Unit 9 — the course's final unit — is built end to end and QA'd: 12 sequences · 107 lessons · 4 cold gates · a practice page (18 items, carrying 8 points of real released FRQ, text-verbatim: 2021 Q5 and 2025 Q6) · the test (18 MCQs + the whole released 2024 Q3 verbatim, 10 points, + an authored 4-point FRQ). 546 driven checks, all green · 18 sealed blind reads, all perfect · validator 0 errors · register at rev 1.8.
One decision leads everything else: the S08A veto is PENDING. A whole sequence (solution redox balancing) was inserted on 7 October without your ruling. The first card names every surface a veto would strike, and the one released exam point that sits on the same ground.
Do one thing: rule on the 16 cards below (the headline carries one sub-choice) — one tap each; every tap sends straight to the inbox. Nothing blocks on them: each card runs on its recorded default until you rule.
The whole-unit verification's verdict: PASS — zero content findings (qa/WHOLE_UNIT_VERIFICATION_U9.md, verdict section). Its four findings are record-wording grain; a hygiene seat is rewording two record files in parallel and touches no student surface.
This page IS the reviewer view, on by default: QA numbers, records and build calls all visible. Students never see any of it. The course page shows it their way, lesson by lesson. Records live in unit9_build/qa/; rulings in unit9_build/blueprints/ADJUDICATIONS.md and BUILD_STATE.md.
Every number below is from qa/WHOLE_UNIT_VERIFICATION_U9.md — a fresh seat re-ran the whole battery on the final state and every recorded number reproduced exactly.
Your review criteria, verbatim (your doc) — lessons are good enough if they are 8/10 on these. A compact copy sits beside every note box.
Is this content better than the current alternative students have?
The headline first, then the stack grouped by family. Every card names its default — the build already runs on it. Each card's small header carries its queue number and its §-cite.
Card 1/16 · HEADLINE · §E-G3 · §E-P1 · §E-T1.2 · verification §7 (the veto-surface list, verified complete)
veto pendingregisterWhat happened: five atoms (TEC-071A–E, balancing half-reactions in acidic and basic solution) were operator-inserted on 2026-10-07. You have not ruled. The sequence was built and QA'd like every other (blind 28/28), but the closure does NOT discharge the veto.
A veto strikes exactly four authored surfaces, all cleanly (each strike argued in ADJUDICATIONS §E-G3, §E-P1 and §E-T1; the register's rev-1.4 one-tap revert stands):
S08A itself — its five lessons (TEC-071A–E)
Gate G3 — items 5 and 6 (TEC-071C/D)
Practice — item U9-P1-15 (the acidic NO₃⁻ → NO balancing MCQ)
Test — item U9-T1-13 (the AsO₂⁻ basic-solution two-flaw check)
One released point sits on the same ground: 2024 Q3(d)(i) — a net-ionic point inside the test's verbatim released FRQ. A veto does not touch it automatically; the next card holds that choice, and it is yours alone.
My recommendation: Approve — the exam itself asks this ground (2024 Q3(d)(i)), and every S08A surface passed the full battery.
Card 1b · rides card 1 · §E-T1.2 ("James's call, not pre-decided") · TEST_RECORD_U9
questionsThe situation: the test carries the real 2024 Q3 verbatim (10 points). Its part (d)(i) asks for a net-ionic equation — the ground S08A teaches. If S08A stays, nothing arises. If you veto, one released point assesses something the course no longer teaches.
The two honest options, laid out, not pre-decided:
Keep 2024 Q3 whole — exam fidelity wins; students meet one point cold, as a real sitting would
Swap 2024 Q3 for another released FRQ — the test assesses only taught ground; the verbatim-whole 2024 Q3 is given up
No recommendation on this one — it is yours. (If you approve S08A on card 1, tap "Not needed".)
Card 2/16 · ADJUDICATIONS §B6 · §E queue 2 · the U6 THC-078 seam · your standing U6 card
wordingstandingThe course says: every standard quantity in U9 wears its mark (ΔG°, ΔS°, E° — 2 965 degree marks, zero wrong glyphs). TEC-011 teaches what the mark means: it pins the states, not the temperature.
The seam: built U6 ships ΔH_f bare, and its THC-078 folded 25 °C into "standard conditions" — the lineage that produced one U9 item teaching the uncorrected claim (caught and fixed under warrant; the S02 strings now carry the convention honestly). U9 recaps cite the routine, never U6's typography.
Cost: confirming U9 costs nothing — it is built. Re-opening U6's wording is a separate maintenance round, one tap here when you rule.
My recommendation: Confirm U9's grain; leave U6 on your standing card until a U6 maintenance round exists.
