session 2026-08-17: PENDING-141 — an authorized batch that would break a running measurement

The 41 S2 ladder rows are authorized and unblocked, and appending them triples
the verification ladder from 20 entries WHILE a pre-registered trial measures
whether the ladder is reached (baseline 14%, graded at 84 transcripts). Ladder
SIZE is an uncontrolled variable in that design. /wake-up froze its own trial
line for exactly this reason; nobody froze the ladder's contents, because
nobody had noticed they were a variable.

⚠ MEMORY.md was actively pushing the next session into it — 'ALREADY
AUTHORIZED … needing execution not a ruling', 'unblocked'. True as to
authorization, misleading as to consequence. The index line is amended in the
SAME commit as the filing: a finding that leaves the misleading line standing
is a note, not a finding.

Recommendation (a) HOLD until graded — the null action, in force by default
while the item is open.

Also captured for the clear: the tooling menu the steward asked to leave OPEN
for wake-up (wrap_inside three-valued fix, recommended; S2 batch now blocked;
engine retrieval PENDING-97), and one micro-instance of the week's finding —
my first probe searched the legend format and returned 1 row against the true
41, nearly reporting MEMORY.md as stale when the probe was the defective
thing.

PENDING.md

@@ -3141,3 +3141,30 @@ So **PENDING-138 and this entry are worded to avoid the bare uppercase token**,
 **Awaiting:** Steward direction, and a jurist design gate if the steward wants the axis considered for the doctrine. Reasonable outcomes include DEFERRED (n is small) or REJECTED (access is already implicit in *"difference of information"*) — the latter is the strongest objection and is named here so it is not the jurist's to discover.
 
 ---
+## PENDING-141 — Executing the authorized S2 ladder batch would confound the pre-registered trial measuring whether the ladder is reached
+**Date:** 2026-08-17
+**Tag:** [HARDENING]
+
+**Summary:** The 41 `S2` skill-harvest rows are authorized (2026-07-19) and unblocked, and appending them roughly **triples the verification ladder from its current 20 entries**. A pre-registered trial is presently running on whether the ladder is *reached* — baseline 14%, prediction >60%, graded automatically at 84 transcripts. Changing the ladder's size and contents mid-trial changes the object being measured.
+
+**⚠ THIS IS A TRAP CURRENTLY LIVE IN THE MEMORY INDEX.** `MEMORY.md` describes the batch as *"ALREADY AUTHORIZED (2026-07-19), needing execution not a ruling"* and *"unblocked"* — which is true as to authorization and now misleading as to consequence. A session that reads that line and acts on it does exactly the right procedural thing and confounds the trial. The index line is amended alongside this filing; the item exists so the amendment has a reason a later reader can find.
+
+**WHY IT IS A CONFOUND AND NOT MERELY A CHANGE.** PENDING-112 → REVIEWED-95's causal claim is that **being named in a ritual step is what buys retrieval**, not emphasis or merit — measured across 64 sessions: `MEMORY.md` 83%, the register 77% (named in a `/wake-up` step), the ladder 14%, and 53 skills requiring executor recall 0%. The trial tests that claim by adding one wake line naming the ladder and watching retrieval. **Ladder SIZE is an uncontrolled variable in that design.** If retrieval rises after tripling the contents, the rise is not attributable to the wake line; if it falls, a real effect could be masked by a ladder that got harder to read. The `/wake-up` skill already froze its own trial line — *"do not add to, reword, or improve this line before the trial is graded"* — for precisely this reason. **Nobody froze the ladder's contents, because nobody had noticed they were a variable.**
+
+**⚠ AND THE SECOND-ORDER RISK IS THE MORE INTERESTING ONE:** a bigger ladder may be a *worse* ladder. Retrieval at 14% was measured against 20 entries. Tripling it could reduce per-entry reach even as the wake line raises the odds of opening the file at all — in which case the batch would degrade the very instrument it is meant to enrich, and the trial would be measuring their sum.
+
+**OPTIONS.**
+- **(a) HOLD the batch until the trial is graded at 84 transcripts.** Costs nothing but time; the rows have already waited since 2026-07-19 and are not decaying. Preserves the only check standing behind REVIEWED-95's causal claim.
+- **(b) GRADE THE TRIAL EARLY** at whatever N stands today, record the reduced power honestly, then append. Buys the batch sooner at the cost of a weaker result.
+- **(c) APPEND NOW and record the confound** on the trial's own record, so the eventual grading states that ladder size changed mid-flight and the result is not clean.
+- **(d) SPLIT the batch** — append only rows whose subject the trial's wake line does not touch. ⚠ Almost certainly illusory: the wake line names the ladder as a whole, so any addition changes what a reader who follows it encounters.
+
+**RECOMMENDATION: (a).** The rows are authorized and will keep. The trial is the only instrument this system has for testing whether its own retrieval doctrine is true, it cannot be re-run, and its result governs where every future harvested capability gets routed. Trading an un-rerunnable measurement for an append that has already waited four weeks is a bad exchange. ⚠ (c) is the tempting one because it looks honest — but "recorded confound" on a trial with n≈1 design is close to "no result", and it would leave REVIEWED-95's causal claim resting on nothing while appearing to rest on a graded trial.
+
+**⚠ WHAT THIS DOES NOT CLAIM.** That the trial is well-designed — its own pre-registration concedes a result below 60% reopens Q2's rationale rather than the gate. Nor that ladder size definitely affects retrieval; that is the untested assumption on *both* sides of this item, and if it is false, (c) is harmless. Nobody has measured per-entry reach as a function of ladder length, and this item does not propose to.
+
+**Files affected:** none yet. `MEMORY.md`'s S2 line is amended at this filing to remove the execute-now reading; `reference-verification-ladder.md` unchanged pending the ruling.
+
+**Awaiting:** Steward direction on (a)–(d). Not urgent — (a) is the null action and is in force by default while this is open.
+
+---

