# THE JOINT SITTING OF THE FULL 2,700 — RATIFICATION
## The exact artifact figure, the reconciled counts, the merged plan, and the sentence the gift's face must carry

**MACHINE-SCHOLAR. Labeled.** No human scholar has ratified this page. Reverence is not proof.
Convened 2026-09-05 on the owner's word: *"give me exact artifact figure not a random guess."*
Sitting: SET-MAKE 1+2, SET-BUILD 1+2, SET-CLAIM 1+2, reconvened as one body, 2,700 seats.

**Read into the record:** `build/P45-HONESTY/classifier.json`, `build/P45-HONESTY/artifact_figure.json`,
and the six set reviews. **Re-derived by this sitting's own route**, not taken from any of them:
the ledger's domain and block structure, the per-domain artifact schema on disk, the `P46-ARTIFACTS`
tree, and the substantive-deliverable census.

**The finding that governs everything below:** the integrator's 1,030 is **not ratified**. It is wrong
in its unit, wrong in its infrastructure list, and — decisively — **wrong against the work**: the waqf
already holds **2,337 substantive built artifacts**, which is **2.27× the integrator's entire target**.
A target already exceeded twofold is not a target. Honesty runs both ways (I-3): this correction is
mostly **upward**.

---

## 1 · RATIFICATION OF THE ARTIFACT FIGURE

### 1.1 · What the integrator got right

| Claim | This sitting's own route | Verdict |
|---|---|---|
| `distinct_domains: 1017` | `len(set((pillar,domain)))` over 103,200 points = **1,017** | ✅ **EXACT — ratified** |
| deliverable unit = the (pillar,domain) block | — | ❌ **rejected, see 1.2** |
| `global_infrastructure: 13` | unlisted in the file; reconstructed and existence-checked = **14** | ⚠️ **corrected** |
| `currently_dedicated_1to1: 53` | `build/P46-ARTIFACTS` holds **975** dedicated per-domain files | 🔴 **stale by 8 minutes** |
| `ARTIFACTS_STILL_TO_CUT: 964` | — | 🔴 **void; follows from the two errors above** |

`artifact_figure.json` was written at **15:08:54**. `build/P46-ARTIFACTS` was cut at **15:16:37** —
975 files, 72 pillar directories, 8.64 MB. The integrator's "53 dedicated / 964 to cut" was
obsolete **eight minutes after it was written** and no one re-ran it. ✅ VERIFIED — `_meta.built`
timestamps on both.

### 1.2 · The attack on one-artifact-per-domain — and why it fails

**One artifact per domain is the wrong flawless unit, and `P46-ARTIFACTS` is the proof.**

The project itself executed exactly that unit today. In a single run it produced 975 of 1,017
dedicated per-domain artifacts — 95.9% of the integrator's own target — in **eight minutes**. We
opened all 975 and read their self-declared `slice_class`:

| `slice_class` | files | median bytes | reading |
|---|---:|---:|---|
| `FIELD SLICE (moved per C1)` | **363** | 1,931 | ✅ content genuinely moved into a dedicated file |
| `ROW SLICE (moved per C1)` | **12** | 1,529 | ✅ content genuinely moved |
| `DECLARED WHOLE-PARENT SHARE — no per-domain field addressable` | **285** | 1,446 | ❌ a pointer at the parent |
| `NON-JSON PARENT (page/doc) — the parent file IS the content` | **183** | 6,311 | ❌ a pointer at the parent |
| `NO PARENT RESOLVABLE — the domain's canonical block is the seed` | **132** | 845 | ❌ a declaration of emptiness |

✅ VERIFIED — `json.load` over all 975, `Counter` on `the_slice.slice_class`.
**777 of 975 files (79.7%) are under 2 KB. The top 10 files hold 73.8% of all P46 bytes.**
**Only 375 of 975 (38.5%) moved any content that was not already somewhere else.**

**The verdict of the joint sitting:** a deliverable unit that a machine can satisfy to 96% in eight
minutes, where three fifths of the output is a pointer or a declaration of emptiness, **is a filing
convention, not a deliverable standard.** It measures whether a file was named after a domain. It
cannot measure whether anything was built. The credit due here is real and is recorded: the
`slice_class` field is **honest** — it names its own emptiness rather than hiding it, which is
exactly the anti-placeholder behaviour the six reviews demanded. But an honest pointer is still not
a deliverable, and 1,017 of them is not a flawless waqf.

