِينَ).  
- The diptote rule requires **fathah** in the genitive indefinite position (not kasrah).  
- This form shows kasrah, not fathah. The rule would NOT flag it.  
- **This is a false positive in the exemplar.** The detector should not have flagged this form by its own declared rule.

True diptotes in the same list: `لِءَادَمَ` (2:34), `إِسْرَٰٓءِيلَ` (2:40/47), `فِرْعَوْنَ` (2:49) — these are proper nouns in genitive position with fathah (no tanwin), correct diptote signals ✅.

The presence of `بِمُؤْمِنِينَ` as exemplar #1 suggests the 436 `diptote_signals` count includes sound masculine plural forms that happen to occupy a genitive context. The exact false-positive count in 436 is not determinable from the JSON alone, but the exemplar evidence is unambiguous.

The meta declares: *"construct-heads excluded by the det/suffix filters as candidates only — treebank confirmation AWAITED."* The declared limitation names construct-heads, not sound-masculine-plural forms. The false-positive class (SMP in genitive) is **undeclared**.

**→ Finding F3 [SHOULD FIX]: Diptote exemplar #1 (`بِمُؤْمِنِينَ` 2:8) violates the layer's own rule; the 436 signal count may include SMP false positives; the undeclared source of inflation must be documented or the detector corrected.**

### Definiteness layer (W2-P6) — mudaf before relative pronoun

**Rule**: four of the seven definite classes machine-detectable: proper noun, al-, pronoun-suffixed, construct-head CANDIDATE.

**Exemplar analysis:**

Indefinite residual exemplars (first 3):
- `صِرَٰطَ` (1:7) — classified **indefinite (nakirah)**
- `ٱلَّذِينَ` (1:7) — classified **indefinite (nakirah)**
- `ذَٰ` (2:2) — classified **indefinite (nakirah)**

**`صِرَٰطَ` (1:7):** In the verse `صِرَٰطَ ٱلَّذِينَ أَنْعَمْتَ عَلَيْهِمْ`, صِرَٰطَ is a **mudaf (construct head)** governed by the relative clause ٱلَّذِينَ. It is definite by idāfa to a definite mudaf-ilayh. The correct class is "construct head (mudaf) — candidate."

The construct-head detector correctly captures `سْمِ` (1:1), `رَبِّ` (1:2), `مَٰلِكِ` (1:4) — cases where the mudaf-ilayh is an explicit proper noun (ٱللَّهِ, ٱلْعَٰلَمِينَ, يَوْمِ). It misses صِرَٰطَ because its mudaf-ilayh is a **relative pronoun** (ٱلَّذِينَ) rather than an explicit noun. The detector does not handle the [noun + relative pronoun] construct pattern.

**`ٱلَّذِينَ` (1:7) as "indefinite":** ٱلَّذِينَ is a **relative pronoun**, not a noun at all. Classifying it as indefinite nakirah is a category error — relative pronouns are neither definite nor indefinite in the noun-morphology sense; they are inherently reference-anchored. The indefinite residual bucket receives this as a false member.

**Scope of the gap:** 17,280 "indefinite (nakirah) - residual" entries. Unknown fraction are:
- True indefinite nouns ✅
- Mudaf before relative pronoun (misclassified, should be mudaf-candidate)
- Relative pronouns themselves (category error — not nouns)

The meta declares: *"the demonstrative/relative/vocative classes and construct confirmation AWAIT the i'rab treebank (owner-gated A1) — declared, not approximated."* This declaration covers relative pronouns as a *class*, but the specific failure here is that the construct-head detector partially works (for noun-before-noun) and silently fails (for noun-before-relative-pronoun), depositing those cases into the indefinite bucket rather than into a "mudaf-candidate" or "undetermined" bucket. The gap is partially declared (relative class awaits treebank) but the deposit behavior is undeclared.