Card 3/16 · 05_operator_adoption D6 (#46) · §E queue 3 · D6(adj)
registerThe course says: TEC-040 treats the temperature side qualitatively; the equation lane runs TEC-041 → TEC-042/043. No "higher T pulls K toward 1" atom exists and none was smuggled in — K's own temperature dependence stays Unit 7 territory (EQM-078/079), retrieved, never extended.
The new evidence for your parked decision: the build met the boundary for real — EQM-078's van-'t-Hoff ground was partially discharged exactly as parked, and TEC-041 pays U7's IOU without any new atom. The parked choice never bit.
My recommendation: Keep D6 as confirmed — no K(T) atom; one tap adds one if you read the ruling differently.
Card 4/16 · ADJUDICATIONS §S03-1 · §S05-1 · §E queue 4–5 · register rev 1.7, wording-only
registerTEC-027: the bracket's "always-favored" example was CH₄ combustion — which fails its own sign pair at literature ΔS° values. The bracket now names H₂O₂ decomposition instead; CH₄ stays TEC-019's formation-sum ground. Wording-only; one-tap revert.
TEC-044: the bracket's rounded 4 × 10⁻³¹ now reads 4.5 × 10⁻³¹ — the U7 pin (one course, one value). Same class, rides this card.
Cost of reverting: TEC-027's old bracket would teach on an example that contradicts the data; TEC-044's would print two values for one constant.
My recommendation: Ratify both.
Card 5/16 · ADJUDICATIONS §B8 · §E queue 6 · register rev 1.7
questionsThe course says: misconception claim-items run only on a registered, sourced belief. Rev 1.7 registers twelve entries, every one sourced to the chemistry-education catalogue; multi-site beliefs register once with licensed sites. The shipped census: 7 Yes / 14 No, no guaranteed-No shape, and the absolute-quantifier tell is dead (verified item by item).
My recommendation: Confirm all twelve.
Card 6/16 · ADJUDICATIONS §B5 · §E queue 7 · §E-S03-pre (the asset-1 refusal) · §E-S11-pre (asset 5 undrawn)
figuresThe course says: each photo carries a felt scene and no data (your 2 Oct rule); everything exact stays a drawn figure. All shipped renders are independently QC'd and file-referenced.
One grant broke honestly: asset 1 was ONE ice/water photo serving two lessons — S01's melting opener AND S03 L09's freezing scene. A melting render cannot caption a freezing pond, so L09 stays prose. The interim rule stands: no second take without you.
My recommendation: Keep the two shipped, leave the three undrawn as grants on file — any of them can be drawn later with one tap, or struck.
Card 7/16 · ADJUDICATIONS §E queue 8 · PRACTICE_RECORD §15.5 · still pending from the U9 blueprint pass
questionsstandingThe question: several released questions are only partly on U9 ground. Part-level consumption would unlock those parts; whole-question-only keeps the ledger simple and every sitting authentic.
What U9 did while you were deciding: consumed only whole reserved questions and drew no split-use remainder. Nothing is waiting on this except future builds' headroom.
My recommendation: Allow part-level consumption, ledgered per part — the remainders are real assessment material no whole-question rule can ever use.
Card 8/16 · §E-P1.2 · PRACTICE_RECORD §15.1 · the CARD-ASSESS-2 class
questionsThe course says: the engine has no free-drawing surface, so Q5(a)'s draw task converts to its stated equivalent, with the released wording quoted inside the prompt ("…asks you to draw an arrow on the diagram showing…"). Same atom, same point value; the other seven released points ride untouched. The test needed no such conversion (2024 Q3 has zero drawing acts).
Cost of rejecting: the whole 2021 Q5 (4 points) comes off the practice page — no draw surface exists to honour the original form.
My recommendation: Ratify the conversion.
Card 9/16 · §E-P1b.1 (the corrected true state) · PRACTICE_RECORD §15.2
questionsCourse surfaces (U9-S09): Mn²⁺/Mn = −1.18 V — printed fourteen times, including one keyed item (TEC-077)
Released material (2025 Q6's own table, on the practice page): −1.19 V
Why it stands: released text is never edited, and the course's value is its own sourced pin. No surface co-prints both, and no point or key anywhere turns on the 0.01 V.
Honesty note: an earlier record claimed −1.18 was "unprinted" — that was false, caught in QA, and corrected before this page was built (§E-P1b.1). What you see above is the verified state.
My recommendation: Accept the cross-page channel — it is how released tables always ride.
Card 10/16 · PRACTICE_RECORD §15.3 · flags P-2/P-3, both verified
figuresThe course says: the released booklets' cell ART for practice Q5 and Q6 could not ride verbatim (no image rights lane), so both cells are redrawn in the course's own cell family. The text of both questions is verbatim. Both reconstructions were accuracy-verified against the released drawings (flags P-2/P-3), and Q6's computed meter face reads exactly the released print: +0.90 V.