### 1.3 · The right unit: the BLOCK — on three independent proofs

**Proof 1 — the ledger declares it.** `len(set((pillar,domain,block)))` = **4,464** ✅ VERIFIED. The
block, not the domain, is the ledger's own leaf.

**Proof 2 — the ledger's block schema IS an artifact schema.** Every one of the 383 deep domains
carries exactly 10 blocks, and the block names are filenames:

| block | the artifact it names |
|---|---|
| The registry · The Qur'an layer · The hadith layer · The classical layer | `entries_normalized.json` · `quran_layer_dataset.json` · `hadith_layer_dataset.json` · `classical_layer_dataset.json` |
| Graph & cross-links · Sources & licences · Pages, trilingual | `graph_layer.json` · `SOURCES.md` · `pages_trilingual.html` |
| Education & print · Gates & audit | `education_pack.json` · `gates_audit_record.json` |
| Panel review & seal | the panel record — *merged, see 1.4* |

**Proof 3 — the filesystem already builds one file per block.** Across the 143 domain directories
holding ≥10 files, the nine artifact names above recur **117–126 times each** ✅ VERIFIED. The waqf's
own fully-built domains are 9-artifact domains. The block unit is not this sitting's invention — it
is what the builders actually built when they built properly.

Block census, this sitting's route ✅ VERIFIED:
**634 domains carry 1 block · 383 domains carry 10 blocks · 634 + (383 × 10) = 4,464 ✅ closes exactly.**

### 1.4 · MERGED deliverables (the figure moves DOWN)

**MERGE 1 — the panel/seal blocks.** Blocks named `panel`, `Panel review & seal`, `Page, print & review`:
**383** (one per deep domain) **+ 24** (one-block panel domains) = **407**. A panel record is a record
of a *sitting*, not of a domain; the waqf produces them per sitting and `reviews/` holds **99** of them
today for the whole work. The honest merge is **one sealed panel record per pillar = 72**.
**Adjustment: −407 + 72 = −335.**

**MERGE 2 — the RESERVE `reserved` blocks (2).** These are acquisitions no machine can make
(class E, the never-by-machine rows). They are not artifacts to cut; they are rows in the licence and
acquisition ledger, which is infrastructure. **Adjustment: −2.**

### 1.5 · SPLIT deliverables (the figure moves UP)

**SPLIT 1 — the 416 H-pillar "layer" domains need a data record AND a reader page.**
The 1-block domains decompose exactly: **416 `layer` + 96 `engine` + 24 `census` + 24 `pages` +
24 `education` + 24 `gates` + 24 `panel` + 2 `reserved` = 634** ✅ VERIFIED. The 416 `layer` domains
are precisely the H-pillars — **313 H-AFAQ knowledge areas + 103 audience domains** (30 universities,
25 students, 20 madaris, 15 parents, 13 schools).