**→ Finding F4 [SHOULD FIX]: Construct heads before relative pronouns (صِرَٰطَ ٱلَّذِينَ pattern) fall into "indefinite residual" rather than "mudaf-candidate"; relative pronouns themselves (ٱلَّذِينَ) also appear in the indefinite bucket. The meta declaration covers the class gap but not the misrouting of detected segments.**

### Plural layer (W2-P3) — relative pronouns counted as sound-masculine

**Rule** (`_meta.law`): *"soundness read from the form's own ending (una/ina; -at); everything else broken; SEGMENT-level counts"*

**Exemplar `ٱلَّذِينَ` (1:7) listed as "sound-masculine" plural.**

ٱلَّذِينَ ends in -ina (ٱلَّذِينَ/-ُونَ pattern). By the layer's declared rule, any -ina/-una segment is counted sound-masculine. But ٱلَّذِينَ is a **relative pronoun**, not a morphological plural of a nominal root. The -ina here is the relative pronoun's own inflection, not the sound masculine plural suffix attached to a verbal/nominal stem.

The meta's declared approximation: *"dual-suffixed and construct-clipped forms follow their ending."* Relative pronouns with -ina endings are NOT named as a declared approximation. They are an undeclared source of inflation in the "sound-masculine: 3,291" count.

Magnitude: Relative pronouns with -ina endings in the Quran include ٱلَّذِينَ (extremely frequent — one of the most common forms in the Quran). The inflation of the SMP count from this alone is probably in the hundreds or low thousands of segments.

**→ Finding F5 [SHOULD FIX]: The plural layer's declared approximation does not mention relative pronouns with -ina endings (ٱلَّذِينَ etc.) being counted as sound-masculine plural. This is an undeclared source of inflation in the SMP count of 3,291.**

---

## MECHANICAL CROSS-CHECKS

### Emphasis stack — 4:119 depth-13 vs. TVF

The deepest emphasis stack is 4:119 with 13 particles (`emphasis_stack_layer.json → totals → deepest`).  
TVF for 4:119: `['FORM: jussive denial (lam-type)', 'MEANING: prophetic', 'MEANING: transmitted']` — 0 emphasis-stack FORM checks.  
This confirms that W5-T4 (built 18:21:11) has not been wired into the TVF generator, consistent with F2.

### Idafa chain lengths — sanity check

`W4-S6/idafa_layer.json` → longest chain: 13 (4:36, `وَبِذِى ٱلْقُرْبَىٰ...`). The verse is the famous list of rights-holders — a genuine long idafa/coordination chain ✅. The "10" chains at 4:163 and 9:70 are proper-noun lists (Prophet lineages, prior nations) ✅. No anomaly.

### Hadhf confirmed-seed vs TVF (summary)

Confirmed seed: `verse: "1:1", hadhf: "the implied verb of beginning", authority: "the consensus shelf: 18 editions at 1:1"` — from `W5-T5/hadhf_candidates.json`.  
TVF 1:1 form checks: **0**.  
Delta: hadhf layer built 23 minutes after TVF. Regeneration is the fix, not a re-derivation. ✅ confirmed structural gap.

