Unit 6 is built end to end and QA'd: 10 sequences · 94 lessons · 506 checks · 231 figures, plus 5 gates, 3 practice pages and the unit test. Every item was cold-solved by a non-author (590/591 keys matched — the one miss is S06's figure-caused read (a marked dot hid the segment); the figure was re-instanced and the items re-solved clean in its fix round (§E-S06)) and blind-read by a key-sequestered fresh reader (597/599 — the two 53/54 records' single clashes are both self-ruled reader errors on the record (S01 item 10; S05 item 26, your card), and S10's one FRQ miss is self-ruled fair).
Do two things. First, rule on the 70 cards that genuinely need you — one tap each, and every tap sends straight to the inbox.
Second, skim the verdict cards (one per sequence, gate, practice page and the test) and tap Looks right or Needs work on each. The adjudication calls sit between, each with a one-tap veto.
This page IS the reviewer view, on by default: QA numbers, records and build calls are all visible. Students never see any of it. The course page shows it their way.
Each card is a genuine fork: your answer changes what a student meets, or what the next unit inherits. The filled button is my recommendation; tap it to confirm, tap the other to overrule, or use Other.
P2 is the released 2023 FRQ Q3, stems verbatim. Its system is CaCO₃ + HCl, which U5 taught — normally barred from checks.
A real released question cannot be re-instanced, and the deliberate-retrieval framing arguably wants the familiar system. The 2021-Q3 precedent is claimed (qa/ASSESSMENT_PREP.md §6.2).
P2 holds until you rule. Everything else on this page ships regardless.
My recommendation: The exemption holds — ship P2.
AP items routinely present the heating curve, and EK 6.5.A.1's constant-temperature clause IS the plateau.
The G3 capstone's multi-segment calculation cannot be performed without reading which segment a path is in.
This is the one adoption that ADDS assessed content beyond the EK list's letter (05_operator_adoption.md, D1).
My recommendation: Keep all three atoms.
The CED treats c as an empirical input everywhere it appears, and no AP item asks why water's value is high.
An ungated supporting-content inset after THC-030 is one tap — switching on need-to-know extras is your call, not the operator's (05_operator_adoption.md, D4).
My recommendation: Leave it out.
One-step items key exactly. Multi-step chains are computed FORWARD from real values: readings to 0.1 °C, keys at 3 s.f., stems state the required precision, wrong answers recomputed from named error paths, table entries spaced well apart.
Classroom-style measured ΔT values within ~2% of literature are licensed, and those stems say "a student measures" (ADJUDICATIONS B3(U6)).
My recommendation: The policy stands.
Passed both seats on take 3 (the one hook of S01's four that survived the independent seat).
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S01.md).
My recommendation: Keep the photo.
Three takes failed (depth-of-field; take 3 also grew two stray background candlesticks). The monochrome candle scene shipped.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S01.md).
My recommendation: Keep the figlib scene.
Content and anatomy passed every take; the independent seat failed all three on sharpness alone. The author seat read takes 2–3 as acceptably sharp.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S01.md).
My recommendation: Keep the figlib scene.
Same pattern as the hand warmer: content clean, three sharpness fails.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S01.md).
My recommendation: Keep the figlib scene.
Three sharpness fails; take 3 (top-down kettle grill) is sharp to the corners by the author's read. The figlib charcoal_grill scene shipped.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S02.md).
My recommendation: Keep the figlib scene.
Take 3 is content-perfect — one spoon, one bowl, steam, no text — and failed only on slightly soft wall tiles. The deterministic heatcontact panel carries the scene.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S03.md).
My recommendation: Keep the figlib scene.
Passed all thirteen checks on take 3.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S03.md).
My recommendation: Keep the photo.
Passed both seats on take 1. No retake, no amendment.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S04.md).
My recommendation: Keep the photo.
Passed on take 3 (take 2 the author rejected itself: a horseshoe drawn as two arcs).
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S05.md).
My recommendation: Keep the photo.
Passed on take 3 once the thermometer lost its printed scale — no reading of any kind is visible.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S05.md).
My recommendation: Keep the photo.
Three sharpness fails; the registered cube-on-palm scene (S01's taught depiction, answer layer off) carries the hook. The record offers a fourth take with a fresh composition on your word.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S06.md).
My recommendation: Keep the figlib scene.
Passed on take 3 after the counter clutter went.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S06.md).
My recommendation: Keep the photo.
Passed on take 3 under the focus-stacked amendment.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S07.md).
My recommendation: Keep the photo.
Passed on take 3; the top-down framing also made the finger count checkable.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S07.md).
My recommendation: Keep the photo.
Passed on take 3; the amendment also tightened the device read — one flame fan from the electrode tips, the real torch's geometry.
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S08.md).
My recommendation: Keep the photo.
Three takes AND a licensed crop all failed on sharpness — the author read every take as PASS. The deterministic s09hook drawing shipped.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S09.md).
My recommendation: Keep the figlib scene.
Same full wipe as S09-H1. The record offers: keep the drawing, adopt any named take, or order a fresh round.
The best takes stay on disk (*.reject3.png); adopting one is one word. The independent seat's sharp-everywhere rule did the rejecting — the leak and content checks passed (qa/PHOTO_QC_U6-S09.md).
My recommendation: Keep the figlib scene.
Passed on take 1 — the prompt composed the depth away (the S09 lesson applied).
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S10.md).
My recommendation: Keep the photo.
Passed on take 1; the declared S02-thread retrieval (R4).
It carries a felt scene and no data; everything exact stays figlib. Judged by both seats; leak check passed (qa/PHOTO_QC_U6-S10.md).
My recommendation: Keep the photo.
One change between neighbouring cases is exactly the example discipline, and choosing the system is THC-001's own taught move.
The recorded alternative: two separate chemical cases (ADJUDICATIONS §S01-8).
My recommendation: Keep the dual-system case.
The register's own title cells carry "heat flows" phrasing that the 7 Oct ban bars from student pages.
Students now see THC-021 as Why the hotter body loses energy and THC-024 as Predict the transfer and the final temperature; the register cells are unchanged (ADJUDICATIONS B7(U6), §S03-2).
My recommendation: The retitles stand.
It restates THC-007's third outcome and coins nothing; it is the example run's same-object boundary case.
Capped at one per quiz (ADJUDICATIONS §S02-1).
My recommendation: Keep it.
One convention keeps the routine drillable; the shortcut reaches the same answer, so nothing an AP item needs is lost. The hydrogenation case counts its unchanged C–H bonds on both sides by design (ADJUDICATIONS §S08-3).
Companion note (S08 spec record): L05 keeps the shortcut list as a named distractor — if you add the honest-shortcut line, that distractor's feedback gains its “reaches the same answer by a different convention” clause.
My recommendation: All-bonds only.
It is the mechanism's own minimal pair: averaged entries drift, single-molecule entries do not. Numbers verified (ADJUDICATIONS §S08-2).
The blind read found its check's stem leading (fine as a scaffolded lesson check, barred from any test — on the record).
My recommendation: Keep it.
The two-level boundary stays clean; the worked lane and the growing table carry L05 (ADJUDICATIONS §S09-2).
One tap adds it later as a table-family figure.
My recommendation: Stay declined.
The column was the AND of the three test columns, and deriving it is the lesson's own routine; the four verdicts moved to one prose line.
Executed because the full table could not fit a phone width; it now measures 382 px against the 408 px wrapper (ADJUDICATIONS §E-S09b G2-residual).
My recommendation: The drop stands.
The course writes ΔH_f; AP tables print the standard-state mark. Unit 9 brings the ΔG°/ΔH° family, and whatever you pick there should match what Unit 6 trains (ADJUDICATIONS §E-S09 part 1; register rev 2.1).
My recommendation: Keep ΔH_f as taught; rule the ° question at the U9 build.
The Hess stacks force step and target into one form, so the fix ran wider than the ruling's letter and was ratified; S07's three reaction-context C(s) correctly stayed (ADJUDICATIONS §E-S10/§E-G4).
The one-word C(s) alternative is on this card; it was declined at the case table because the diamond-vs-graphite teaching point needs the named form.
My recommendation: C(graphite) stands.
The teaching surface shows the summed equation and total as the model; every CHECK uses given-mode stacks that never print sums (BUILD_STATE, S10 spec).
Sum-silent mode is one tap if you want students producing the sum even at the teaching line.
My recommendation: Keep the printed sum on the teaching surface.
The same two reactions recur on purpose: manipulation drills in L08, full Hess calculations in L09 (BUILD_STATE, S10 spec).
The scaffold-convergence card below shares the design family.
My recommendation: Bless the arc.
Every Hess stack derives its annotation from the stored manipulation record — “reversed — sign flipped”, “× 2”, “× ½” — never authored per figure (figlib_u6_hessstack; BUILD_STATE, S10 spec).
They recur in gates and tests, so one blessing fixes the course-wide form.
My recommendation: Bless all three.
S10's L01-04 and L09-49 both state sibling values in their stems. The engine shows one check per screen, and the H1 licence makes the carried value load-bearing — the design was ruled twice (ADJUDICATIONS §E-S10b, flag B1).
Three independent blind eyes have now converged on the same note; the records put the convergence itself to you.
My recommendation: The design stands.
The register defines the plateau rule over pure substances; the item keys on that clause with no colligative theory taught (ADJUDICATIONS §S06-6).
The wording stays agnostic on freezing-point depression — it never claims salt water has no plateau behaviour of its own.
My recommendation: Keep it.
The path starts at lead's freezing point, so it never needs liquid lead's heat capacity; a conditional liquid value (0.15) is registered should a later item want the third stretch (qa/REVIEW_ACCURACY_S06; register rev 1.8).
My recommendation: It stands.
Accuracy confirmed the value at-bp; register rev 1.8 pins the at-transition convention unit-wide, which also pre-empts the 40.7-vs-44.0 water collision between S09 and S10 (qa/REVIEW_ACCURACY_S06).
My recommendation: It stays.
The thermometer barely moves, so the honest verdict is neither; the lesson now opens the sorting with “When the thermometer moves, one glance sorts the dissolution…” (ADJUDICATIONS §E-S05 E1).
My recommendation: The key stands.
Items 34 and 44 are each the only item in their run with an extra option — and the extra option is the key both times (qa/REVIEW_QUESTIONS_S05; the closing blind read sustained the second).
One decision covers both. The fix touches the R113 two-option licence and the THC-007 three-outcome licence, which is why it is yours (ADJUDICATIONS §E-S05 E5, §E-S05b).
My recommendation: Make the third option standing across each run.
Register rev 1.7's scope line marks the course's salts as anhydrous (with forward cover for S07); the caveat never reaches student text (qa/REVIEW_ACCURACY_S05).
My recommendation: The handling stands.
The blueprint's 2,090 J and a boiling bath were impossible together; the lesson now runs a 70.0 °C block giving +2,926 J exact both ways (argued in qa/SPEC_RECORD_U6-S05.md; accuracy verified it).
My recommendation: The rebuild stands.
The S03→S05 iron arc must reproduce S03's drawn mid-course panel, and with real c values that drawn ratio implies about 7 kg of iron per litre of water (ADJUDICATIONS §E-S03 ledger actions).
S05 executed it at 7.00 kg / 1.00 kg water, consistent to 0.05 °C.
My recommendation: Accept it — the drawing governs.
3/2 O₂ → O₃ IS a formation equation of an element's non-standard form; the value (+142.7) is verified (ADJUDICATIONS §S09-5).
The blind reader's literal-reading hesitation is on the record; the lesson shows the allotrope case before the item.
My recommendation: It stays.
Three independent cold readers all answered “the system”; the page's sorting sentences, both QA records and AP's mixture-as-surroundings convention all back the key (qa/REVIEW_LESSON_S05 §5).
It stays as the designed boundary near-miss; the never-reuse-outside-the-course caution is on the record.
My recommendation: The key stands.
S01 teaches a third verdict for a system whose energy does not change; THC-062 as registered classifies by sign only and never claims the two classes exhaust all reactions (qa/REVIEW_ACCURACY_S07, register finding 2).
No current item forces the question — this is a register scope line for the record, not a fix.
My recommendation: Add the one scope line at the next register touch.
(1) Item 1's coining scaffold “The reaction happens once in mole units” — fades to exam register by L07/L08. (2) L06's title A beaker that warms itself — the reflexive IS the lesson's point; alternative on record. (3) The “belongs” family (≈8 sites) — ruled stative attribution, not agency. (4) L03's “water vapor” scene claim — true, value-free, convention pinned at rev 1.9 (qa/REVIEW_LESSON_S07 §5).