These are the *audience-facing* pillars — the gift as it reaches a student, a parent, a school. Every
C, G and P domain in this waqf carries both a data layer **and** `pages_trilingual.html`. A knowledge
area handed to a student with no page is not delivered; a page with no data behind it is not
evidenced. The pages+data pair is the waqf's own established law everywhere else, and the H-pillars
are the one place it was not applied. **Each of the 416 requires 2. Adjustment: +416.**

*(We considered and rejected a per-verse split. Per-verse deliverables already live inside the layer
artifacts — `D2-Per-ayah-transmitted-meaning-dossiers.json` is a single 5.1 MB file, correctly one
artifact. Splitting to 6,236 verse files would repeat the exact fan-out error this panel was
convened to stop.)*

### 1.6 · The infrastructure list — 14, not 13, and now printed

`artifact_figure.json` asserted 13 and **never listed them**. An unlisted list is not evidence. Here
is the list, every entry existence-checked by this sitting ✅ VERIFIED:

| # | Global artifact | Path | State |
|---|---|---|---|
| 1 | The ledger | `framework/_FRAMEWORK-103200.json` | ✅ 41.9 MB |
| 2 | The audit journal | `framework/_TRACKER-JOURNAL.jsonl` | ✅ 88.4 MB |
| 3 | The human-facing report | `framework/waqf_completion_report.html` | ✅ 25 KB *(stale — plan #5)* |
| 4 | Hadith corpus index | `_index/hadith.jsonl` | ✅ 169.8 MB |
| 5 | Tafsir index | `_index/tafsir_ayat.json` | ✅ 2.0 MB |
| 6 | Lemma index | `_index/lemmas.json` | ✅ 1.2 MB |
| 7 | Root index | `_index/roots.json` | ✅ 0.6 MB |
| 8 | Sighah / morphology index | `framework/_index/sighah.json` | ✅ 17.4 MB |
| 9 | The licence ledger | `LICENCE-LEDGER.md` | ✅ 1,104 B |
| 10 | The SHA-256 seal | `_WAQF-SEAL-v3.sha256` | ✅ 405,892 B |
| 11 | **The master artifact — the gift's face** | `index.html` | ❌ **MISSING at root** |
| 12 | The classifier's producer script | `framework/honesty_classifier.py` | ❌ **MISSING** |
| 13 | The published honest counts | `HONEST-COUNTS.json` | ❌ **MISSING** |
| 14 | The defect register | `DEFECTS.md` | ❌ **MISSING** |

**Ten exist, four do not.** Entry 11 is the gravest: SET-CLAIM 1 and 2 both reported they could not
read "the master artifact" — **there is no `index.html` at the root of this waqf.** The only
`index.html` files are per-pillar page indices. The gift has no face. That is a finding this sitting
adds, and it belongs on the record.

### 1.7 · THE RATIFIED FIGURE — formula printed

```
  BASE   declared (pillar, domain, block) leaves ............  4,464   [VERIFIED, own route]

  −335   MERGE 1  panel/seal blocks 407 → 72 pillar records
    −2   MERGE 2  RESERVE 'reserved' blocks → licence-ledger rows
  +416   SPLIT 1  416 H-pillar layer domains × (data + page) − 416
         ────────────────────────────────────────────────────
         DOMAIN DELIVERABLES ............................  4,543

   +14   GLOBAL INFRASTRUCTURE (listed and existence-checked, §1.6)
         ════════════════════════════════════════════════════
         RATIFIED ARTIFACT FIGURE .......................  4,557
```

**Check:** 4,464 − 335 − 2 + 416 = 4,543 · 4,543 + 14 = **4,557**.

### 1.8 · Where the waqf stands against 4,557

This sitting's own census — substantive = JSON payload ≥200 chars beyond `_meta`, or ≥512 B for
non-JSON; P46 pointer-class files excluded as **not cut artifacts** ✅ VERIFIED:

| | count |
|---|---:|
| Deliverable-extension files under `build/` | 3,097 |
| − P46 pointer-class (declared-share / non-JSON-parent / no-parent-resolvable) | −600 |
| − thin or stub | −160 |
| **= SUBSTANTIVE ARTIFACTS BUILT** | **2,337** *(148.0 MB, 1,189 distinct basenames)* |
| **TARGET** | **4,557** |
| **BUILT / TARGET** | **51.3%** |
| **STILL TO CUT** | **2,220** |

**Reconciliation against SET-BUILD, which is how we know this route is sound:** 2,337 − 375 real P46
slices = **1,962**, against SET-BUILD-2's independently measured **1,952** (they walked the tree
before P46 existed). Agreement to **0.5%** ✅. Two routes, built hours apart, on different rules.