The one-tap alternative: render nothing and describe the cell in words instead — fidelity to "no reconstructed art", at the cost of making students read a cell they would see drawn on the real paper.
My recommendation: Keep the redraws.
Card 11/16 · §E-S01b.6b · §E-S02b.5 · the U8 label-floor card's sibling
figuresstandingEnergy-curve labels at phone width (the mb family, 480 px): ≈10.2 px — readable in both colour schemes on its own renders
Figure captions (house page CSS, every figure): 12.8 px
The in-figure floor the engine enforces elsewhere: 13 px
Why nothing was changed: both members pass their own legibility reads; forcing 13 px re-renders a closed family and re-skins the house CSS — a family ruling, so it filed to your card instead of code (the U8 titration-curve precedent).
My recommendation: Per-family norms, each measured on its own renders — the same call the U8 card recommends.
Card 12/16 · §E-G2.1 (the seam ruled lawful)
wordingThe course says: a U9 gate item and a built U7 surface print the same +88 quantity from one lineage at two precision grains. They never co-print on any page, so the QA seam was ruled lawful and the gate shipped.
The one-value option: if you want strict one-course-one-value here, the fix lands on U7-S10, never the gate — a one-value edit on a closed unit's page.
My recommendation: Leave it — the grains never meet, and nothing keys on the difference.
Card 13/16 · §E-S09 addendum (residue row, queued as your footnote)
wordingThe course says: S09 L06's item 25 (the tripled ΔG° > 0 case) keeps "energy released" in its rubric's primary wording and model line 2. The accept-clause already carries the accurate form, so every honest student answer scores.
Why it was left: the line surfaced after the fix warrant closed; changing a closed rubric for a wording students never fail on did not justify a new round.
My recommendation: Queue "energy change" as primary for any future S09 touch — no round now.
Card 14/16 · §E-S11b.5 (W18, escalated — a coverage decision, not a warrant)
registerThe register (TEC-100): mass deposited on or removed from an electrode
The sequence (S11): teaches and assesses the deposition direction only
Why it reached you: a reviewer flagged the gap; narrowing a register statement or adding taught coverage is a register decision — yours, not a seat's.
The two honest fixes: amend TEC-100's bracket to deposition (one wording edit), or add the dissolution direction to S11 (new teaching + checks, a re-open).
My recommendation: Amend the bracket to deposition — the CED's electrolysis calculations run on deposition, and no U9 item needs the reverse.
Card 15/16 · §E-T1b.4 · REVIEW_TEST_U9 row 2 · executed
done under warrantwordingThe deleted sentence: "That is carbon disulfide's normal boiling point: above 320 K, vaporization at 1 atm is favored."
The part's ask: "Calculate the temperature…" — the model now stops at the calculated value
Why it went: the sentence is chemically true (the calculated 320.2 K matches the handbook's 319.4 K — that match is on the record), but it answers a question the part never asked. On a TEST surface, exam fidelity outranks the flourish.
My recommendation: Leave it deleted; the matches-the-world note lives in the record.
Card 16/16 · standing (your U7→U8 queue) · vault candidate ii · the close-out vault-sync note
standingwordingU8's wrong-option frame (your open card): "This option is wrong — you …"
U9's right-answer frame (the same contract family): every correct option's feedback opens "Correct."
The vault's explanation guidance: feedback explains; it does not open on a verdict formula
What happened in U9: the conflict surfaced in review; the contract won in-course (shipped units stay one voice) and the vault holds the conflict on your existing card, deliberately unresolved — never silently settled by a seat.
Cost of reframing: a mid-course voice change desyncs nine shipped units; the retrofit is a course-wide re-open.
My recommendation: Hold both house frames; your one tap settles the family course-wide either way (the vault records the exemption or the retrofit).
Applied without asking (rulings you have already given, or calls that would not change what a student learns): the two record files whose prose failed the record checker are being reworded checker-clean in parallel under the precedent you set at §E-S07b.4 — record wording only, zero student surfaces · the blind folder's scratch directories are deleted at the same hygiene seat · the two proposed engine guard arms ship to the NEXT course through the vault, not here (every U9 bundle is already cut, so they could never fire) · the Incept question pack was again not used for the test (write-mcqs owns question work — your U7 ruling) · the one-process-one-owner instance standard ran as operator ground throughout.
Whole unit · 12 sequences · 4 gates · practice · test · the course's final unit
107 lessons · 546 driven checks all green · 166 figures · 18 perfect sealed blind reads · zero content findings at verification. Open the course page in the next tab and dip in anywhere.
Judge it on your criteria (the 8/10 bar):
The full box sits above the 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 APCHEMU9 with your name on it. If a send fails, the note is kept in this browser, a banner shows it ("Not sent"), and the page retries by itself. Collected notes get themed into the report before any lesson changes are made (your 7 Oct rule).