---

## FINDINGS — SUMMARY TABLE

| # | Finding | Wing | Severity | Action required |
|---|---|---|---|---|
| **F1** | Speaker-detector fails to recognize Shia attribution formulas (`صلى الله عليه وآله`, `(ص)`). At least 8 shia_tusi rows confirmed false negatives out of 45 downgraded (Downgrades 1 & 2 at 2:31 and 2:57 are explicit cases). The stated reason "quote does not attribute a saying to the Prophet ﷺ by the detector's markers" is factually incorrect for these rows. | W7-M1 | **FIX-FIRST** | Add Shia invocation variants to the detector's marker list: `صلى الله عليه وآله`, `(ص)`, `روي عن النبي ... وآله`. Re-run detector on all 451 shia_tusi rows. Re-classify false negatives. |
| **F2** | TVF (built 18:21) predates the hadhf layer (built 18:45) and has not incorporated W5-T4 emphasis checks or W5-T5 hadhf checks. Confirmed hadhf at 1:1 gets 0 form checks in the TVF. 4:119 (deepest emphasis stack: 13 particles) gets 0 emphasis form checks. Coverage_proof (built 18:12) is also stale — predates the TVF. TVF law: "the file regenerates from the layers, never by hand." | W8-V5 / W8-V10 | **FIX-FIRST** | Regenerate TVF and coverage_proof after all built wings have been read: W5-T4 → add emphasis-stack form checks; W5-T5 → add hadhf form checks (using confirmed_seed for 1:1 minimum). Re-run coverage_proof on the new TVF. |
| **F3** | Diptote exemplar #1 (`بِمُؤْمِنِينَ` 2:8) is a sound masculine plural showing kasrah (-ina ending), not a diptote. The detector's own rule ("FATHA and no tanwin in genitive indefinite") would not flag it. The 436 signal count may include SMP false positives. The undeclared inflation class is "SMP in genitive context." | W2-P5 | **SHOULD FIX** | Correct the exemplar list; audit the 436 signals to estimate SMP false positive rate; add SMP to the meta's declared approximation list, or correct the detector to exclude SMP (-una/-ina suffix class) from the diptote-signal scan. |
| **F4** | Construct heads before relative pronouns (صِرَٰطَ ٱلَّذِينَ, 1:7) fall into "indefinite (nakirah) — residual" instead of "construct head (mudaf) — candidate." Relative pronouns (ٱلَّذِينَ, 1:7) also appear in the indefinite bucket. The meta's declaration covers relative pronouns as a class awaiting the treebank, but does not specify that these segments are being deposited into the indefinite residual bucket rather than a neutral "undetermined" class. | W2-P6 | **SHOULD FIX** | Reroute [noun + relative-pronoun] sequences to "mudaf — candidate" or a new "undetermined (pending treebank)" class. Reroute relative pronouns themselves out of the definite/indefinite classification. Amend the meta to name the misrouting. |
| **F5** | Plural layer counts relative pronouns with -ina endings (ٱلَّذِينَ) as sound-masculine plural. The declared approximation in the meta names dual-suffixed and construct-clipped forms but not relative-pronoun false members. The SMP count of 3,291 is inflated by an undeclared amount. | W2-P3 | **SHOULD FIX** | Add "relative pronouns with -ina/-una endings (ٱلَّذِينَ, ٱلَّذُونَ) counted in SMP total — undeclared approximation; estimate magnitude in v2 note." The v2 note already flags a word-level census as needed; relative pronouns should be explicitly named there. |
| **F6** | Surah al-Fatiha (1:1–1:7) is entirely absent from dossiers_v2. All 7 verses return empty. Combined v1+v2 coverage is 3,883/6,236 (62%). Downstream code querying dossiers_v2 for Fatiha returns nothing. | W7-M6 | **EXECUTION WARNING** | Document explicitly in the dossier meta that Fatiha is in P30 canonical dossiers (v1 format) not in v2. Add a `fatiha_in_p30_only: true` flag or equivalent. Any code that merges v1+v2 dossiers must union them before querying. |

---

## CLEAN PASSES

| Area | What passed |
|---|---|
| Weakness classes (5 roots) | All 5 independently re-derived: سوي, ضوأ, وقي, حيي, أوي — all classifications correct by i'lal definitions ✅ |
| Prophetic PASS marks (6 rows) | All 6 sampled rows contain direct attribution formula (`قال رسول الله ﷺ`, `قال النبي ﷺ`, `سمعت رسول الله ﷺ`) — detector PASS marks earned ✅ |
| Downgrade 3 (2:59 wasit) | Plausible on visible evidence — companion-level chain ✅ |
| Dossier ordering | Prophetic before companion for 2:255 and 2:187 — oldest-first maintained ✅ |
| P30 depth non-overwrite | Meta claim verified — v2 rows are reference pointers; P30 canonical depth untouched ✅ |
| TVF checks for 2:255, 2:187, 4:11, 24:45 | No check is factually WRONG or overclaimed; the checks that exist are correct and bounded ✅ |
| Dual form check at 24:45 | `رِجْلَيْنِ` (two legs) is a genuine dual — the check is correct for this verse ✅ |
| Coverage_proof 6,236/6,236 | Every verse has ≥1 TVF entry — technically correct; appropriately labeled "MEASURED, not claimed" ✅ |
| Diptote proper-noun examples | `لِءَادَمَ`, `إِسْرَٰٓءِيلَ`, `فِرْعَوْنَ` — correct diptote signals (proper nouns with fathah in indefinite genitive) ✅ |
| Idafa chain lengths | Longest chain (13, 4:36) is a genuine coordination list; 10-chain verses are proper-noun enumerations ✅ |
| TVF outcome-claim law | Checks are CHECKABLE/DETECTABLE, not CORRECT/CONTROLLED — no overclaims in the body ✅ |
| All six structural proofs (re-verified from W8-V10) | 6,236 verses with TVF entries; letter-coverage claim is measured ✅ |