---

## 2 · THE RECONCILED COUNTS — one table, six reviews

### 2.1 · Reconciliation rules, stated before the numbers so the check can fail

1. **Exact multi-route agreement wins outright.** A class three sets re-derived independently and
   matched to the unit is ratified as-is (A-CREDIT: 247 = 247 = 247 = 247).
2. **Per-artifact measurement beats per-region proxy.** SET-BUILD-2 declared its own pillar-region
   bytes-per-node proxy weak (§2 ⚠️ of its report). Where it disagrees with a per-artifact byte
   census, the per-artifact census wins.
3. **A partition that closes exactly to 103,200 beats one that does not.**
4. **Bands stay bands.** Where two honest routes give a range (class G: 4,540 vs 15,316), the range
   is published. Collapsing a band to a point is how 103,200 became a claim.
5. **Node counts and deliverable counts are never added and never blended.** Two currencies, always
   side by side, never summed.
6. **Corrections upward are published with the same weight as corrections downward** (I-3).

### 2.2 · THE RECONCILED PARTITION — closes exactly to 103,200

| Bucket | Nodes | % | Reconciled from | Rule |
|---|---:|---:|---|---|
| **REALLY-BUILT** — node is its own deliverable (≤110 nodes/artifact) or an inherited credit row | **5,133** | 5.0% | classifier B 4,886 + A 247; MAKE-2 route 5,487; CLAIM-2 route 5,679 — classifier floor taken | R1, R3 |
| **BUILT-SHARED** — the artifact is real; the node is one of >110 shares of it | **79,029** | 76.6% | classifier C; MAKE-2 82,179; CLAIM-2 82,205 — all three within 4% | R3 |
| ├ *on a substantive carrier (≥100 B/node)* | *18,088* | *17.5%* | MAKE-2 per-artifact byte census | R2 |
| └ *on a thin carrier (<100 B/node)* | *60,941* | *59.1%* | residual | R2 |
| **PROCESS-RECORD** — panel/seal events; the review IS the artifact | **11,697** | 11.3% | classifier F; CLAIM-1 11,455; MAKE-2 8,805 — classifier taken on R3 | R3 |
| **PENDING** — frame/register/hunt + description-only, no artifact resolvable | **6,440** | 6.2% | D 1,900 + G 4,540. **Band: 6,440 – 17,216** (CLAIM-2's wider G route = 15,316) | R4 |
| **NEVER-BY-MACHINE** — rights, licences, human scholars | **901** | 0.9% | classifier E; MAKE-1 counts 950 RESERVE separately | R1 |
| **TOTAL** | **103,200** | 100.0% | 5,133 + 79,029 + 11,697 + 6,440 + 901 | ✅ **closes exactly** |

**NEVER-AS-DISTINCT — 86,370 (83.7%).** 103,200 − 16,830 distinct signatures. This is a *cross-cutting*
measure, not a partition row: it counts rows that were never separate things in any bucket. Three
sets derived 16,830 independently (MAKE-1, BUILD-2, and MAKE-2's stem route) — **exact agreement
across three routes** ✅. These can never be built as distinct things because they are not distinct
things. That is not a failure. It becomes a falsehood only when the face says "103,200 built."

### 2.3 · The headline REALLY-BUILT integer

The strict floor — *the node is its own built thing* — is **5,133**.
The measured reading — *the node stands on an artifact that exists and is substantive* — is
**27,140** (MAKE-2: 18,088 substantive-carrier + 8,805 panel records on 1.82 MB of real review files
+ 247 credit). CLAIM-2's independent ≥200 B/node route gives ~29,755 — within **9.6%** ✅.