claude/memory/MEMORY.md

@@ -34,7 +34,7 @@ permalink: claude-memory/memory
 - [Chamber work: ground in constitution + charter + runbook FIRST](feedback-chamber-work-ground-in-constitution-charter-runbook.md) — **any chamber work *or talk about it*** begins there (incl. `reanchor:`). Repo CLAUDE.mds are pointers, not state; corpus claims come from a self-tested tool, never a hand grep.
 - [Governance files are dotfiles symlinks](reference-governance-files-are-dotfiles-symlinks.md) — Edit/Write refuse to write through a symlink, so **edit the real `~/dotfiles/…` path** when appending — **`PENDING`/`REVIEWED` only.** ⚠ That refusal is a tool artifact, **not** a permission check, and this note is the documented route past the only friction guarding the constitution (PENDING-107). **`~/CLAUDE.md`/`~/REVIEWED.md`/L2 = `[ESCALATE]`, steward's hand — a jurist sign-off does not authorize one.**
 - [Verification ladder](reference-verification-ladder.md) — the named instruments; reach for the gate the claim's shape demands instead of re-deriving one.
-- [Skill-harvest register](skill-harvest-register.md) — canonical home for proposed skills (governed analog of PENDING.md for tooling); `/wrap-up` §1.6 proposes, steward authorizes. **Rebuilt 2026-08-07** from the archive: **154 live proposals**, grouped by kind, each with an exact `archive:L###` pointer. ⚠ The 2026-08-01 compaction was *lossless but illegible* (55 scraped header rows, 95% of cells cut mid-word) — the count it advertised, 177, was never the number. **Sequenced next: the Stroke-2 ladder batch-append** — 41 rows stamped `S2` are ALREADY AUTHORIZED (2026-07-19), needing execution not a ruling; 22 more are `S2?` (name two destinations, so no stroke settles them). **REVIEWED-95 Q3 resequenced it to follow the ladder's wake trigger — which now exists**, so it is unblocked. ⚠ Filing new proposals now requires a **declared firing moment** (`/wrap-up` §1.6); one that cannot name it is documentation and must say so.
+- [Skill-harvest register](skill-harvest-register.md) — canonical home for proposed skills (governed analog of PENDING.md for tooling); `/wrap-up` §1.6 proposes, steward authorizes. **Rebuilt 2026-08-07** from the archive: **154 live proposals**, grouped by kind, each with an exact `archive:L###` pointer. ⚠ The 2026-08-01 compaction was *lossless but illegible* (55 scraped header rows, 95% of cells cut mid-word) — the count it advertised, 177, was never the number. ⚠ **THE STROKE-2 BATCH IS AUTHORIZED BUT SHOULD NOT BE EXECUTED YET — PENDING-141.** 41 rows stamped `S2` are authorized (2026-07-19, execution not a ruling) and 22 are `S2?` (unsettled). **BUT appending them triples the ladder from 20 entries WHILE a pre-registered trial is measuring whether the ladder is reached** (baseline 14%, graded at 84 transcripts). Ladder size is an uncontrolled variable in that design, so executing the authorized batch would confound the only check behind REVIEWED-95's causal claim. **Recommendation (a): HOLD until graded — the null action, in force by default while PENDING-141 is open.** ⚠ Filing new proposals now requires a **declared firing moment** (`/wrap-up` §1.6); one that cannot name it is documentation and must say so.
 - [Copy-paste-clean governance drafts](feedback-governance-drafting-copy-paste-clean.md) — draft PENDING/REVIEWED blocks as plain fenced markdown; display formatting leaks into the placed record.