Also deliberate and on the record: L01/L07's complement item pair reads cleanly in both stems.
My recommendation: All four stand.
(1) Item 18's distractor “The bond energy.” — the lesson reserves the two names first; one tap narrows the option. (2) “Our two totals” article voice. (3) The “carries a plus sign” family — stative, with the S07 “belongs” ruling. (4) The L01/L02 notes now say “of the atoms”; the CED's own noun is “system” — one word either way, course-wide. (5) One signRetry rides the positive +86 key, beyond the negative-keys-only convention (qa/REVIEW_LESSON_S08 §5; SPEC_RECORD_U6-S08).
My recommendation: All five stand.
S08 L01/L02 reuse the licensed U2 render with its double-headed arrow; the blueprint imagined an up-arrow/down-arrow pair (SPEC_RECORD_U6-S08).
The prose carries the direction meanwhile; a directional-arrow param is a one-tap template change.
My recommendation: Keep the shared render.
Accuracy found true exceptions, so the absolute was softened at register rev 2.0, with the L06/L07/FRQ echoes following (ADJUDICATIONS §E-S08 E2).
My recommendation: It stands.
“Demands” sits beside your trigger list's “insists”; the balance is a constraint, not an actor (qa/REVIEW_LESSON_S09 §5).
Carried on the same lexicon card, awaiting your word on the trigger list: “you let X choose/decide” (two sequences now) and “the subtraction lands at…”.
My recommendation: Keep “requires”; add the let-X-decide trigger.
The new arrow-vs-label guard forced S08's two tight renders to the one-depiction equation form; S02's teaching run keeps role words (ADJUDICATIONS §E-S07b/§E-S08b coordination note).
The divergence is deliberate — teaching surfaces name roles, application surfaces name species — and flagged for one tap.
My recommendation: Keep both styles.
Your 8 Oct rule prefers energy-transfer wording; the exam prints the noun, and the course keeps the exam's form (ADJUDICATIONS §E-S04b; qa/REVIEW_LESSON_S04 §5).
Applies forward to Units 7–9 either way.
My recommendation: Keep the exam's noun.
This course opens wrong-option feedback “This option is wrong — you …”; the vault's newer R156 says reasoning only, no verdict prefix.
Units 4, 5 and now 6 all shipped the contract form (17 openers in S04 alone); every reviewer routed it here rather than fork the course mid-build (qa/REVIEW_LESSON_S04 R-1).
Your ruling sets the voice for Units 7–9 and any feedback pass over U4–U6.
My recommendation: Keep the course opener — one voice per course.
The cap was applied to wrong-option feedback unit-wide; four S01 key-feedback lines and S03's 37-word item 8 stand on the scope ruling that the cap covers wrong options only (ADJUDICATIONS §E-S01b F2, §E-S03 addendum).
My recommendation: Wrong options only — a key's reasoning earns its length.
A why item whose stem states the verdict tests the reason, not the verdict: S01's explain items run it by blueprint design, S06 uses it at checks 14/25, S03's check 15 shares the shape.
The same-case-vs-fresh-case question is recorded open (qa/REVIEW_QUESTIONS_S06, check 29).
My recommendation: Bless the form.
(1) “about 2 °C” → “about 2.4 °C” in L04/L05, matching the iron's 2-s.f. companion line. (2) “A water bath contains 2.00 mol of water” → “A beaker holds…” (36 g is no bath). Both non-blocking Leaves, listed for you (qa/REVIEW_ACCURACY_S04 rows 2 and 7).
My recommendation: Leave both — the adopted blueprint values stand.
The eqstack family renders on a fixed 620-unit frame floor — content-tight frames are one line if ever wanted, at the cost of re-rendering every instance (qa/REVIEW_FIGURES_S10).
The ½-glyph vs n/m-slash coefficient convention is held course-wide on purpose; the standing Leave bars fixing one instance (the P3 round re-confirmed it).
PENDING since Unit 1 (the 2021 Q2 row in past_exams/USAGE.md).
It now gates three one-U6-part shorts — 2022 Q4(c), 2023 Q6(b.ii), 2025 Q4(c.ii): spending one part on a U6 page splits a question across units (qa/ASSESSMENT_PREP.md §6.1).
My recommendation: Whole questions only; the three shorts stay unreserved.
Parts (a)/(c)/(d) are pure Unit 6; (b) is backward retrieval only (U3 particle speeds — no forward content), so the test stays topic-pure in your ruling's sense (qa/ASSESSMENT_PREP.md §6.3).
The alternative is the U1-style pure-parts excerpt.
My recommendation: Keep the whole question.
A new last-taught-U6 short sits in the registry, but the 2026 scoring guidelines are not archived, so it cannot be converted (qa/ASSESSMENT_PREP.md §6.4).
It will be offered for practice once the SG lands in past_exams/2026/.
The alt never speaks the scored 38.5 — it says where the column top sits, and the student earns the read by converting position to value, the item's own skill (ADJUDICATIONS §E-T1 E5).
Dropping it would blind a screen-reader student entirely. The blind flag vs examiner divergence is recorded; this joins the standing course-wide blind-description question carried from U5/S01.
My recommendation: Keep the alt as the accessible data channel.
A numeric item gives exactly ONE substantive retry (a sign nudge or a named wrong-answer nudge) and that retry consumes the try (build_course.py submit(); the G5 drives were restructured after reading it verbatim).
Format slips are free retries, never fails: not-a-number, right value at the wrong precision — and an exact integer typed with a decimal spelling lands as a format retry, not outright acceptance (the 8 Oct decimal-spelling behaviour, on the record).
My recommendation: The contract stands.
mathtext has no token for it, and figlib's drawn ΔH_rxn subscript kerns badly — S07 and S08 both dropped planned equation figures over it; lesson titles ban the raw token (BUILD_STATE, S07/S08 specs).
One tap orders the mathtext token and the re-render; the kerning fix rides the same change.
My recommendation: Add the token at the next engine touch.
The blind reader noted a near-miss between one prompt's unit phrase and some solution routes' natural unit; the rubrics are route- and unit-agnostic and scored 10/10 cleanly (ADJUDICATIONS §E-P3, blind notes).
My recommendation: Leave it — the rubric absorbs every route.
Cards on the unit test read “Check what you just learned” / “Free-response practice” with a “Check n of 20” progress line — lesson register on a test (qa/REVIEW_TEST_T1, W4).
It is one hard-coded branch in build_course.py shared with U5-T1 — one build change fixes both units; deferred to close-out as a cross-unit cosmetic (ADJUDICATIONS §E-T1 E6).
My recommendation: Order the fix, both units.
A bundled figure ships what the drawing actually shows (plus a variant tag) — never the spec plus its flaw switch, never the flaw's name.
The S02 blind reader judged the format fit for purpose and recommended it as standard; three figure families now resolve this way, and the pattern caught two real leak classes (§E-S06, §E-G3).
Held as PROPOSED for your word (vault-sync 2026-10-09 @ e7af59d).
My recommendation: Ratify.
(1) A compared pair whose second frame elides the repeated noun into a pronoun is a REAL flag when both readings are true — the fix is the repeated noun, not a checker exemption (the S07 E3 ruling).
(2) The equation-lines / antonym-chain scope line: worked equation lines and deliberate antonym chains are exempt from the one-frame rule (the S07/S09 adjudicated exemptions).
Held as PROPOSED for your word (vault-sync 2026-10-09 @ e7af59d).
My recommendation: Ratify both.
The pack's QC seat fails natural depth-of-field on sight: 8 of the 19 granted hooks died on SHARP_NO_DOF across S01–S09, while the author seat read most takes as fine.
S10's two hooks passed on take 1 by making the scene FLAT — a wall or fence almost touching the subject, so no depth exists to blur (qa/PHOTO_QC_U6-S10).
That prompt pattern conflicts with the 2 Oct natural-focus ruling — a real tension, surfaced for you and David (the pack's owner), held as PROPOSED (vault-sync 2026-10-09 @ e7af59d).
My recommendation: Scope the seat's sharp-everywhere rule for hook photos, with David.
Applied without asking, under your option-(b) delegation: D2 (the pilot question is moot — the full-course go ended the pilot phase), D3 (rearrangement atoms stay grouped), D5 (reduced per-variable treatment for q = nΔH), D6 (work defined at first use, never assessed). The record is 05_operator_adoption.md; veto any line there.
Every call below was made on the record under your option-(b) delegation and already executed. Tap Stands to confirm, or Veto with a line. The record itself sits collapsed under each card.
B
The call: The unit's binding conventions: U5's §B set carries unchanged, plus seven U6-specific calls (B1–B7).
What changes: Every spec wave built against these; the pool ledger became coordinator-owned at process grain.
## B. Conventions confirmed for Unit 6 (bind every sequence)
- **B1(U6) — U5's §B conventions carry unchanged:** B1(U5) option counts (2–4 lesson checks,
never padded; gates/tests exactly four); B3(U5)'s carries from Unit 4 (B4 instance-pool
discipline, B7 no fluency rows on calculation lessons, B9/F1 verdict+reason split with the
≤1/lesson ≤3/sequence cap, B10 metaphor discipline, F2 sig-fig tiers from each item's own
givens, H1 scaffold-chain restatement only where the carried value is load-bearing); the
blueprint decisions-heading format B2(U5).
- **B2(U6) — why-items in lessons may use the engine's `frq` flow item** (build_course.py
treats frq as a first-class check). Type-(c) "why" items and capstone explain items run as
frq with model answer + rubric-style points, or as separate verdict/reason MCQs — the spec
author picks per item and declares the split. Bundled verdict+reason options stay capped
per B1(U6).
- **B3(U6) — numeric policy (from S04-2, S05-6, S07-7, adopted as one rule):** one-step items
key exactly; multi-step chains with real constants are computed FORWARD from the real
values (thermometer readings to 0.1 °C; keys at 3 s.f.; stems state the required
precision; wrong-map values recomputed from their named error paths; table entries spaced
>10× the rounding wobble). Supplied constants are real literature values (rounded forms
licensed, named per item) and review-accuracy checks every one. Classroom-style measured
ΔT values that land within ~2% of literature ΔH are licensed and stems say "a student
measures" (S07-7). On James's card.
- **B4(U6) — the instance-pool ledger is blueprints/POOL_LEDGER.md**, coordinator-owned,
grain = PROCESS (substance + what happens to it), per U5 §S06-10. Checks never use any
taught instance, any other sequence's reserved entry, or an undeclared cross-sequence
retrieval. Licensed taught-thread retrievals (teaching surfaces only, never checks):
methane combustion (S07/S08), lime-kiln CaCO₃ decomposition (S01 scene / S07 named +178
case), NH₄NO₃ cold pack (S01 taught / S07 L09 worked), carbon-charcoal combustion
(S02 thread / S10 capstone arm).
- **B5(U6) — photo-real hook licences: ALL NINETEEN GRANTED** (S01 ×4 iced tea, candle, hand
warmer, cold pack; S02 ×1 charcoal; S03 ×2 spoon-in-soup, hot-vs-cold mug; S04 ×1 beach;
S05 ×2 quench barrel, real coffee-cup calorimeter; S06 ×2 ice-in-hand, steam off a pan;
S07 ×2 stove flame, snapped cold pack; S08 ×1 atomic-hydrogen torch; S09 ×2
graphite-and-diamond, spirit burner; S10 ×2 snowbank, charcoal grill). Grounds: each
carries a felt scene and no data; everything exact stays figlib (James's same-day rule,
2 Oct decision). Conditions: David's pack, FILE-REFERENCED never base64, independently
QC'd (qc-figure) then review-figures on the render, each listed in its sequence record.
The count (19 vs U5's 8) sits on James's card — he may strike any to monochrome figlib.
- **B6(U6) — "energy change" in words until THC-061 coins ΔH** (register scope decision;
corrects the coordinator's own S02 brief, which wrongly said "ΔE/ΔH arrow" — the register
governs, the brief did not). No ΔE anywhere in Unit 6.
- **B7(U6) — student-facing titles never carry banned agency phrasings even where the
register's operator-facing title cells do.** Executed for THC-021 ("Why the hotter body
loses energy") and THC-024 ("Predict the transfer and the final temperature"); the
register keeps its own cells; SPEC_CONTRACT already warns authors. Watch item for the S07
spec: the blueprint's L04 title "Why the energy moves the way it does" must clear the
heat-agency/quantity-never-acts lexicon on validation — if it hits, retitle to the same
pattern ("Why the energy ends up where it does" or similar), keys untouched.C
The call: All ten blueprints' parked decisions answered on the record — the §S01–§S10 rulings.