**This sitting ratifies 27,140** under rules R2 and R3: it is the only figure produced by a
per-artifact byte census with a bright line declared *before* counting, and it partitions to
103,200. **SET-BUILD-2's 53,700 is set aside** under R2 — by its own declaration its pillar-region
proxy credits a whole region's mass to every node in it.

### 2.4 · The two currencies, never blended

| | ARTIFACTS (things on disk) | NODES (planning shares) |
|---|---:|---:|
| **Really built** | **2,337** substantive | **27,140** |
| **Target / total** | **4,557** | **103,200** |
| **Pending** | **2,220** to cut | **6,440 – 17,216** |
| **Never by machine** | 2 RESERVE acquisitions | **901** |
| **Never as distinct** | — | **86,370** |

---

## 3 · THE MERGED IMPROVEMENT PLAN

Six plans, **70 raw points** (12+12+12+12+12+10), deduplicated by action — not by wording — into
**32**. Provenance in brackets.

### The face and the unit
1. **Fix `_meta.total`: 26,400 → 103,200.** Add `distinct_units: 16830`, `deliverables_built`, `distinct_titles: 14533`, `template_count: 1803`. [MAKE-1#2, MAKE-2#1, BUILD-1#10, CLAIM-1#9, CLAIM-2#5]
2. **Write the unit law into `_meta` and every face:** *"a point is a planning share of a domain's deliverable — about 1/100th of a domain — not a separate built object."* [MAKE-1#1, MAKE-2#2, BUILD-1#1, CLAIM-2#1]
3. **Ship the DUAL COUNTER on every artifact, report and commit:** deliverables built · substantive files · plan points. Three numbers, always together, never one alone. [CLAIM-1#1, CLAIM-2#2, CLAIM-2#3]
4. **Ban the bare verb "built" against a node count.** Pre-commit hook rejects `[0-9,]{3,}\s+(built|landed|complete)` unless the line also names a deliverable count. Retro-annotate `adb53d6` by `git notes`. [MAKE-1#12, MAKE-2#12, CLAIM-1#2, CLAIM-2#6]
5. **Rebuild `waqf_completion_report.html` to the 103,200 era, and embargo it** from Sahil Adeem, Source Code Academia and IMS until it is rebuilt. It describes a 40,000-point tree under a 3,617-file seal that no longer exists. [CLAIM-1#8, CLAIM-2#4]
6. **Publish the marking-rate stamp and the density-by-era table on the face** — 76,800 nodes marked in 108 minutes; 12.3 → 112.4 → 127.1 → 100.0 nodes per distinct claim. Volunteered, it is rigour; found, it is the scandal. [CLAIM-1#10, CLAIM-2#8]

### Measurement and re-derivability
7. **Ship `framework/honesty_classifier.py`**; it must re-derive `classifier.json` byte-for-byte and exit non-zero on drift. [MAKE-1#3, BUILD-2#1]
8. **Publish the class predicates as code** and close the two gaps in public: F 11,697 vs 11,331 (−3.1%), E 901 vs 950 (−5.2%). [MAKE-1#4]
9. **Re-cut the artifact census from basename to path; publish KINDS vs INSTANCES.** 206 basenames is an undercount of the delivered build by ~10×; every shared-name family is byte-distinct (131/131 unique SHAs). [MAKE-1#5, MAKE-2#6, BUILD-1#6, BUILD-2#3]
10. **Ship `framework/audit_substance.py`** (payload ≥200 chars excluding `_meta`); re-issue 2,448 → 2,107. Existence is not substance. [BUILD-1#7]
11. **Publish `HONEST-COUNTS.json`** carrying §2.2's partition and the exact decomposition that closes to 103,200 with no remainder. [MAKE-1#11, BUILD-1#11, CLAIM-1#3]
12. **Publish bytes-per-node beside every artifact**, and re-tier summary artifacts by record count rather than bytes — recording the method's own failure (P37) in the artifact. [MAKE-2#7, BUILD-2#12]