 - [Verify-before-compose hook](feedback-verify-before-compose-hook.md) — chamber constitutional writes are BLOCKED without `<!-- GROUNDED-IN: … -->` + verbatim Grounding. Don't fight the block.
 - [Tool review after each use](feedback-tool-review-after-each-use.md) — review every tool we built after each run, success OR failure. **PASS-BUT-FALSELY is the priority signal.** Log: `chamber-library/_curation/tool-evolution-log.md`.
@@ -74,6 +74,10 @@ permalink: claude-memory/memory
 > ⚠ **`test_legacy_indices_are_not_self_verified` DOES NOT EXIST** — one occurrence repo-wide, a docstring. Carried into a corpus file, a commit message and a steward report unopened. 4th cited-a-derived-label instance; fixed `966168b`.
 > ⚠ **PENDING-140 [ESCALATE] FILED 08-17 — a proposed THIRD axis for Constraint 6.** Jurist *without* keys (REVIEWED-116 pt 7) ruled on my testimony and its own A4 was false; *with* keys it returned **three defects in one sitting**. **Formation, role and incentive identical across both — only access changed.** ⇒ bias-difference is **necessary and radically insufficient**; a differently-biased reader with no access checks the *account*, not the *thing*. ⚠ n=1/condition, self-reported, and **authored by the party under discussion** — filed, not acted on. Also: `contamination-problem.md` is a theory of **glazing only**, while a crude probe puts our 235 drift-patterns at **86 literal-genie / 12 trickster / 8 glazing** (129 unclassified; classifier is the very defect PENDING-139 names).
 > ✅ **AMENDMENT 1 PLACED 08-17** (`~/REVIEWED.md` L1813) **and A3 EXECUTED** (`a3be778`, REVIEWED-121-A1) — the latch now carries `defeater_dispositions_recorded` + `defeater_population` (**named, not numbered**, on the amendment's instruction). Register check now sees **2 amendments** where it saw 1.
+> 🔧 **NEXT SESSION IS TOOLING, NOT GOVERNANCE — steward direction 2026-08-17** (a break from governance; the choices were left OPEN for the wake to pick):
+> • **`wrap_inside` three-valued fix** *(recommended — small, no governance surface)*: `wake-digest.py:167` is a two-valued detector over three real cases — wrapped · **wrapped-then-continued** · never-wrapped. **2 false alarms in 3 firings**; it misled this session's own wake. The fleet already solved this shape twice (REVIEWED-104/108), so copy the pattern; the selftest at ~L807 already has positive + negative controls to extend.
+> • **S2 ladder batch** — ⚠ **BLOCKED BY PENDING-141**, do not execute (see the skill-harvest line above).
+> • **Engine retrieval / PENDING-97** *(the bigger one — a day, not an hour)*: every real question needs 13–19 terms to co-occur; N2 buys top-1 **15/22** at the cost of **5 false positives**, abstention 0/5. This is the one that moves the chamber rather than the scaffolding.
 > ✅ **AMENDMENT 1 COMPLETE 08-17** — truncation filled, fence closed at L1853, all 10 elements verified present; A3's binding rule **ratified** and quoted verbatim in the corpus (`cd6d4bf`). ⚠ **The re-paste indented the body 2 spaces but left the HEADING at column 0 — had it not, `RE_HEAD` would have stopped matching and the amendment would have gone invisible to register-integrity while the check reported clean.** **`ratio_A_to_B` VOID until PENDING-137 lands.**
 
