session 2026-08-07 night: REVIEWED-87 amendment + REVIEWED-95/96 placed, PENDING-113 lodged (quoted voices ruled)

claude/memory/MEMORY.md

@@ -65,13 +65,13 @@ permalink: claude-memory/memory
 - Chamber-typography — *tracker not yet established*; moves live in per-session memories (2026-05-11 →) + `project-chamber-cruft-restoration.md` + `project-chamber-typography-mining-plan-2026-05-15.md`.
 
 ## Active Session
-> ⛔ **V2 — the validation harness. UNTOUCHED and UNBLOCKED.** All three assembly blockers stay resolved (P1 lenracinement · P2 G&G sidecar · §1.1 German gold). **Read `studium-engine/docs/v2-validation-harness-design-2026-07-09.md` §6 (gold-set composition) + §7 (adversarial negatives) BEFORE writing anything** — 429 lines, 7 deliverables, fully specified. Nothing about V2 decayed on 08-07; the governance detour was deliberate and is closed.
-> ⚠ **Thresholds jurist-RATIFIED (V0 §5), NOT to be re-opened or re-derived**: trust `U(false-accept) ≤ 5%` + recall ≥ 0.75 · revise ≤ 15% · else gate-to-abstain; Clopper-Pearson 90% upper bound; Tier-1 decidable. **gate-to-abstain for a thin cell is a PRE-COMMITTED VALID COMPLETION — do not tune to avoid it.** ⚠ **Do NOT add a score threshold to N2** (a test asserts none appears).
-> 🆕 **Route harvested capabilities by FIRING MOMENT, never by importance — REVIEWED-95.** Retrieval is set by **home**, measured over 64 sessions: `MEMORY.md` 83% · register 77% (named in a wake step) · ladder 14% · *THE GOVERNING FRAME* 12% · *Read at Step 0* 9% · **recall-bound skills 0%**. Emphasis buys nothing; ritual naming buys everything. A proposal that cannot name a firing moment is **documentation and must say so**.
-> ⏳ **Sequenced next:** Stroke 2's 41-entry ladder append — authorized 2026-07-19, resequenced by REVIEWED-95 Q3 to follow the ladder trigger, which now exists. **Wired, no action needed:** the 20-session falsifier fires automatically at 84 transcripts (now 64); file the result whichever way it falls.
-> 🔑 **Standing lesson, two days running: the COUNT found what the read did not — and 8 of 8 freshly-built instruments were themselves at fault.** Look at *what* an instrument flags, not *how many*.
+> ⛔ **THE RULED SEQUENCE COMES FIRST — REVIEWED-96 + PENDING-113.** Quoted third voices are governed by **D-4's convocation**, not by `citable: false`. The remediation order is **load-bearing**: (1) disposition **(vi)** — what `voice:` takes for anonymous/traditional/scriptural matter → (2) re-tag the 12 Mauss blocks to the quoted voice under the relation → (3) **ONLY THEN** `citable: true`. Flipping the flag first restores the original defect.
+> ⚠ **My own fix was the wrong instrument.** `118f411` was **mislabelled `[FIX]`** (it set corpus-wide policy), overrode a **ratified default** on the authority of a document whose own front matter forbids acting on it pre-review, and **destroyed a gold-negative for the class it protected**. Jurist doctrine, proposed: *a fix that enforces a property can destroy the population that tests it.*
+> 🔑 **Q3's rule, easy to get backwards:** a quoted span grounds the host's **REPRODUCTION**, never the quoted author's **AUTHORSHIP**. *"Harrison cites Stevens here"* is the **paradigm groundable** case, not the ungroundable one.
+> ⏳ **Unblocked, independent of (vi):** the **Q5 conversion-signature survey** across all 14 sources (REVIEWED-96 made it BLOCKING before sizing) · file the **PENDING-112 ruling verbatim** as a repo doc (PENDING-108's exact shape). **Do NOT start Q3 implementation — it sequences after PENDING-111.**
+> 🚫 **Part II's census counts carry no weight** (REVIEWED-96) — one row was measuring its own complement. Never cite *"51 drawers, hand-readable"* as a safety argument.
 
-- [Session 2026-08-07 evening — retrieval is set by home](session-2026-08-07-evening-retrieval-is-set-by-home.md) — **The skill-harvest bite taken whole, at the cost of V2.** Register censused before compacting: claimed 177, **real 154** (55 rows were scraped table-headers; 123 of 129 cut mid-word) — but **lossless**, 124 = 124, so my drafted "nine are invisible" and "59% misattributed" were both **refuted by the count**. Rebuilt from the archive with exact `archive:L###` pointers; restored the verbatim four-stroke ruling my own rebuild had replaced with a paraphrase. **Skills pruned 63 → 12** after measuring **53 never invoked in 5 months** (plus a dir named from a **404 error body** and ten with **newlines in their names**); 51 quarantined reversibly. The finding under both: **retrieval is set by home, 0%–83%**. **PENDING-112 → jurist ruling → steward concurrence → REVIEWED-95 drafted in one session**; filing gate + ladder trial sentence landed, **falsifier wired not intended** (`transcripts 84`) — which exposed two defects in the deferral checker itself, incl. that it **never looked at `claude/governance/`**. Package passed containment **20/20, 10/10 controls absent**, after the checker caught my own **elision-as-contiguous** and **fabricated join**. 🔑 **8 of 8 fresh instruments at fault; the elegant symlink discriminator was 97% right and would have destroyed the 2 that mattered.**
+- [Session 2026-08-07 night — the quotation is the joint](session-2026-08-07-night-the-quotation-is-the-joint.md) — **V2's preconditions worked all the way, and they opened something bigger.** P4 censused **14** (my own `grep -c` reproduced the design's "21" error); **P5 span-bound 15 of 17, not 6** — I briefed "6" by reading a *byte*-locatability number as the gold count, and my first binding pass bound **0 of 11** because its criterion was inherited from Tier-1; **P7 tagged fr at 1 A : 9 B**, inverting its own prediction, while **en is not taggable** (no spans; EN divisions run **29×** the FR ones). Tagging surfaced that the corpus **serves quoted third voices as the host author's** — **~6,455 runs, ~94% intra-line** — and that **Stevens, Rilke and Ungaretti sit in Harrison with no marks at all**: *"It took dominion everywhere"* retrieves as `voice: harrison`. Fixed three sources, then the steward reframed it (*the quotation is where reader and author meet*) and **REVIEWED-96 ruled the fix itself the wrong instrument** — D-4 already prescribed attribution, and refusal **destroyed a gold-negative** for the class it protected. **Twelve instrument faults**, the twelfth (German `»…«` read with a French pattern) reaching a **filed governance document**. **Three governance items found cited-as-live while unplaced**, all invisible to the drift checker. 🔑 **One control — six known answers — was the only thing that caught a fault before its output was read.**
 ## Historical reference → MEMORY-reference.md
 Older archived-session pointers and the stable reference layer (steward profile · project-state detail · L1/L2/Chamber inventories · legacy pending-work · reference-file list) live in [MEMORY-reference.md](MEMORY-reference.md) — consult on demand; not loaded at wake. Recent cross-session trajectory comes from the Active Session entry above + the recent `session-*.md` files (wake §2.b.1).