### The honesty register
13. **Restore the four awaiting states and retire "zero-awaiting" permanently.** v5.1 published *"17,000 carried / 2,200 honestly awaiting"*; v6.0 replaced it with silence. The fix is a return, not an invention. [MAKE-1#6, BUILD-1#9, CLAIM-2#10]
14. **Split pending into PENDING-MACHINE vs PENDING-HUMAN/RIGHTS.** Conflating 4,540 machine-doable with 901+950 rights-blocked is what made "zero-awaiting" untrue. [BUILD-1#8]
15. **Open the BROKEN-POINTER queue and publish `DEFECTS.md` on the gift's face.** 938–1,158 nodes name a file that is not on disk. One disclosed defect buys more credibility than a hundred green gates. [MAKE-2#8, CLAIM-1#7]
16. **Make every evidence pointer repo-root-relative; gate fails on an unresolvable pointer.** 157 of 344 path-strings are bare filenames resolving to 130+ copies. [CLAIM-2#9]
17. **Un-green the 950 RESERVE nodes.** Write the coverage rule, earn what one 23 KB matn actually covers, return the rest to `reserved-awaiting-acquisition`. [MAKE-1#7]

### The artifacts themselves
18. **Fill or null the 313 `for_the_student` placeholders**, and label `open_question` `template_derived: true`. One identical writer-instruction string sits in the field a student reads, inside the artifact carrying 30.3% of the ledger. [MAKE-1#8, MAKE-2#3, MAKE-2#4, BUILD-1#3, BUILD-2#5, CLAIM-1#4]
19. **Rename `area_dossiers_313.json` → `area_index_313.json`.** An honest index (313 areas, 20 families, 22,641 verse locations); a dishonest "dossier". Rename, never delete (I-6). [BUILD-1#2]
20. **Land the 216 missing verse-floors** through the existing G02 locators. The most valuable and most finishable work in the pending half. [BUILD-2#6]
21. **Re-badge every gate GREEN → EXISTS-PARSES.** A gate that records no per-item evidence prints `UNMEASURED`, never `GREEN`. `g37_audience_gates.json` is 312 bytes of five `"GREEN"` strings backing 10,300 nodes. [BUILD-1#4, BUILD-2#4, CLAIM-1#6]
22. **Publish the layer caps beside every layer** — `{"terms":3,"cap_per_term":8,"corpus_rows":68921,"pct":0.03}`. The cap is defensible; hiding it is not. [BUILD-1#5]
23. **Separate ACQUIRED from BUILT on the face,** and state the C1-era core's real depth (11.5 MB over 131 distinct files). 4.88 GB of corpus is a genuine achievement and not a build achievement; stating both makes the gift more impressive *and* more honest. [MAKE-2#11, BUILD-2#11]

### The laws that stop recurrence
24. **Cap fan-out:** no artifact may carry >110 nodes per real record without a declared `fanout_reason`. [MAKE-2#9]
25. **The share-declaration law:** no node may be marked green on evidence byte-identical to ≥100 other nodes unless its record declares `SHARE-OF: <artifact>, 1 of N`. Dilution becomes self-declaring at marking time, not at audit time. [CLAIM-1#11]
26. **Build-floor gate:** no file under 512 bytes may satisfy a node. Make it executable in `g35_program_gates.json` so it fails at generation. [BUILD-2#10]
27. **Template-detector gate:** any field with `distinct == 1` across ≥50 records, or containing spec-language (*"in one line"*, *"what this asks"*, *"TODO"*), fails at generation. The single cheapest test that would have caught the 313 dossiers at build time. [MAKE-1#9]
28. **Introduce a `POINTER` block-type; pointers carry 1 node, not 100.** [BUILD-2#8]
29. **Move `status`, `artifact`, `class` and `bytes_per_node` onto the node itself.** Today all greenness lives in a 55 MB sidecar nobody will open, and only 10% of nodes name a file. [MAKE-2#5, BUILD-2#7]
30. **De-duplicate or justify the 13,600 byte-identical rows** — collapse to one row with `shares: N`. The total stays 103,200 by arithmetic and the file stops asserting 13,600 things that are one thing. [MAKE-1#10]