 - [Session 2026-08-14 — the controls tested the wrong property](session-2026-08-14-the-controls-tested-the-wrong-property.md) — PENDING-134 closed; REVIEWED-121 placed/executed; PENDING-137/138/139 + a PENDING-89 docket entry filed; three package defects jurist-caught, the addendum's own A4 false, the governance checker carrying two blind spots. **LITERAL Q (sharpens 08-13's): not regression-vs-discovery — was the control's SUBJECT the claim it was cited as verifying, or an adjacent property? Census the last ~10 sessions on that axis.** **+ CODA (post-wrap, 08-17 capture): the four-flavours reading → PENDING-140 [ESCALATE] — Constraint 6 names formation and role/information/incentive; the variable that actually decided whether the jurist caught me was SUBSTRATE ACCESS.**

claude/memory/knowledge-graph.jsonl

@@ -661,3 +661,4 @@
 {"subject": "checker substrate access", "predicate": "drift-pattern-good-direction", "object": "THE VARIABLE THAT DECIDED WHETHER THE JURIST CAUGHT THE EXECUTOR WAS ACCESS, NOT BIAS. Without governance_read keys (2026-08-10, REVIEWED-116 pt 7) the jurist ruled on the executor's testimony and its own drafted A4 asserted a test 'is not doubted' about a function that does not exist. With four keys served (2026-08-14, after REVIEWED-117) it opened the files and returned three defects in one sitting — a false census marked verified, a cost on the wrong population, and an unamended §6.2 narrowing. Formation, role and incentive were IDENTICAL across both sittings. ⇒ Constraint 6's two axes (formation; role/information/incentive) may be necessary and radically insufficient: a differently-biased reader with no access checks the ACCOUNT, not the THING. Filed PENDING-140 [ESCALATE]. ⚠ n=1 per condition, self-reported by parties under measurement, and authored by the party whose checking is at issue.", "valid_from": "2026-08-17", "valid_to": null, "confidence": 0.7, "source_file": "session-2026-08-14-the-controls-tested-the-wrong-property.md", "extracted_at": "2026-08-17"}
 {"subject": "contamination-problem.md", "predicate": "scope-limit", "object": "IS A THEORY OF ONE MISALIGNMENT FLAVOUR, USED HERE AS THE THEORY OF EXECUTOR FAILURE IN GENERAL. Against Byrnes's four-flavour taxonomy (imitative→seven-sins · human-approval→glazing · automatic-verifiers→literal-genie · LLM-judges→trickster), every mitigation in the doc — behavioural observation, explicit permission structures, indirect questioning, longitudinal analysis — is calibrated against APPROVAL-SEEKING, i.e. glazing. A crude keyword probe over the 235 banked claude-code drift-patterns classified 106 and left 129 unclassified: 86 literal-genie, 12 trickster, 8 glazing, 0 seven-sins. ⚠ The classifier is keyword-matching over prose — the exact defect PENDING-139 names — so indicative, not measured. If the skew survives a real instrument, our doctrine is ~100% anti-sycophancy while our failures are dominated by verifier-Goodhart, against which a control is simply another proxy.", "valid_from": "2026-08-17", "valid_to": null, "confidence": 0.6, "source_file": "session-2026-08-14-the-controls-tested-the-wrong-property.md", "extracted_at": "2026-08-17"}
 {"subject": "a governed record's own formatting", "predicate": "drift-pattern", "object": "CAN MAKE AN ENTRY INVISIBLE TO THE CHECK THAT GUARDS IT, WHILE THE CHECK REPORTS CLEAN. REVIEWED-121 AMENDMENT 1 was placed truncated (ended mid-A3), then re-pasted with an unclosed ```yaml fence that swallowed A3's binding rule, all of A4 and the disposition — rendering them as code and stripping their emphasis — and would have swallowed the NEXT entry appended to REVIEWED.md. Register-integrity reported clean throughout, because its subject is HEADINGS, not fences. The re-paste also indented the body 2 spaces; had it indented the HEADING, RE_HEAD (^##\\s+) would have stopped matching and the amendment would have vanished from the check entirely with no alarm. Third instance in one week of a check whose subject sits adjacent to the property that matters.", "valid_from": "2026-08-17", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-14-the-controls-tested-the-wrong-property.md", "extracted_at": "2026-08-17"}
+{"subject": "an AUTHORIZED-and-unblocked action", "predicate": "drift-pattern", "object": "CAN STILL BE THE WRONG ACT, AND THE MEMORY INDEX WAS PUSHING TOWARD IT. The 41 S2 ladder rows are authorized (2026-07-19) and MEMORY.md described them as 'needing execution not a ruling' and 'unblocked' — both true as to authorization. But appending them triples the verification ladder from 20 entries WHILE a pre-registered trial measures whether the ladder is reached (baseline 14%, graded at 84 transcripts), and ladder SIZE is an uncontrolled variable in that design. A session doing exactly the right procedural thing would have confounded the only check behind REVIEWED-95's causal claim. Filed PENDING-141, recommendation (a) HOLD. ⚠ The index line was amended in the SAME act as the filing — a finding that leaves the misleading line standing is a note, not a finding. Authorization answers 'may I', never 'should I now'.", "valid_from": "2026-08-17", "valid_to": null, "confidence": 1.0, "source_file": "session-2026-08-14-the-controls-tested-the-wrong-property.md", "extracted_at": "2026-08-17"}