What changes: Register rev 1.5 (misconception registration, verbatim-reuse extension, the sublimation fence) executed with the record.
## C. Per-sequence rulings (numbered as in each blueprint's decisions section)
### §S01 (13 lessons)
1. **Four photo hooks: GRANTED** (B5). 2. **Heat-or-work channel verdicts: YES** — naming
the channel on a described case IS THC-008's contrast behaviour; the exclusion bars work as
an assessed QUANTITY and work-only item sets; zero numeric work anywhere; keyed drivers
stated in the stem. 3. **Constant-temperature case keyed "the observation does not say":
YES** — the atom never claims the converse, and the honest third option protects THC-050/051
from a planted belief. 4. **Energy-unchanged case as a described hypothetical: YES** — the
register's own bracketed example is abstract; naming a real near-zero reaction would need
untaught data. 5. **L13 six checks (4 verdict + 2 explain): YES** — U5 capstone precedent;
the explain items run per B2(U6). 6. **The two dissolution equations shown once as context:
YES** — U4 prior knowledge, never assessed here. 7. **Cold-pack salt unnamed until L11:
YES** — each lesson stays on its own atom; L11 connects the name. 8. **The kiln dual-system
case: YES** — one change between neighbours is exactly the discipline, and choosing the
system is THC-001's own taught move (retrieval, not front-loading); THC-043 formalises the
dissolution case later. On James's card with the two-separate-cases alternative. 9. **Pools
reserved: YES**, via POOL_LEDGER.md (S01 keeps the self-heating can; S02 swaps — see the
ledger). 10. **Thirteen scenes: STAND.** Accuracy flags routed to REVIEW_ACCURACY_S01:
kiln framing, the sherbet endothermic claim, CaCl₂/KCl/urea dissolution signs.
### §S02 (5 lessons)
1. **Equal-levels case keyed "the system's energy does not change": YES** — restates
THC-007's third outcome, coins nothing, and is the example sequence's same-object boundary.
One per quiz. 2. **Gridline axes in lesson checks, exam-drawn figures from practice on:
YES** — teaching form in lessons, exam fidelity at assessment surfaces (meta-directive
credited). 3. **Words-only "energy change" label: YES** (B6(U6)). 4. **Charcoal photo:
GRANTED** (B5). 5. **Taught instances + pools: YES** via the ledger (self-heating can ceded
to S01). 6. **Draw item supplies the size unsigned; the student earns side, direction and
sign: YES** — constants-supplied rule honoured while the atom's whole behaviour is assessed.
7. **Five scenes: STAND.** Charcoal −394 kJ (rounded from −393.5) to REVIEW_ACCURACY_S02
with the per-mole convention stated at spec.
### §S03 (6 lessons)
1. **Two photo hooks: GRANTED** (B5); the forge scene stays prose as proposed. 2. **Student
retitles of THC-021/024: YES** (B7(U6)); register cells unchanged; on James's card.
3. **"The cold gets in" registered as THC-021's documented misconception + ONE bare Yes/No
claim item in L03: YES** — the vault's own heat rule documents the belief, so this is
registration, not invention. Register rev 1.5 executes it. 4. **MECH-HEAT-COLLISION verbatim
reuse in the capstone: YES** — verbatim beats paraphrase for a protected sentence;
usedVerbatimIn += THC-024 (rev 1.5); the engine's generated canonical check follows the
register automatically. 5. **Described joule amounts as givens, comparisons only: YES** —
THC-023/024's scope works from energy amounts; no equation exists here; carries L06's
six-check licence (U5 capstone precedent). 6. **Thermal-equilibrium checks on touching
bodies only: YES** — the register defines the term over thermal contact; no scope fight.
7. **Final-temperature deferral line: STANDS** exactly as drafted (no forward term).
8. **heatcontact + convergecurves with the listed refusal guarantees, particleboxes ported:
YES** — U5 family discipline. 9. **Pools: YES** via the ledger. 10. **Six scenes: STAND.**
### §S04 (15 lessons)
1. **Beach photo: GRANTED** (B5). 2. **Rounded real c values, named per item: YES** (B3(U6);
review-accuracy verifies each against literature). 3. **No fluency row on the L08 ΔT coin:
YES** — B7 + the U5 integrated-rate-law precedent; the signed numeric run is the drill.
4. **Single-ask ranking items: YES** — the vault's one-plain-question rule; largest +
smallest + reversed-ask walks the ranking. 5. **The aluminium-match payoff line: YES** —
one closing line + a three-row supplied table, never assessed; shows why anyone solves for
c. 6. **signRetry on every negative key, none elsewhere: YES** — exactly the contract's
rule. 7. **Pools: YES** via the ledger (lead's two processes split — see ledger). 8. **One
all-positive L09 numeric (q = 83.6 J): YES** — the 7 Oct find-the-quantity rule satisfied
without front-running L10's sign convention. 9. **No misconception items: YES** — none
documented; the traps live as keyed boundary cases; the 7 Oct rule bars invention.
10. **Fifteen scenes: STAND.**
### §S05 (7 lessons)
1. **Both photos: GRANTED** (B5). 2. **Identify-the-metal capstone frame: YES** — the AP
lab's own use of the calculation (meta-directive credited); 0.903-vs-0.897 within 1%,
entries far apart. 3. **Mixture mass + water's c, assumption PRINTED in every such stem:
YES** — the AP convention, stated honestly. 4. **S05 stays calorimeter-constant-free: YES**;
the perfect-cup line stands ("we treat the cups as absorbing nothing"). No supplied-constant
item anywhere in Unit 6's authored items this build — if a real FRQ conversion later carries
one, it goes to James at that conversion (the exclusion's supplied-constant licence is for
real exam material, not authored items on an untaught routine). 5. **Open-container boundary
item: YES** — the validity condition is THC-041's own clause. 6. **Rounding policy: YES**
(B3(U6); on James's card). 7. **Pools: YES** via the ledger. 8. **Seven scenes: STAND.**
Spec-wave note carried from the blueprint: THC-041's register clause "no energy escapes
elsewhere" is paraphrased in student text as "exchanges energy with nothing else" (S43
sweep); the register sentence is not student-quoted.
### §S06 (14 lessons)
1. **Both photos: GRANTED** (B5). 2. **Capstone at eight checks: YES** — the segment-decision
items ARE the routine's categorisation step drilled first (design-check-sequence); U5
precedent licenses the over-four warning. 3. **One cooling bare item keyed with a negative
total: YES** — the register says "a path that crosses one or more phase changes"; the sign
machinery is taught; AP runs cooling paths (meta-directive). Molten-lead instance per the
ledger. 4. **The register's five-segment ice→steam example: MOVES to the practice page and
the gate pool** — the lesson teaches the routine at three stretches (one new difficulty at
a time); the five-stretch run is the same routine run longer, exactly what practice and
gates are for. BINDING on the U6-G3 builder and the practice builder: the five-segment
path MUST appear in at least one of those surfaces. 5. **Pools: YES** via the ledger.
6. **The salt-water boundary item: YES** — keyed by the register's own pure-substance
clause; no colligative theory taught. 7. **Sublimation stays out of S06: YES** — THC-085
(S10) coins it; the forward-term fence now protects it (rev 1.5). 8. **Fourteen scenes:
STAND.**
### §S07 (9 lessons)
1. **Both photos: GRANTED** (B5). 2. **One methane-stove thread L01–L08: YES** — the
register's own bracketed examples name methane at four atoms; one dataset per stage (U5
zinc-thread precedent); checks stay on the ledger pools. 3. **Name the lime kiln as the
+178 kJ case: YES** — the name-the-specific-thing rule; CaCO₃ → CaO + CO₂ is the real
reaction with that value (REVIEW_ACCURACY_S07 verifies); declared retrieval of S01's kiln
scene (B4(U6)). 4. **HCl–NaOH as L09's scaffolded case: YES** — the standard n = M·V
solution-calorimetry case; the architecture gives U8 the capstone for ITS practice, it does
not bar the instance here; no acid–base vocabulary beyond the names. 5. **NH₄NO₃ worked
thread: YES** — declared deliberate retrieval (B4(U6)); checks sit elsewhere. 6. **Capstones
at seven and eight checks: YES** (U5 precedent). 7. **Classroom-style measured ΔT within 2%:
YES** (B3(U6); stems say "a student measures"; REVIEW_ACCURACY_S07 checks plausibility; on
James's card). 8. **No misconception items: YES** — none documented; the sign-flip error is
a confronted wrong-map. 9. **kJ/mol_rxn after coining the plain phrase: YES** — the CED and
the exam write it (meta-directive credited); L01 coins "kilojoules per mole of reaction"
first, symbol arrives as its label. 10. **Nine scenes: STAND**, with the B7(U6) watch item
on the L04 title.
### §S08 (8 lessons)
1. **Torch scene + photo: GRANTED** (B5) — REVIEW_ACCURACY_S08 verifies the device physics
(real device; the welding heat is H-atom recombination). 2. **Diatomic non-example in L04:
YES** — it is the mechanism's own minimal pair (averaged entries drift, single-molecule
entries do not); numbers to review-accuracy; the CED-nuance caveat noted on James's card.
3. **All-bonds convention ONLY, no changed-bonds shortcut line: ADOPTED (the stricter
option)** — one convention keeps the routine drillable; the shortcut reaches the same
answer, so nothing an AP item needs is lost; the hydrogenation case counts its unchanged
C–H bonds on both sides by design. James may add the honest-shortcut line with one word.
4. **Six checks at L05, five at L07: YES** — multi-step routines under the 2026-09-25
scaffolding rule; U5 precedent. 5. **The N₂+O₂ engine case stays: YES** — the coarse
estimate (+225 vs ≈+180) is ON-THEME for the estimates lesson and gets one honest line
reusing L04's point; REVIEW_ACCURACY_S08 verifies both values. 6. **Pools: YES** via the
ledger (NH₃ ceded to S10 already; HCl/HBr formations owned here). 7. **Eight scenes:
STAND.**
### §S09 (7 lessons)
1. **Both photos: GRANTED** (B5). 2. **Three-level shared-baseline diagram: NO for this
build** — the two-level boundary stays clean; the worked lane and the growing table carry
L05; the licence question goes on James's card (one tap adds it later as a table-family
figure). 3. **Growing formation table: YES** — L03's zero convention lands as its own
reveal; the one-lesson O₂ wonder is the headache working as designed. 4. **Physical +
non-combustion bare items in the capstone: YES** — the atom statement says "chemical or
physical process"; the two cases carry the exact sign-slip spots (meta-directive credited).
5. **Ozone's formation equation as the surprising yes-case: YES** — it IS a formation
equation of an element's non-standard form; the instance reuses a taught substance in a
different performance (licensed); value (+142.7 kJ/mol) to REVIEW_ACCURACY_S09. 6. **The
whole-number-relaxation lines: STAND as drafted** (06_crossref item 2 executed). 7. **Pools:
YES** with the CH₄ → SiH₄ swap (the author's own offer; methane is S07/S08's taught
thread) — ledger executes it. 8. **Seven scenes: STAND.**
### §S10 (10 lessons)
1. **Both photos: GRANTED** (B5); the charcoal-grill scene is a DECLARED retrieval of S02's
charcoal thread (B4(U6)). 2. **glossaryAudit += sublimation → THC-085: YES** — the S04-1
precedent (integrated rate law); register rev 1.5 + validator re-run executed with this
record; the forward fence now protects the term (and bars it from S06, consistent with
§S06-7). 3. **Fractional coefficients outside formation equations, ONE recap line at L05:
YES** — the 06_crossref item-3 constraint honoured; the line as proposed. 4. **No ΔH_sub
symbol — words only: YES** — the register audits no such coin; the one-coin boundary holds.
5. **One bond-table-unusable (solid species) case in L10: YES** — THC-072's gas-phase scope
is taught, so the discrimination is real, not invented. 6. **One both-tables tiebreak keyed
to formation enthalpies over the estimate: YES** — THC-073 teaches exactly that; the reason
sampled in the SEPARATE reason item (meta-directive credited: AP asks reliability questions).
7. **Zero verdict+reason bundles: YES.** 8. **Pools: YES** via the ledger (HBr formation
ceded to S08; swap per ledger). 9. **Ten scenes + real enthalpy choices: STAND**;
REVIEW_ACCURACY_S10 confirms every supplied value and sum.§S03-S
The call: S03's two spec-stage deviations adopted: taught threads carry the example runs; recall MCQs sit mid-teaching per R158.