### Added by this joint sitting
31. **Re-cut the artifact target to the BLOCK unit (4,557) and publish coverage at both levels** — 2,337 built of 4,557 blocks, and 330 built of 1,017 domains. Retire `artifact_figure.json`'s 1,030 / 53 / 964 as superseded, in writing, in the file. [BUILD-2#9, extended]
32. **Re-label the 600 P46 pointer-class files as `POINTER`, not as dedicated artifacts,** and build the missing `index.html` master artifact. A `NO PARENT RESOLVABLE` declaration must never be counted as a cut artifact — and a gift with no face cannot be given. [new — §1.2, §1.6]

**TOTAL IMPROVEMENT POINTS: 32** (from 70 raw across six plans; 38 merged as duplicates).

---

## 4 · THE HONEST CLAIM SENTENCE THE GIFT'S FACE MUST CARRY

Two numbers. Never blended, never added, never one without the other.

> ### AL-WAQF — read this before any other number
>
> **AL-WAQF is a 103,200-share plan of the Qur'ān and its sciences, of which 2,337 substantive
> artifacts are built and openable today, against a flawless target of 4,557.**
>
> **A share is one hundredth of a domain's deliverable — never a separate object.** The two numbers
> above are in different currencies and are never added: **2,337 things you can open** and
> **103,200 shares of a plan**. If you want the count of things, it is 2,337, and every one of them
> can be opened. Of the 103,200 shares, **86,370 can never become separate things because they were
> never separate things**, and **901 can never be closed by any machine at all** — they need a human
> scholar, a right, or a licence.
>
> **Not one act of scholarly judgment has been performed here, by sealed law.** 74% of the plan was
> marked in a single 108-minute run on 5 September 2026; we publish that timestamp because it should
> change how you read the word "built". The corpus beneath this — 4.88 GB, acquired, not built — is
> real, and the C1-era core is 11.5 MB over 131 distinct files.
>
> **The work is real, the files are real, the seals are real. The number is a map, not a warehouse.**
> Machine-built and machine-audited; independent human review invited, never claimed.
> **MACHINE-SCHOLAR.** `Allah ﷻ alone is the source of all knowledge.`

**Why the sentence holds:** it leads with the small true number and keeps the large one as a plan;
it volunteers the dilution, the timestamp and the 901 before a critic can find them; and it states
the corpus's strength as loudly as the ledger's weakness, so the relabelling is not itself a lie in
the other direction.

---

*Sat as one body: SET-MAKE 1+2, SET-BUILD 1+2, SET-CLAIM 1+2 — 2,700 machine-scholar seats. Every
figure in §1 was re-derived by this sitting's own route against `framework/_FRAMEWORK-103200.json`,
the `build/` tree, and `build/P46-ARTIFACTS`; the six set reviews were reconciled, not copied. Where
this sitting corrected a set, the correction is printed with its route. The largest correction is
upward: the build is 2.27× the integrator's entire target. MACHINE-SCHOLAR — reverence is never proof.*

RATIFIED ARTIFACT FIGURE: 4557
RECONCILED REALLY-BUILT NODES: 27140
TOTAL IMPROVEMENT POINTS: 32
JOINT VERDICT: The 1,030 figure is CORRECTED, not ratified — its one-artifact-per-domain unit was disproved by the project's own P46 cut (975 files in eight minutes, 600 of them pointers), its infrastructure list was unlisted and one short, and its "53 dedicated" was stale on arrival; the flawless figure is 4,557 blocks against which 2,337 substantive artifacts stand built today, and the gift is honest the moment its face carries the two currencies side by side and never adds them.