---

## VERDICT

### **FIX-FIRST**

**Numbered findings:**

1. **F1 [SPEAKER-DETECTOR — FIX-FIRST]** The speaker-detector has a systematic false-negative gap for Shia attribution formulas. At least 8 shia_tusi rows in the 45-row downgraded set carry `روي عن النبي صلى الله عليه وآله أنه قال` — a direct prophetic attribution in Shia form — yet are downgraded with the stated reason "the quote does not attribute a saying to the Prophet ﷺ by the detector's markers." That reason is factually incorrect. Downgrades 1 (2:31) and 2 (2:57) are explicit false negatives; the content is documented ṣaḥīḥ/ḥasan prophetic material (creation-of-Adam hadith; truffle hadith). Must be fixed before the strata layer is considered verified.

2. **F2 [TVF STALENESS — FIX-FIRST]** The TVF (18:21) and coverage_proof (18:12) predate the W5-T5 hadhf layer (18:45). The confirmed hadhf seed at 1:1 — unanimously verified by 18 editions — is absent from the TVF as a form check. The deepest emphasis stack (4:119, 13 particles) gets 0 emphasis-stack form checks. The TVF law requires regeneration after every wing landing; two wings (W5-T4, W5-T5) have landed since the last TVF build. The coverage_proof's specific-layers count does not include hadhf or emphasis checks. Regenerate both files.

3. **F3 [DIPTOTE — SHOULD FIX]** `بِمُؤْمِنِينَ` (2:8) is listed as the first diptote exemplar but is a sound masculine plural with kasrah — not a diptote by the layer's own fathah rule. The 436 signal count may include SMP false positives. Correct exemplar; audit count; declare or fix the SMP gap.

4. **F4 [DEFINITENESS — SHOULD FIX]** Construct heads before relative pronouns (صِرَٰطَ ٱلَّذِينَ, 1:7) fall into "indefinite residual" rather than "mudaf — candidate." Relative pronouns themselves (ٱلَّذِينَ) also appear in the indefinite bucket. The 17,280 "indefinite (nakirah)" count is inflated by misrouted segments. Reroute these to an appropriate class or declared "undetermined" bucket.

5. **F5 [PLURAL — SHOULD FIX]** Relative pronouns with -ina endings (ٱلَّذِينَ) are counted as sound-masculine plural via the form-ending rule, inflating the SMP count of 3,291 by an undeclared amount. The declared approximation in the meta does not name this class. Add it to the v2 note.

6. **F6 [DOSSIER FATIHA — EXECUTION WARNING]** Surah al-Fatiha (1:1–1:7) is absent from dossiers_v2. Any downstream system querying v2 for Fatiha returns empty. Must be documented in meta; union with P30 v1 dossiers must be enforced before any Fatiha query.

**Clean passes:** Weakness classes (5/5 ✅) · Prophetic PASS marks (6/6 ✅) · Dossier ordering ✅ · P30 depth preservation ✅ · TVF checks non-overclaimed (5 verses ✅) · Diptote proper-noun exemplars ✅ · Structural proofs ✅.

**The Word+Meaning engines are structurally sound and their declared approximations are largely honest. Two systematic errors must be fixed (F1 speaker-detector Shia gap; F2 TVF/coverage-proof staleness) before W7-M1 and W8-V5/V10 can be considered verified layers.**

---

*PA2 (Opus) · PANEL WAVE A · 2026-09-03 · `10-WAQF/reviews/PANEL-P31-W2plus.md`*