What changes: Check cases stayed as blueprinted; nothing moved downstream.
## §S03-S — S03 spec-stage deviations (adjudicated 2026-10-08) 1. **Example runs carried by the taught threads (soup/spoon, coffee/air), not the blueprint's tile-and-bench / thermometer-in-roast: ADOPTED** — the blueprint had a within-lesson B4 collision (run instance = check instance); the ledger's "distinct" note resolves it exactly this way; check cases stay as blueprinted. The single-slot frame rewrite after the check_parallel FLAG is the 2026-10-08 parallelism bar working. 2. **Recall MCQs placed mid-teaching (after the headache, before the line that uses the fact) per R158, not as literal first check: ADOPTED** — R158 is the governing vault rule; the blueprint's "first check" phrasing was shorthand, not a contrary ruling.
§F-POOLS
The call: The pool ledger's nine ambiguities ruled: three licences declared, two entries stand, two retrofits, one echo kept, one fresh case ordered.
What changes: POOL_LEDGER.md carries the rulings; S01's propane-torch item became an oil lamp.
## §F-POOLS — coordinator rulings on POOL_LEDGER.md's nine ambiguities (2026-10-08) - **A1 iron-oxidation family: LICENCE DECLARED** — S01 hand-warmer (taught) + S02 L05 rusting sketch (taught) + S01 L01 rusting-nail check, which stays: it tests the system/surroundings PARTITION, not the energetics (different-atom licence, B4(U5) precedent). - **A2 ice-on-palm: LICENSED taught-thread bridge** S01 L04 → S06 L01 (the S06 hook deliberately leans on THC-004 retrieval). - **A3 iron-block-into-water: LICENSED thread** — S03 L06 poses the final-temperature question, S05 L07 resolves it; that arc is the design. - **A4 S07 NaOH bare item: C22 STANDS (swapped)** — NaOH dissolution is S01-taught; the calculation-lane argument does not clear B4. S07's spec wave takes the ledger's replacement. - **A5 N₂+O₂ step datum: C-NO2 STANDS** — S10's toolbox reuses its own carbon case. - **A6 methane licence: EXTENDED to name S10 L09** (register-bound −890.3 Hess step); teaching surfaces only, never checks. - **A7 S01 retrofits: item 20 is ALREADY re-instanced** (§E-S01 E10, burning log in a fireplace) — the ledger's match-burning suggestion is OVERTAKEN; wood combustion is shared WITHIN-OWNER with S01's campfire item (distinct scenes, same sequence — licensed, pools are per-sequence property). The propane-torch item RETROFITS to an oil lamp burning lamp oil (same job: combustion exo verdict, no data; propane combustion is S09-taught) — batched into the S01 F-application round after review-lesson lands. - **A8 aluminium echo: DELIBERATE and kept** — S04's unassessed closing line foreshadows, S05's worked example performs; no check collision. - **A9 S06 L03 transfer slot: the S06 spec wave picks a fresh condensation case** (the milk-bottle entry reusable within-owner if L02 leaves it unspent) — goes in the S06 author brief.
§E-S01
The call: S01's first round: one Essential (the concrete-sets stem gained a surroundings witness) and eleven more rows, all executed.
What changes: The fb-voice and explain-form questions went to your cards; everything else is applied and re-read.
## §E-S01 — S01 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/3W/6L, questions 1E/5W, cold-solve 54/54)
- **E1 (questions Essential, item 27 concrete-sets):** ADOPT the stem rewrite — "Fresh
concrete sets on a building site, and the air just above the slab warms. Is the setting
reaction endothermic or exothermic?" — with both feedback lines recomputed. Grounds: the
sibling items all give a surroundings witness; the rewrite is smaller than the L06
criterion-sentence route (that alternative stays on James's card in the record).
- **E2 (accuracy W1, L10):** ADOPT "The skin and the air lost exactly the energy the salt
gained." — the lesson's own figure and nq already say both.
- **E3 (accuracy W2, scenes family):** TEMPLATE-LEVEL fix — figlib_u6_scenes suppresses the
dashed system boundary whenever no system/boundary label is supplied; L05's box, when
drawn by its own lesson, encloses the burning wax and wick only. Re-render + re-screenshot
all affected scenes; REVIEW_FIGURES_S01 re-checks on the render.
- **E4 (accuracy W3, L08):** ADOPT "How does the energy that lifts the lid move from the
steam to the lid?" Key untouched.
- **E5 (questions W, item 18):** duplicate distractor feedback split — each wrong option's
fb recomputed from ITS named error path.
- **E6 (questions W, item 37):** key trimmed to "The campers and the air gain it."
- **E7 (questions W, item 41):** distractor recast to the exact reverse trade, fb recomputed.
- **E8 (questions W, L04/L05 Y-N pattern):** L05 run reordered to charcoal→dew→frost→ice→term
(keys/content untouched; kills the shared Y N Y Y shape).
- **E9 (questions W, item 54 rubric):** pt 1 reads "names at least one of the flask, the
moisture, the board, or the nearby air" (disjunctive).
- **E10 (B4 forward-guard, item 20):** the charcoal-briquette check shares the
charcoal-combustion process S02 teaches on (its taught thread). RE-INSTANCE item 20 onto a
burning log (wood fire) — no sequence claims it; fb recomputed; POOL_LEDGER gains the line.
- **E11 (engine, coordinator-executed):** the numeric retry string pluralises blindly
("1 significant figures") — made number-aware at template level; mutation untouched
(string cosmetic), clickthrough re-verifies.
- **E12 (register, coordinator-executed, rev 1.6):** scopeDecisions gains the S01 energy-
reading scope line (accuracy register-finding (a)): in THC-002/003 lessons "energy"
means the thermally read energy — what the thermometer reports; THC-043's
system-convention closes the bookkeeping. Register finding (b) (touching-things heat
coinage; no radiant channel items unit-wide) recorded as a standing authoring constraint —
already relayed to the S03 author; SPEC_CONTRACT note at next engine touch.
- Accuracy's 6 Leave rows: LEAVE. Questions observation (1) (fb voice frames vs the vault
Round-3 line) stays on the existing James feedback-voice card (U4/U5 lineage); observation
(4) (verdict-stated-in-stem explain items) recorded for James per the blueprint's design.§E-S01b
The call: The S01 lesson pass's seven fixes adopted as written; the blind read's bridge-item flag held for the application.
What changes: The application batched these with the figure rows below.
## §E-S01b — review-lesson work order (adjudicated 2026-10-08) Role A 9/9. F1–F7 ADOPTED as written in qa/REVIEW_LESSON_S01.md: F1 item 26 propane torch → oil lamp (executes §F-POOLS A7; key/position unchanged; ledger line); F2 the eleven wrong-option fb trims to ≤30 words (exact trims from the record; the four KEY-fb lengths stay — U5 F4's scope was wrong options — on James's card); F3 L04 recap split; F4 item-25 fb recast on the beaker alone (one-line accuracy confirm ordered with the application); F5 L11 headache closed with the parallel pair; F6 kettle stem thermometer fix; F7 "settles the label". Blind-read F2 (item 28) STANDS per the reviewer's ruling. Application agent batches F1–F7 + any REVIEW_FIGURES_S01 fix rows, then: lexicon sweeps, clickthrough, CHANGED_IDS.md (bare ids, no key-derivable prose — the U5 catch), fresh cold re-read of changed items.
§E-S01c
The call: S01's figures: four template faults in the scenes family, all fixed at template level; 20 of 28 figures were already clean.
What changes: Every affected instance re-rendered and re-opened in both themes.
## §E-S01c — review-figures work order (adjudicated 2026-10-08; 20/28 clean, 0 Essential, give-away 0) G1–G4 ADOPTED, all template-level in figlib_u6_scenes.py: G1 arrow-tip/label slot collision (a_out tip vs label slot; one fix covers L08/L09/L10 + the L08 cork-overlap x-offset); G2 vessel_rig line-through-text (water-level vs label y; wall vs long sys labels); G3 dissolve_trade's pointer arrows must not reuse the coined energy-arrow style (plain lead lines or drop); G4 iced-tea "the glass" label slot abuts the wall. After the template fixes: re-render + re-open every affected instance light+dark. The sweep's blind spot (arrow-vs- text collisions invisible to text-only geometry) is a VAULT-SYNC CANDIDATE (extend the figure sweep to arrow-vs-text hit-testing). E3 confirmed on the render; 480 px 0 overflow.
§E-S02
The call: S02's round: both Essentials were the coordinator's own — the bundler's promised flaw neutralisation had never been implemented.
What changes: The drawn-facts format became the course standard.
## §E-S02 — S02 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/3W/4L keys 24/24, questions 2E/4W cold-solve 24/24) - **E1 (Essential, TOOLING — the coordinator's own bundler):** _blind_bundle.py strips `flaw` but its promised variantTag neutralisation was NEVER IMPLEMENTED (docstring only — the overpromise is the coordinator's, on the record). Fix IN THIS ROUND, owned by the fix agent: for any figure whose spec carries a flaw switch (and for options composites, per panel), the bundle carries the RESOLVED DRAWN FACTS (the renderer's validated level/ arrow/label values as drawn, plus a variantTag letter) — never raw spec+flaw, never the flaw's name. Prove it: cut a trial S02 bundle and verify item 19 reads as drawn-wrong, item 18's four panels are distinct with the clean one unnamed, item 24 is coherent; leak-grep extended with flaw-switch names. NO S02 bundle existed before this, so nothing was corrupted; S01/S03/S04 bundles carried no flawed figures (S04's figures are tables) — no retro-cut needed, verified by grep. - **E2 (Essential):** items 23/24's shared longest key → "The arrow is drawn from the wrong level"; re-run the vault KEY TELLS script and APPEND the dated "KEY TELLS: 0 of 19" line to qa/AUTHORING_S02.md. - **E3 (W1):** shared flaw set reordered [arrow, side, sign(, size)]; keys move 2,3 → 0,3. - **E4 (W2)** item 16 restated stand-alone; **E5 (W3)** item 19 stem harmonised; **E6 (W4)** draw checklist names the acetone. - **E7 (accuracy W1):** L02 headache fig products.e 120 → 150. **E8 (accuracy W2):** the one-flaw-broke-both-comparisons sentence added to L05's worked audit. **E9 (accuracy family-level):** figlib_u6_twolevel prints the kJ unit on gridded-read figures (axis or entry-side), and L01/L03's invisible charcoal level data ALIGN to one dataset (one scene, one truth). - Ledger note: U6 image-option clean slot used = panel 3 (item 18) — any later U6 image-option item varies the slot.
§E-S02b
The call: S02's close: one figures Essential — per-panel autoscale broke COMPARED renders; a shared-scale mechanism now pins the morph run.
What changes: The run re-opened together, light and dark; blind read 24/24.
## §E-S02b — S02 closing application round (adjudicated 2026-10-08; blind 24/24, figures 1E/2W/1L, lesson 0E/5W) - **G1 (figures Essential, template):** figlib_u6_twolevel gains a validated shared-scale mechanism (axisMax pinned to 300 across the five L02 morph-run figures; shared emax computed inside render_twolevel_options) so COMPARED renders share one scale — the "same diagram, levels swapped" caption becomes visually true and panel J's double-length arrow normalises (the two W rows fall out of the same fix). Refusal battery re-proved and extended with a planted mismatched-axisMax case; the L02 run re-opened TOGETHER light+dark. - **F1–F5 ADOPTED** as written in qa/REVIEW_LESSON_S02.md (eleven fb trims; four agency swaps incl. the one key-fb wording, no key moves; the staging opener; two sentence splits; checklist line 5 names the acetone-vapor level). - **Leave:** the U2-inherited pecurve bond-energy label clip joins the CROSS-COURSE OWINGS (shared-figlib template note, both units together — never a U6-only fork). - Application agent: G1 + F1–F5; rebuild; re-prove battery; re-render/re-open the morph run and composite; sweeps; clickthrough; bundle re-cut (item 17's checklist is bundled) with CHANGED_IDS (checklist-surface 17; fb-only 3, 4, 16, 18, 22, 23, 24; figure-only the L02 run + composite); mini re-read of the named surfaces closes S02.
§E-S03
The call: S03's round: one convergent Essential (“Nothing passes out of the spoon” → “No substance…”) and the ordered-quiz feedback-channel fix.
What changes: Item 26's one-word fix, missed by coordinator oversight, was adopted in the addendum.