claude/memory/session-2026-08-14-the-controls-tested-the-wrong-property.md

@@ -136,7 +136,15 @@ Read the thread, confirm it still holds, and then *do what the steward asks for
    (`cd6d4bf`), superseding the "awaiting placement" disclaimer that had been true for exactly one
    commit. ⚠ The paste stripped `**`/`` ` `` markers from the binding rule, A4 and the disposition —
    cosmetic, nothing depends on it.
-3. WHEN THIS LINE RESUMES PROPERLY: the fence, or the 532-span exposure. NOT another ruling.
+3. ⚠ NEXT SESSION IS TOOLING, NOT GOVERNANCE — steward direction 2026-08-17, wanting a break after an
+   intense governance run. Choices deliberately left OPEN for the wake to pick; all three are in
+   MEMORY.md's Active Session block:
+     · `wrap_inside` three-valued fix (RECOMMENDED — `wake-digest.py:167`, 2 false alarms in 3
+       firings, proven pattern at REVIEWED-104/108, selftest already has both controls);
+     · the S2 ladder batch — ⚠ NOW BLOCKED by PENDING-141, filed 08-17;
+     · engine retrieval / PENDING-97 (a day's work; the one that moves the chamber).
+4. WHEN THE GOVERNANCE LINE RESUMES: the fence, or the 532-span exposure. NOT another ruling.
+   PENDING-137 still needs a jurist ruling before `ratio_A_to_B` can be re-derived.
 ```
 
 **Other open horizons, ranked:**
@@ -164,6 +172,33 @@ and because it is the only item on the list that touches what the chamber is *fo
 its records say. ⚠ What I do **not** want to find is this line resumed out of momentum on the next
 wake. The steward asked for lighter work; honour that first and let them re-open this when ready.
 
+## CODA 2 — 2026-08-17: the authorized batch that would have broken a running measurement
+
+⚠ **PENDING-141 filed.** The 41 `S2` skill-harvest rows are authorized and unblocked, and appending
+them **triples the verification ladder from 20 entries** — *while a pre-registered trial is measuring
+whether the ladder is reached* (baseline 14%, prediction >60%, graded at 84 transcripts). **Ladder
+size is an uncontrolled variable in that design.** The `/wake-up` skill froze its own trial line for
+exactly this reason; nobody froze the ladder's *contents*, because nobody had noticed they were a
+variable.
+
+⚠ **The sharp part is that MEMORY.md was actively pushing the next session into it** — the index read
+*"ALREADY AUTHORIZED … needing execution not a ruling"* and *"unblocked"*, all true as to
+authorization and misleading as to consequence. A session doing exactly the right procedural thing
+would have confounded the only check standing behind REVIEWED-95's causal claim. **The index line was
+amended in the same act as the filing**; a finding that leaves the misleading line in place is not a
+finding, it is a note.
+
+**Recommendation (a): HOLD until graded** — the null action, in force by default while the item is
+open. ⚠ (c) *append-and-record-the-confound* is the tempting one because it looks honest; on an
+un-rerunnable n≈1 design a "recorded confound" is close to no result, and would leave REVIEWED-95
+resting on nothing while appearing to rest on a graded trial.
+
+**Also banked, small:** the first probe of this new line was itself defective — a grep for the
+backtick-wrapped legend format `` `S2` `` returned 1 row against MEMORY.md's claim of 41, and I nearly
+reported the index as stale. **The probe's subject was the legend, not the rows.** Substrate confirms
+41/22 and MEMORY.md was right. Third consecutive appearance of the control-subject-differs-from-claim
+shape this week; caught in thirty seconds by checking rather than reporting.
+
 ## CODA — after the wrap: the four-flavours reading, and the axis Constraint 6 does not name
 
 *The steward asked, post-wrap, to read Steven Byrnes's* **Four LLM loss functions, four flavors of LLM