## §E-S03 — S03 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 1E/3W/10L, questions 1E/7W/24L, cold-solve 32/32, keys 32/32, no key changes anywhere)
- **E1 (accuracy Essential = questions W, convergent):** L03 misconception-item key fb opens
"Nothing passes out of the spoon." — false at particle level. ADOPT the one-word fix:
"No substance passes out of the spoon."
- **E2 (questions Essential, item 12):** key fb + distractor fb together define item 13's
both options (ordered-quiz feedback channel, 2026-09-29 rule). ADOPT the record's two fb
rewrites; the synonym teaching stays in lesson text; no key moves.
- **E3 (accuracy W1):** L04 cooling series → "60 °C, 40 °C, 28 °C, then 21 °C" (drops must
shrink — the law L05 teaches).
- **E4 (accuracy W2):** L05 convergecurves final 70 → 55 (the soup/spoon thread's ONE
endpoint; L03's 65/45 and L04's 60/50 straddle it, nothing else moves).
- **E5 (accuracy W3):** "at its open surface".
- **E6 (questions W rows 3/6/8/16):** ADOPT the record's exact rewrites.
- **E7 (questions W, bricks fb):** "Each brick is in thermal contact with the bench and the
air it touches, not with the other brick."
- **E8 (questions W, items 25/32):** the two ", so" key-shape trims, meaning unchanged.
- Leaves stand (vacuum-flask radiative leak on the deliberately-omitted list; the licensed
900 J/50 J item; check-15-shape why-stem recorded for James).
- **Register/ledger actions (coordinator, with this record):** POOL_LEDGER gains (i) the
soup/spoon thread endpoint datum "settles at 55 °C" (figures may not drift); (ii) the A3
arc constraint — S05 L07's resolving dataset must be consistent with S03 L06's drawn
mid-course panel (iron 90→70 °C while water 15→30 °C ⇒ body heat capacities
C_iron ≈ 0.75·C_water, final ≈ 47 °C; with real c values that is roughly 7 kg of iron
per litre of water — the S05 author picks exact values, REVIEW_ACCURACY_S05 verifies);
(iii) wording note: "glowing"/"radiator" are permitted appliance words on contact items
(the contact, not radiation, carries the taught transfer).
### §E-S03 addendum (coordinator confirmations, 2026-10-08)
- The fix agent's E6/E1 conflict resolution on item 16 is CONFIRMED — E1's convergent
one-word ruling governs that fb line; E6 applied to items 3/6/8 only.
- Item 8's key fb at 37 words STANDS — §E-S01b's scope ruling (the ≤30 cap binds wrong-option
fb; key-fb lengths are on James's card) applies unit-wide.
- Questions-W item 26 ("identical aluminum blocks", one-word) was dropped from E1–E8 by
COORDINATOR OVERSIGHT, not ruling — ADOPTED now; applied with the S03 review-lesson
application round (no separate micro-round).§E-S03b
The call: S03's close: five lesson rows, three figure template fixes, the bricks stem made distance explicit — and the radiation clause DECLINED.
What changes: S03 closed on a no-faults mini re-read.
## §E-S03b — S03 closing application round (adjudicated 2026-10-08; blind 32/32, figures 0E/3W, lesson Role A 8/8 + 0E/5W) - **F1–F5 ADOPTED** as written in qa/REVIEW_LESSON_S03.md (F1 = the carried item-26 word; F2 "settle"; F3 the second-mug sentence; F4 the two named-substance restatements; F5 the two 31-word fb trims). - **G1–G3 ADOPTED** from qa/REVIEW_FIGURES_S03.md: G1 inset caption drops below the panel (template, figlib_u6_heatflowfam); G2 convergecurves note moves to the left third (template — also clears the blind read's strike-through flag and the axis-skimming label, re-check both on the re-render); G3 is F3 (one spec sentence, same fix). - **Blind flag 1 ADOPTED**: bricks item stem → the two bricks rest on two separate benches, a hand's width apart (kills the indirect-path reading before the answer; key/options/fb untouched beyond what E7 already set). - **Blind flag 2 DECLINED**: no radiation-acknowledging clause in the vacuum-flask fb — the radiative leak stays on the deliberately-omitted list (accuracy's ruling); adding it would coin an untaught channel. Recorded for James with the Leave. - **S01 four-site retrofit ADOPTED** (routed observation): specs/U6-S01.json's "touching things" ×4 → "touching bodies", aligning the heat coinage with item 6's adjudicated fix, S03's taught wording and the rev 1.6 constraint. S01 is closed but NOT blessed — still in its authoring chain, so current rules apply (step-0 freeze check consulted). Prose-only; S01 lexicon sweeps re-run; noted in S01's CHANGED record trail. - Application agent: apply all of the above, re-render affected figures light+dark, sweeps, clickthrough, blind/20261008_S03/CHANGED_IDS update (stem changes: 26 AND the bricks item — both get full re-reads; fb-only: 4, 31), bundle regen leak-free, then a mini cold re-read closes S03.
§E-S04
The call: S04's round: zero Essentials; six question rows adopted (the tense run, the option reorder, the warming pie).
What changes: The two optional parity edits went to your card, not this round.
## §E-S04 — S04 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/0W/5L, questions 0E/6W, cold-solve 62/62, keys 62/62)
All six questions-W rows ADOPTED as written in qa/REVIEW_QUESTIONS_S04.md: (1) items 1–4
present-tense harmonisation; (2) item 8 active voice (the plumber); (3) item 23 stem →
"What is the first sample's temperature change?"; (4) item 29's 26.5 fb names the addition
error; (5) item 38 options reordered m/c/q/ΔT, key → C (kills the L09 4-option B-run; unit
4-option spread 0/2/2/0); (6) item 41 flipped to "A pie warms in an oven" (q > 0; fluency
string + + − −; also de-duplicates the cooling-pie scene vs L01). Accuracy's 5
Simplification-Leaves stand; its two optional parity items ("about 2.4 °C"; the L14 beaker
swap) go to James's cards, not this round. Standing items routed: fb-prefix voice (existing
card); KEY-TELLS spec-JSON mode added to the CLOSE-OUT TOOLING LIST (third sequence where
the set-level hunt stood in). Fix agent: apply the six rows, re-cold-solve changed items,
re-verify strings/positions, sweeps, bundle cut + CHANGED_IDS (stem changes 1–4, 8, 23, 41
full re-read; option-order 38; fb-only 29).§E-S04b
The call: S04's close: three items' answers were printed on the page above them — capstones re-instanced to silver and tin, the block renumbered.
What changes: Ledger: silver spent, tin entered; q-as-heat went to your card.
## §E-S04b — S04 closing application round (adjudicated 2026-10-08; blind 62/62, figures 0E/2W, lesson 4E/20W)
- **E1–E4 ADOPTED** as written in qa/REVIEW_LESSON_S04.md (E1 item 61 → silver 25.3, the
C-Ag ledger entry spent; E2 item 62 → tin 0.227 — iron correctly rejected under B4; E3
item 50 renumbered 300.0 g/−1,155 J with key −10.0 unchanged; E4 the layout-term prose
line). One-line accuracy confirms on the silver and tin CRC values ordered with the
application.
- **F1–F20 ADOPTED** as written (incl. the beach-headache payoff with the corrected
sea-water-c claim, the worked-label normalisation, the elided-subject family, and the
three fb trims).
- **F21 (coordinator, from the blind read's sustained flags):** item 24's stem rewritten so
the "equal" key is not presupposed away ("How does the first sample's temperature change
compare with the second's?" or the blind record's own wording); item 20/55 double
possessives are already F-rows; L04-15's close-capacity pair stays (touch-only rule — the
item is untouched; on the Leave list).
- **G1–G2 ADOPTED** from qa/REVIEW_FIGURES_S04.md: G1 _glyph_beaker_tub full-size beaker
("the same beaker" drawn the same); G2 figlib.py render_anneq viewBox cropped to content
extent + margin (the U5 copy shares the fault — already on the cross-course owings list).
- **Routed:** the 17 fb-prefix openers stay on James's fb-voice card, R156 reference added;
"q-as-heat noun vs the 8 Oct heat rule" joins James's cards (AP exam noun kept, applies
forward). Ledger: C-Ag marked SPENT (S04 item 61); NEW entry tin molar-capacity
conversion (S04 item 62) — distinct process from S06's tin pool entry, noted.
- Application agent: E1–E4 + F1–F21 + G1–G2, re-render + re-open affected figures
light+dark (incl. every anneq instance after the crop — all sequences use the template:
re-open S01/S02's eqstack/anneq renders too and report), sweeps, clickthrough,
CHANGED_IDS (full re-read 20, 24, 43, 50, 55, 61, 62; fb-only 17, 21, 22, 23, 25, 29,
44, 60) checked by build_tools_check_changed_ids.py if present, bundle re-cut leak-free;
mini re-read closes S04.§E-S05
The call: S05's round: the ‘any dissolution’ absolute scoped, the zinc knife-edge re-picked to 33.4 °C, eleven W rows adopted.
What changes: KEY TELLS 0 of 45 on the record.
## §E-S05 — S05 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 1E/2W, questions 2E/11W, cold-solve 54/54, keys 54/54, A3 consistent)
- **E1 (accuracy Essential):** the L05 "any dissolution" absolute takes the scope line
("When the thermometer moves, one glance sorts the dissolution…"); NaCl item and key stand.
- **E2 (questions Essential, item 21):** stem drops ", from the water's measurements".
- **E3 (questions Essential, item 44):** the one-criterion sentence enters the L06 article
("A phase transition is one substance changing its own state, with nothing mixed in.");
item stands; fb(a) rewritten per the record.
- **E4 (accuracy W, zinc knife-edge):** RE-PICK adopted — start 90.0 °C, key 33.4 °C (both
solution routes agree; kills the edge structurally rather than patching the grader). The
changed item joins the fresh re-read list.
- **E5 (questions key-tell, item 34):** the key option's clause shape/length aligned with
its two distractors (cosmetic only; the singleton third option STAYS); the structural
question — a standing "neither" third option across the L05 run, touching the R113
two-option licence and the THC-007 three-outcome licence — goes to James's card as the
reviewer proposed. KEY TELLS re-run must print 0 of 45.
- **E6 (questions W rows 2, 4, 7, 16, 20, 23, 27, 48, 49, 54):** ADOPTED as written (incl.
the two numeric wrong-map gap fills and the L04 positive-half criterion sentence).
- **E7 (precision statements):** wherever a decimal key enforces sig-figs (−2,926.0 class),
the stem states the required precision — audit all nine numerics, add where missing. The
engine-level sig-fig-nudge observation joins the close-out tooling list.
- Accuracy's Leaves stand; the MgSO₄ hydrate caveat and the L05 standing-third-option
question are James cards. Fix agent: apply E1–E7, re-cold-solve changed items, KEY TELLS
0-line pasted, sweeps, bundle cut + CHANGED_IDS, then S05's closing passes.§E-S05b
The call: S05's close: one leftover reveal number fixed, sixteen lesson rows and the calorimeter label fix shipped; item 26's key RULED standing.
What changes: S05 closed on the items-3/41 mini re-read.
## §E-S05b — S05 closing application round (adjudicated 2026-10-08; blind 53/54 — reader error; figures 0E/1W; lesson 1E/16W, Role A 53/54 + 53/54 with the item-26 key RULED standing on 3/3 cold-reader evidence, James note carried) - **E1 (lesson Essential):** item 53's revealSteps "95.0" → "90.0" (the lone pre-re-pick leftover in the built file; reveal-only). - **F1–F16 ADOPTED** as written in qa/REVIEW_LESSON_S05.md (incl. the porridge → cooked rice re-instance with ledger C18 update; the L01 alternation kill via item 3's option flip; the two S04-F12 label-class variants to the fixed strings; the two over-cap trims). - **G1 (figures W):** figlib_u6_calorimeter's label layer places each right-side label at its anchor's height with short horizontal leaders (kills the leader crossings); both calorimeter instances re-rendered + re-opened light+dark. - **Blind flags:** item 44's structural fourth-option tell JOINS the item-34 James card (the standing none/neither design question — one answer covers both; no unilateral fork). The item-52 grader drive is ORDERED in the application's clickthrough: 0.523 must grade wrong, 0.5210-form must draw the format retry. - Item 26's never-reuse-outside-the-course caution re-affirmed (ledger note exists). - Application agent: E1 + F1–F16 + G1; zero-greps (porridge, 95.0, middle moment, two checks below, one step per line); sweeps; clickthrough incl. the grader drive; ledger C18 line; bundle re-cut leak-free; CHANGED_IDS (full re-read 3, 41; fb/reveal-only 44, 52, 53, 54); mini cold re-read of items 3 and 41 closes S05.
§E-G1
The call: G1's round: the chocolate datum cut 45 → 15 kJ per 300 g (whole chocolate is ~30% cocoa butter), plus four smaller rows.
What changes: A fresh cold re-read of items 3, 5, 6, 7 closed the gate.
## §E-G1 — U6-G1 fix round (adjudicated 2026-10-08; cold-solve 8/8 from the render, blind 8/8, 0 Essential, 4 W + 1 blind flag) - **E1 (item 3):** "a thermometer in the solution shows the temperature rising" (the reading rises, not the instrument). - **E2 (item 5 fb(a)):** "the diagram whose levels drop but whose label reads +45 kJ" (disambiguates J from M). - **E3 (chocolate datum):** 45 → 15 kJ per 300 g (whole chocolate ≈30% cocoa butter; ≈45 J/g) — stem, all four panel labels, both fb echoes; panels re-rendered and the composite re-opened light+dark; the gridline-honesty eye re-applied after the relevel. - **E4 (item 7 fb(d)):** trailing clause trimmed per the record (protects item 8's key idea). - **E5 (blind flag, item 6):** option B lengthened to a parallel two-clause form per the reader's alternative (the key's hedged canonical shape stays; the length/hedge tell dies); fb adjusted; key untouched. - Then: rebuild, gate sweeps, bundle re-cut leak-free, and a fresh cold re-read of items 3, 5, 6, 7 closes U6-G1.
§E-G2
The call: G2 closed in one round: an accept list now catches unrounded correct entries — and S05's zinc twin got the same fix.
What changes: Closed on this record.
## §E-G2 — U6-G2 fix round (adjudicated + coordinator-executed 2026-10-08; blind 7/7, QA 0E/1W/4L, five values confirmed) - **W1 ADOPTED and executed:** G2 item 7 gains `"accept": ["31.58", "31.581", "31.5813"]` — the engine's valueOk honours accept lists (build_course.py:452), so an unrounded correct entry now draws the sig-fig format retry, never the generic wrong. Grading metadata only; key and student text untouched; no re-read owed. - **The S05 zinc twin (L07 item 53, key 33.4) gets the same accept list** (["33.40", "33.405", "33.4052"]) — the same fault class the G2 reviewer flagged for close-out, executed now; S05 is closed but unblessed and the change is grading metadata (step-0 freeze check consulted; recorded here and in S05's CHANGED trail). - Leaves stand (the two judged-fresh adjacencies noted to the ledger; the ordered-gate rule-feedback channel inherent to a cumulative checkpoint — recorded). **U6-G2 CLOSED** on this record.
§E-S06
The call: S06's round: the one cold-solve miss was figure-caused — a marked dot hid the 13 px segment it marked; re-instanced 380 → 450 °C.
What changes: The coordinator implemented the heatcurve bundle resolver before S06's first bundle shipped.
## §E-S06 — S06 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/0W/6L keys 76/76, questions 1E/2W cold-solve 76/77 — the one miss figure-caused) - **E1 (questions Essential, check 32):** the marked-point dot occludes the 13.1 px gas segment it marks. ADOPT the primary fix: the mercury curve's end re-instanced 380 → 450 °C (gas segment ≈47 px), tick + alt updated, check 30's option c follows 380→450; PLUS the family refusal — a marked point's segment must draw ≥36 px — with its planted suite case. Keys and stems unchanged. - **E2 (W, check 70) and E3 (W, check 12):** the two recall checks move ABOVE the idea lines that state their answers (the L05/L09/L12 order); wording/keys untouched. - Leaves stand; the 14/25 verdict-stated what→why form JOINS the existing James open-form card; the 3-opt 'a' 43% note carried to the unit roll-up. - Fix agent: E1–E3; re-render + re-open the two changed curves light+dark; refusal suite extended and green; sweeps; clickthrough; bundle cut (python3 _blind_bundle.py U6-S06 20261008_S06, leak-NONE — first S06 bundle) + CHANGED_IDS (figure-change 30, 32 full re-read; position-only 12, 70); then S06's closing passes. ### §E-S06 addendum (coordinator, 2026-10-08) - The fix agent's STOP on the bundler was RIGHT (the mandated condition worked). The coordinator implemented _resolve_heatcurve in _blind_bundle.py (drawn stretches in order from the renderer's own segment law + flaw transforms; uniform clean/flawed; variantTags; flaw names added to the leak list — one early edit aborted on a stale match and was redone; the first S06 bundle cut in that window was DISCARDED by the re-cut). Verified: composite 4 distinct panels, plain reads stripped-as-truth, leak NONE. - Check 31 SHARES the changed mercury render: it JOINS the full re-read list (30, 31, 32).
§E-S06b
The call: S06's lesson pass: the unit's only heat-agency hit fixed (“the heat for the melt”), eight rows adopted, the mirror collision resolved as a declared licence.
What changes: The 14/25 what→why form joined your open-form card.
## §E-S06b — S06 lesson-pass work order (adjudicated 2026-10-08; Role A 77/77 ×2 — three independent full passes with the blind read; 1E/10W) - **E1 ADOPTED:** L11's "the heat that melts the ice" → "the heat for the melt" (the sequence's only heat-agency hit; the 8 Oct rule binds). - **F1–F8 ADOPTED** as written (draw stem labelling clause; item-73 4-s.f. statement; item-75's 74.5 mid-chain-rounding map — key does not widen; the L02 option flip killing the a b a b open; the L03 payoff line insert-adjacent to the verbatim clause; the L08 proportionality payoff; the kJ↔J chains onto their own convert step; the seven fb trims incl. the plateau-as-actor repair). - **F9 COORDINATOR-EXECUTED:** the shower-mirror (S06 L03) / bathroom-mirror-fogging (S05 L06) process collision resolved as a DECLARED different-atom licence (§F-POOLS A1 precedent — S05's check sorts the transition, S06's explains the transfer); ledger entries merged under one process row with both licensed uses; the S06 SPEC_RECORD's "claimed by no entry" claim corrected on the record (coordinator addendum); no student text changes. - Reader flags adjudicated by the reviewer STAND (sequential-gating declines; item 4's James note; item 20's §S06-6 card). Application agent (after REVIEW_FIGURES_S06 lands): E1 + F1–F8 + any figure rows; grader drives 75/73 as ordered; bundle re-cut; CHANGED_IDS (full re-read 7, 36, 73; fb/reveal-only 37, 49, 71, 75, 76, 77); mini re-read closes S06. ### §E-S06b addendum — figures row (0E/1W, 2026-10-08) - **G1 ADOPTED (template):** figlib_u6_scenes `_scene_body` system-label leaders end on the GLYPH BODY, not the boundary rect; every system-labelled scene across S01–S06 re-rendered and re-opened (the template serves S01's whole scenes family — the re-open covers both units of use). Leaves: the no-drawn-ice beaker (honest); the two liquid depictions (S03 spread vs S06 cluster) — VAULT-SYNC CANDIDATE register note, any redraw belongs to S03's next opening, not this round.
§E-S07
The call: S07's round: the HI items' iodine state fixed g → s (+53 belongs to solid iodine); two prose absolutes softened.
What changes: The neither-vs-ΔH≈0 scope question went to your card.
## §E-S07 — S07 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/3W keys 55/55, questions 0E/1W cold-solve 55/55) - **E1 (accuracy W1):** the HI items' I₂ state symbol g → s, twice (the +53 kJ/mol_rxn value belongs to solid iodine); key 26.5 unchanged; the two items get a fresh cold re-read. - **E2 (accuracy W2):** L01's "Every reaction…" absolute softened per the record (the S01 third class survives). - **E3 (accuracy W3):** the "rarely finish" majority word → the "many … many others" form. - **E4 (questions W1):** Q41's stem "Which fraction" → "Which quantity" (options unchanged). - Register rev 1.9 COORDINATOR-EXECUTED (methane H₂O(l)/H₂O(g) convention pinned + ledger line + relay to the in-flight S08 author); the "neither vs ΔH≈0" scope question on James's cards. Fix agent: E1–E4, re-solve the two HI items + Q41, sweeps, bundle cut + CHANGED_IDS, then S07's closing passes. ### §E-S06b second addendum (coordinator, 2026-10-08) - The application agent's ONE beyond-verbatim surface (item 77's good-string "the plateau releases" → "the solidifying lead releases …") is RATIFIED — the work order's own zero-grep made it unavoidable, the actor repair matches F8g's adopted form, and the surface already sat in the fb/reveal-only CHANGED class. Properly flagged, not silent. - G1's implementation (per-glyph sys_end + refusal on label-without-anchor) EXCEEDS the order in the right direction — recorded as the template's new law.
§E-S07b
The call: S07's close: the kiln line now READS as its declared retrieval; 22 rows; the twolevel family gained an arrow-vs-label REFUSE guard.
What changes: Four wording alternatives sit on your S07 card.
## §E-S07b — S07 closing application round (adjudicated 2026-10-08; blind 55/55, lesson Role A 55/55 ×2, figures 0E/2W, lesson 1E/22W)
- **F1 (Essential) ADOPTED:** the kiln opening becomes the declared retrieval ("We've seen
that a lime kiln heats calcium carbonate until it decomposes.") — the §S07-3 licence's
own condition.
- **F2 (the delegated E3 ruling) RATIFIED:** the pronoun-explicit variant ships; the
reviewer's two-true-antecedents reasoning is exactly right and becomes the vault-sync
candidate line for [[Line up comparisons so only the difference varies]].
- **F3–F22 ADOPTED** as written (incl. the as-modified rows keeping the canonical sentence
verbatim on its own line, the five fb trims, item 17's option-noun recast with key
position unchanged).
- **G1 (figures, template) ADOPTED:** figlib_u6_twolevel gains an arrow-vs-label collision
guard (REFUSE, never draw through ink); affected specs adjust label text/level data as
needed to pass; re-render EVERY twolevel across S02/S07/S08 and re-open.
- **G2 (figures, template) ADOPTED:** the L06 particleboxes carries its kelvins in the box
labels (home-register form) and figlib_u6_heatflowfam WARNs when the drawn arrow ratio
< 1.05 (an invisible difference under a look-for-it line is a broken figure).
- James cards += item-1 wording, L06 title, the "belongs" family, the L03 water-vapour note.
Application agent: F1–F22 + G1–G2, re-render/re-open, sweeps, clickthrough, bundle re-cut,
CHANGED_IDS (full re-read 17; fb/good-only 33, 36, 44, 48, 54, 55), mini re-read closes S07.§E-S08
The call: S08's round: the −890 diagram beside gas totals now names its states (the Essential); the rehoused misconception got its fourth option.
What changes: A course-wide grep confirmed no other −890-beside-gas-numbers site.
## §E-S08 — S08 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 1E/3W keys 37/37, questions 0E/3W cold-solve 42/42)
- **E1 (accuracy Essential):** L08's −890-labelled diagram beside the 2,648/3,466 gas totals
takes the two state-naming sentences per the record (the rev 1.9 execution L04 already
models); then grep the course for other −890-beside-gas-numbers sites and report.
- **E2 (accuracy W2):** "Every entry…" → "Nearly every entry…" + the L06/L07/frq echoes (the
register's own THC-073 wording softened at rev 2.0 — coordinator-executed, James card).
- **E3 (accuracy W3):** L04 drops "each" from the 439 line (first-BDE honesty; L03's form is
the model). **E4 (accuracy W4):** the L07 table's C=O entry gains "(in CO₂)" per the rev
2.0 convention line.
- **E5 (questions W1):** Q17 gains the fourth option "Every molecule that contains a C–C
bond." (the rehoused misconception; key unchanged, lands at d).
- **E6 (questions W2):** "543" deleted from Q31's wrong{} (signRetry owns the magnitude —
the near-match leak dies).
- **E7 (questions W3):** Q34's "your two totals" → "the two totals".
- Leaves stand (Q21's fb-prints-a-later-key logged for the blind record; the generic Q41/42
licence per S07 precedent). Fix agent: E1–E7, re-solve changed items, shim re-run after
Q17's new option (0-line), sweeps, bundle cut + CHANGED_IDS, then S08's closing passes.§E-S08b
The call: S08's close: the CHANGED_IDS key-leak class RECURRED — the checker was written to the rulings' spec and run over every bundle on disk.
What changes: The checker now ships with review-lesson (vault, landed @ e7af59d).
## §E-S08b — S08 closing work order (adjudicated 2026-10-08; blind 42/42 + Role A 42/42 ×2; lesson 1E/10W; blind flags folded)
- **F1 (Essential, PROCESS) ADOPTED:** blind/20261008_S08/CHANGED_IDS.md recut to bare ids
+ re-read scopes ONLY (its key-status prose is the U5 CHANGED_IDS leak class recurring);
the checker build_tools_check_changed_ids.py PORTED into course/ (source: unit5_build/
course or the vault — find it; if neither has it, write it to the 2026-09-28/29 rulings'
spec) and run on EVERY bundle dir on disk (S01–S08, G1–G3) — report every hit. VAULT-SYNC
CANDIDATE: ship the checker with review-lesson (the class has now recurred on a fresh
course).
- **F2–F11 ADOPTED** as written in qa/REVIEW_LESSON_S08.md §4.
- **Blind flags:** L06-29's subtraction-direction stem → the "broken total minus formed
total" wording (the set's one real second-defensible risk — fixed BEFORE S10 teaches the
products-minus-reactants convention); L03-13 "in every molecule" → "in any particular
compound"; L03-18's two-names note and L04-21's scaffold-order note go to James's cards;
the 32/33-via-34 lane-order note joins the bundle record's standing line.
- SEQUENCING: the S08 application runs AFTER REVIEW_FIGURES_S08, which itself runs AFTER
the S07 application's twolevel-guard re-render (the guard touches S08's renders).
### §E-S07b/§E-S08b coordination note (coordinator, 2026-10-08)
The new twolevel arrow-vs-label guard refused two shipped S08 renders (L08 flow[12]/[16]).
The S08 owner slot was empty (its application round queued), so the COORDINATOR patched the
two labels to the one-depiction equation form ("N₂ + O₂"/"2 NO"; "CH₄ + 2 O₂"/"CO₂ + 2 H₂O";
roles:true captions carry the role words; axisMax pins kept; both pass _validate). The S07
application's re-open covers the patched renders; REVIEW_FIGURES_S08 re-checks on the
render. S09's stalled questions reviewer resumed (outage wave continues).§E-S09
The call: The canonical-core validator ruling (option B) and S09's accuracy pass: the phosphorus coin item re-instanced onto tin.
What changes: Register rev 2.1; the ΔH_f degree-mark question went to your card.
## §E-S09 (part 1 — accuracy; questions pass still out) + the canonical-core ruling (2026-10-08) - **Canonical-core (from §E-S07b's F8 conflict): OPTION (B) EXECUTED** — the validator derives each canonical requirement as the clause core (leading "because " stripped, the MECH-BOND-ENERGY halves precedent); the SHARED lexicon sentence stays as Unit 3 declared it (the U5 D5 lesson); mutation 148/148 with a standalone-core control. - **S09 accuracy (1W):** the phosphorus coin-quiz item re-instances onto TIN (grey/white, grey keyed) — sulfur declined because S(s) is already an L01 yes-case at coarser grain (the reviewer's own pool note); fb mirrored; one-item Role A cold read ordered with the fix. Register rev 2.1 executed (phosphorus scope note); the ΔH_f-no-degree convention question joins James's cards (rule before U9's ΔG° family). Fix round waits on the S09 questions pass.
§E-G3
The call: G3 closed: the bundler's clean-heatcurve field names could answer the plateau-mapping item — every heatcurve now resolves to drawn facts.
What changes: G3 and S06 bundles re-cut and proven name-free.
## §E-G3 — U6-G3 close (adjudicated + coordinator-executed 2026-10-08; blind 6/6, QA cold 6/6, 0 Essential, all values confirmed) - **W1 EXECUTED (coordinator, bundler):** ALL heatcurves now route through _resolve_heatcurve (a clean spec's mp/bp field NAMES answered the plateau-mapping item); G3 + S06 bundles re-cut, mp/bp names proven absent, leak-grep NONE. S06's already-run blind read stands (77/77 computed from stretches; the channel gap is noted on its record — no retro re-read: the reader's per-item reasoning never cited the field names). - Leave rows stand (the cross-gate c-share counts at unit sign-off — close-out list). **U6-G3 CLOSED.** Gates: G1 ✅ G2 ✅ G3 ✅ · G4 queues on S07 closure · G5 on S10.
§E-S09 part 2
The call: S09's fix round: four question rows (the grammar tell, the CaO datum, two order swaps) and the ordered decimal-alias drives.
What changes: S09's closers followed on the fresh bundle.
## §E-S09 part 2 — fix round (adjudicated 2026-10-08; accuracy 0E/1W keys 34/34, questions 0E/4W cold-solve 52/52) - **E1 (accuracy W):** the L01 phosphorus coin-quiz item re-instances onto TIN (grey/white, grey keyed — already ruled in part 1); fb mirrored; one-item cold read with the round. - **E2 (questions W1):** items 3 and 5's stems reordered to the run's one frame (kills the fronted-possessive no-case grammar tell); **E3 (W2):** the CaO stem gains its neutral state datum; **E4 (W3):** the two standalone 2-option order swaps (L02, L05; fb follow); **E5 (W4):** verify the engine's decimal-alias grading on L06's ±249 and the Hg items — drive 249.0/0.0 on the page; add ".0" wrong-map aliases ONLY if the drives show the named fb being missed. - Leaves stand (the verdict-run pool-truth strings; the H₂S one-value ruling with the blind-read flag). Fix agent: E1–E5, re-solve changed items, shim re-run (0-line), sweeps, clickthrough incl. the ordered drives, bundle cut (first S09 bundle — the growing tables flow through strip(); CHECK the bundle for any table field NAME that answers an item, the §E-G3 W1 class, and report) + bare-ids CHANGED_IDS, then S09's closing passes. ### §E-S09 part 2 reconciliation (coordinator, 2026-10-08) The E1 parenthetical "(grey/white, grey keyed)" was the COORDINATOR'S ERROR — grey tin converts into white above 13.2 °C and every table zeros WHITE tin (grey −2.1); keying grey would recreate the phosphorus error class E1 exists to fix. The fix agent's reconciliation (white keyed — the stays-form and the tabulated zero agreeing) is CONFIRMED; the within- lesson Sn-states adjacency ruling stands. The agent's refusal to apply a wrong parenthetical verbatim is the STOP-rule discipline working.
§E-S09b
The call: S09's close: six prose rows, the pencil dropped to the bench, and the case table lost its derivable column (your card) to fit a phone width.
What changes: Two mutation cases added; S09 closed on its mini re-read.
## §E-S09b — S09 closing application round (adjudicated 2026-10-08; blind 52/52, lesson 0E/6W with 6/6+6/6 changed-id reads, figures 0E/2W)
- **F1–F6 ADOPTED** as written in qa/REVIEW_LESSON_S09.md §4 (the conditional-imperative
staging form F1 is a NEW S36 sub-class — vault-sync candidate; F4's "demands"→"requires"
with the boundary question on James's lexicon card).
- **G1 (figures W1) ADOPTED:** figlib_u6_s09hooks `_glyph_graphite_diamond` drops the
pencil barrel 18 units onto the bench; re-render + re-open light/dark.
- **G2 (figures W2) ADOPTED:** the L04 case table prints states in its equation cells and
shortens the three question headers (kills the hidden-verdict-column scroll at 480 and
the clipped header at 760); all 16 verdicts unchanged; re-open at 760 + 480 both themes.
- Then: sweeps, clickthrough, rebuild; NO bundle recut owed (no stem/option/key surface
moves); a small fresh non-author re-read of the eleven changed fb/prose surfaces + the
two re-rendered figures closes S09.
### §E-S08b close (coordinator, 2026-10-08)
The re-read's one non-blocking fault (item 20's model line "the two totals" forward-
referencing L06) fixed by the coordinator with the reader's smallest fix ("makes the
computed total slightly wrong") — model-line-only, one string, rubric untouched, grep 0 on
the page. **U6-S08 GAUNTLET CLOSED.**
### §E-S09b G2-residual ruling (coordinator, 2026-10-08)
The applied G2 means cannot reach the ordered outcome (min-content 602 px vs 542/408
wrappers — the agent's arithmetic accepted). RULING, two parts: (a) ENGINE — table cells
may break an equation's NBSP run at the → arrow regardless of run length (extending the
U5 wrap-guard's arrow licence to in-table cells; mutation case owed); (b) SPEC — the L04
casetable DROPS its fifth "formation?" column: it is the AND of the three test columns,
and deriving it is the lesson's own routine; the four verdicts move to one prose line
directly under the figure ("Only the second equation passes all three tests, so only the
second is a formation equation."). The three test columns and all 12 remaining verdicts
stay. Projected min-content ≈330 px — fits 408 without scroll. On James's card (the
dropped-column design call). Re-open at 760 + 480 both themes; the S09 mini re-read then
covers it.§E-S10
The call: S10's round: the acetic scaffold pair got its stacks on-screen (the engine shows one check per screen); C(s) → C(graphite) in formation contexts.
What changes: Register rev 2.2; your card keeps the one-word C(s) alternative.
## §E-S10 — S10 QA wave 1 fix round (adjudicated 2026-10-08; accuracy 0E/3W keys 61/61, questions 1E(2 items)/4W cold-solve 59/59, mode-C SAFE) - **E1 (questions Essential):** the L09 acetic scaffold nq PAIR each gains the given-mode acetic hessstack the preceding MCQs carry (the one-check-per-screen engine fact). - **E2 (accuracy W1):** "measured values" ×4 in L10 → "each substance's own value"; the S09 ROOT (the naming line's "measured at 25 °C") → "tabulated at 25 °C" — a COORDINATOR one-word fix applied after the G2-residual executor releases U6-S09.json (register rev 2.2 (c) carries it; the S09 mini re-read covers the word). - **E3 (accuracy W2):** "…into one the table never lists." - **E4 (accuracy W3):** C(s) → C(graphite) in S10's formation contexts (rev 2.2 (d); the James card keeps the one-word alternative). - **E5–E7 (questions W1–W3):** the kJ/mol_rxn options; the continuity stem; the two fb-trim sentences on the both-tables verdict item. **E8 (W4):** the −196.0 echo flag added to SPEC_RECORD's key-echo list. **E9 (accuracy watch):** the L01 iodine item's stem gains "in that order". - Register rev 2.2 COORDINATOR-EXECUTED (validator PASS). Fix agent: E1, E3–E9 (E2's S10 half; its S09 root is the coordinator's); re-solve changed items; sweeps; bundle cut (first S10 bundle — hessstack figures resolve via the bundler? CHECK: hessstack is a NEW kind — the bundler's strip() path must not leak manipulation records or sums; verify and STOP if it would, the §E-S06/§E-G3 precedent) + bare-ids CHANGED_IDS; then S10's closers.
§E-G4
The call: G4 closed: three wording rows — the particle-level distractor dressing, the kinetic-energy ownership, the 5-mol stem.
What changes: A fresh cold re-read of the three items closed the gate.
## §E-G4 — U6-G4 fix round (adjudicated 2026-10-08; blind 6/6, QA cold 6/6, 0 Essential, 3 W) W1–W3 ADOPTED as written in qa/REVIEW_GATES_G4.md (the particle-level distractor dressing on THC-064; the THC-066 kinetic-energy ownership; the "carries out 5 mol of reaction" stem). Leaves stand (the U6-P2 concrete-direction note joins the practice builder's constraints — relay if practice is in flight). Fix agent: W1–W3, re-solve the three items, bundle re-cut leak-free, fresh cold re-read of the three closes U6-G4. ### §E-S09b G2-residual, second ruling (coordinator, 2026-10-08) As-ruled lands at 433 px vs the 408 wrapper (25 px licensed scroll remains at 480; 760 fully fixed). ADOPTED: the in-cell licence extends to breaks before "+" — the SAME template law eqstack's U5 canvas rule already applies (breaks before + / → / ⇌), measured to 382 px (fits with 26 px spare). The C(s) alternative declined (collides with the diamond-vs- graphite teaching point; the James card keeps it). Mutation case owed; re-open at 480 both themes; the S09 mini re-read then covers the final render. ### §E-S10/§E-G4 close notes (coordinator, 2026-10-08) - The fix agent's ONE beyond-verbatim surface (the reason-item fb(d) recast into one compared frame, forced by the ordered check_parallel step) is RATIFIED — the §E-S06b precedent; fb-only, properly flagged. - The 46× C(s)→C(graphite) breadth (beyond the "formation contexts" letter) is RATIFIED on the agent's one-form bookkeeping argument (the stacks force step and target into one form); S07's 3 reaction-context C(s) correctly left (rev 2.2 (d)'s scope). **U6-G4 CLOSED** (REREAD_G4_CHANGED 3/3 on disk). Gates: G1–G4 ✅ · G5 queues on S10 closure. - The bundler gained the hessstack item-stack REFUSE defence + "op" in the leak set — coordinator notes the pattern now covers all three resolved families. ### §E-S09b count correction (coordinator, 2026-10-08) The re-read's count note accepted: the F1–F6 set is TEN surfaces, not the records' "eleven" headline (off-by-one, nothing unapplied). Corrected here rather than editing closed records. **U6-S09 GAUNTLET CLOSED** (re-read all-PASS incl. the 382/408 table, the hook, and the tabulated root word).
§E-PRAC
The call: The practice round: P1 stopped answering (a) before asking it, and the rubric-literacy form came back (a point's TYPE, never its content).
What changes: P2's ship-gate stays your CaCO₃ + HCl card above.
## §E-PRAC — practice fix round (adjudicated 2026-10-08; QA 20/20 cold + blind 20/20; P2 FAITHFUL; 2 Essential / 4 W + 3 blind flags)
- **E1 ADOPTED:** P1's (c)/(d) neutral rewrites ("magnitude of q for the solution"; the
equal-|q| claim recast) — the page no longer answers (a) before asking it (the blind's
flag 1 is the same class, satisfied by these rewrites).
- **E2 ADOPTED:** the two currency-frame idea lines recast to the U5 rubric-literacy form
(the point's TYPE, never its content) — restores the THC-041 asked-never-printed design.
- **W1 ADOPTED:** "surface area" → "particle size" in the P2 frame. **W2 ADOPTED:** one
declaration line naming the three engine-shape moves; PLUS the blind's flag 3 — the
second P2 item's stem restates the 1.00 M and the 100.09 g/mol molar mass (cross-item
data the one-screen engine hides). **W3 ADOPTED:** the ledger prose trimmed off the
student page (the U5 source-line form). **W4 ADOPTED:** the 23.97 → 24.0 display step.
- Blind flag 2 (bundler part labels) ALREADY EXECUTED by the coordinator; bundles re-cut.
- Leaves stand; P2's ship-gate stays the CaCO₃+HCl exemption on James's card.
- Fix agent: E1/E2 + W1–W4 + the stem restatements; re-solve changed parts; sweeps;
clickthrough rubric drives; bundles re-cut leak-free + bare-ids CHANGED_IDS; a fresh
reader re-reads the changed parts to close the practice gauntlet.§E-S10b
The call: S10's close: five blind rows adopted (the dry-ice clause, the trimmed mechanism key, the kit wording, formulae in the alt).
What changes: The LAST sequence gauntlet closed on its mini re-read.
## §E-S10b — S10 closing application round (adjudicated 2026-10-08; blind 58/58 + frq fair-miss, figures 0E/1W, lesson 0E/5W with 14/14 ×2) - **F1–F5 ADOPTED** as written in qa/REVIEW_LESSON_S10.md §4 (the equation-flips agency class with its doer restored; the solids-rule fb mirror; "is chosen for" with the James boundary note; "add up to"). - **G1 (figures W) ADOPTED:** the L06 "below" → "shown" stem word. - **Blind flags:** B1 (L09-49 stating sibling answers) — LEAVE, the third independent eye on the twice-ruled scaffold-ladder design (one-check-per-screen + H1 licence); the convergence goes on James's card beside L01-04's. B2 ADOPTED — the dry-ice fb gains the liquid-CO₂ defusing clause (fb-only). B3 ADOPTED — L10-58's mechanism-bearing longest key trimmed per the blind record. B4 ADOPTED — L10-56's kit wording says the steps come WITH their ΔH values. B5 ADOPTED — L06-32's alt speaks the formulae its options use. - Housekeeping: the stale pre-fix _s10_family/_zoom sheets regenerated or deleted. - Application agent: all rows; sweeps; clickthrough; bundle re-cut (B3/B4 touch option/stem text) + bare-ids CHANGED_IDS; a small fresh re-read of L10-56, L10-58, the dry-ice fb and the L06 stem closes S10 — and with it the LAST sequence gauntlet. ### §E-S10b close (coordinator, 2026-10-09) The re-read's two fb-only faults fixed by the coordinator with the reader's exact strings (the −49.2 why restored; the mirror defusing clause into the No fb). 4/4 cold stands. **U6-S10 GAUNTLET CLOSED — all ten sequences closed; validator 0 errors course-wide.**
§E-G5
The call: G5's round: the drive harness's silent no-match caught — three wrong-map paths were undriven while the record claimed 65/65.
What changes: Fail-loudly drive harnesses are now vault law (landed @ e7af59d).
## §E-G5 — U6-G5 fix round (adjudicated 2026-10-09; blind 14/14, QA cold 14/14, 0 Essential, 3 QA-W + 2 blind flags; all twelve value flags confirmed incl. the −46.0 one-dataset ruling) - **W1 ADOPTED (harness + record):** the drive harness's "toluene" fragment never matched the C₇H₈ stem — three wrong-map paths silently undriven while the record claimed 65/65. Fix: fragment → "C₇H₈", a no-match-is-a-FAILURE guard added, the battery re-run, the record line corrected ON the record (never silently). The silent-no-match class is a VAULT-SYNC candidate (drive harnesses must fail loudly on zero matches). - **W2 ADOPTED:** the hydrazine scene → the fuel-chemist frame (real thrusters never burn N₂H₄ with O₂); numbers unchanged. - **W3 ADOPTED:** the fb(c) dangling clause → the record's rewrite. - **Blind flags ADOPTED:** item 12's key trimmed / distractor weighted to kill the long-key tell, AND the stem gains "with the water a gas in both" (the exclusion self-carrying). - Leaves stand (the −46.0 call CONFIRMED as the rev 1.9 discipline working; G5 joins the close-out 480-px sweep note). Fix agent: W1–W3 + the two blind rows; re-solve changed items; bundle re-cut; a fresh re-read of items 7, 9, 12 closes U6-G5.
§E-T1
The call: The test round: Q18's table recast to MIXED data — killing a format leak and strengthening the sufficiency demand in one fix.
What changes: The test-page label cosmetic went to close-out (your card).
## §E-T1 — U6-T1 fix round (adjudicated 2026-10-09; examiner 20/20 purity+fidelity PASS 0E/4W, blind 18/18 + 4/4 + 10/10 with 3 flags) - **E1 (blind F1 + examiner W2, one fix):** Q18's table recast to MIXED data where only the ΔHf column suffices (kills the Q16 format-leak structurally AND strengthens the sufficiency demand) + Q18's alt neutralised per the examiner's line. Verdict option set unchanged. - **E2 (examiner W1):** Q14's stem drops "average" (the word-match crown; the James note on the S08 lane precedent kept). - **E3 (examiner W3):** Q07's fb(d) trim (it pre-states FRQ-1(d)'s award sentence). - **E4 (blind F3):** Q04 — one distractor dressed with a parallel qualifier shape (the G4 W1 precedent); the key's "on average" is the mechanism's own qualifier and stays. - **E5 (blind F2) RULED LEAVE + James card:** the thermometer alt's drawn-position description is the figure's ACCESSIBLE DATA CHANNEL — it describes exactly what the drawing shows (the read is earned by converting mark-position to a value, the item's own skill); dropping it would blind a screen-reader student entirely. The divergence (blind flag vs examiner measurement) is recorded; it joins the standing engine-level figure-read-alt card from U5/S01. The frqstim refusal's literal-grep limitation noted on the same card. - **E6 (examiner W4) DEFERRED to close-out:** the test-page lesson-card labels are a build_course.py template branch shared with U5-T1 — a cross-unit cosmetic; on the close-out list + James card, not a this-round spec edit. - Fix agent: E1–E4; re-solve changed items; KEY TELLS re-run; bundle re-cut + bare-ids CHANGED_IDS; a fresh re-read of Q04/Q07(fb)/Q14/Q18 closes U6-T1. ### §E-T1 close note (coordinator, 2026-10-09) The fix agent's one declared beyond-verbatim surface (the Q18 alt spelling "delta H f" — the engine's own no-raw-token alt assert caught the examiner's literal) is RATIFIED; the §E-S06b flag-don't-improvise precedent held again. ### §E-T1 final close (coordinator, 2026-10-09) The re-read's one fault (Q07 key-fb's portable inverse law — the FRQ-1(d) point in quotable form) fixed by the coordinator with the reader's exact data-first rewrite; the anaphora defence considered and REJECTED with the reader (the sentence quotes out of context as a complete law). grep 0 on the page; build/validator/mutation green. **U6-T1 CLOSED.**
§E-P3
The call: P3 closed: one rubric line named the wrong table on the exact point under test — fixed with the record's string.
What changes: The practice set P1–P3 is complete.
## §E-P3 — P3 fix round (adjudicated + coordinator-executed 2026-10-09; QA cold 10/10 both flags confirmed, blind 10/10; 1E/5W/4L + 2 blind notes) - **E1 + W1–W3 + W5 COORDINATOR-EXECUTED** with the records' exact strings (the rubric's "those averages"; the "are:" fragment; "justify your choice"; the manipulation-shape de-leak; "labeled"). Build/validator/mutation/bundle/checker all green after. - **W4 DECLINED:** the ½-glyph vs n/m-slash pair is the hessstack family's ONE convention, held course-wide — the S10 figures pass's standing Leave explicitly bars fixing one instance (the two reviewers' conflict resolved FOR the standing ruling). - **Blind notes:** the "vapour" flag DECLINED (US student-facing convention); the kJ/mol_rxn prompt-alignment near-miss on James's card (rubrics are route/unit-agnostic and score cleanly); the serve-together constraint confirmed declared. - The changed surfaces are rubric/prompt/idea/alt wording only (no keys, no data); the two 10/10 cold reads stand. **U6-P3 CLOSED — the practice set (P1/P2/P3) is complete.**
One verdict card each: what it holds, and the five QA records' headline results, parsed from the records themselves.
Your criteria (as you pasted them, 7 Oct 2026 — the doc): lessons are good enough if they are 8/10 on these.
Gut checks: Are we confident that this is better than the alternative? Are we confident that a student would benefit from taking this lesson?
The doc's tab is the physics one; its accuracy line reads as chemistry accuracy here. Wording otherwise verbatim.
Sequence S01 of 10
13 lessons · 51 checks (49 MCQ + 2 numeric) · 28 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S02 of 10
5 lessons · 24 checks (19 MCQ + 4 numeric + 1 draw) · 36 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S03 of 10
6 lessons · 32 checks (32 MCQ) · 8 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S04 of 10
15 lessons · 62 checks (47 MCQ + 15 numeric) · 27 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S05 of 10
7 lessons · 54 checks (45 MCQ + 9 numeric) · 16 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S06 of 10
14 lessons · 77 checks (59 MCQ + 17 numeric + 1 draw) · 25 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S07 of 10
9 lessons · 55 checks (43 MCQ + 12 numeric) · 18 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S08 of 10
8 lessons · 41 checks (31 MCQ + 10 numeric) · 31 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S09 of 10
7 lessons · 52 checks (43 MCQ + 9 numeric) · 10 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Sequence S10 of 10
10 lessons · 58 checks (31 MCQ + 27 numeric) · 32 figures
Read it as a student on the course page.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Gate U6-G1
8 items (8 MCQ; MCQs four-option) · 2 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Gate U6-G2
7 items (5 MCQ + 2 numeric; MCQs four-option) · 1 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Gate U6-G3
6 items (4 MCQ + 2 numeric; MCQs four-option) · 1 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Gate U6-G4
6 items (4 MCQ + 2 numeric; MCQs four-option) · 0 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Gate U6-G5
14 items (6 MCQ + 8 numeric; MCQs four-option) · 3 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Practice U6-P1
2 FRQs, 10 rubric points · 2 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Practice U6-P2
2 FRQs, 10 rubric points · 2 figures
Holds for your CaCO₃ + HCl card above — P2 ships only past it.
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Practice U6-P3
3 FRQs, 10 rubric points · 3 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Unit test U6-T1
18 MCQ + 2 FRQs (14 points) · 10 figures
Judge it on your criteria (the 8/10 bar):
The full box sits above the verdict cards.
Everything you tapped or typed on the review tab gathers here, in page order. It saves in this browser as you go; sending twice is safe — later notes supersede earlier ones.
Everything lands under APCHEMU6 with your name on it. If a send fails, the note is kept in this browser and the page retries by itself.