AP Chemistry Unit 7 — Equilibrium — your whole-unit review

Keep pH out of Unit 7's Le Châtelier observables?

The course says: stress → observable items use colour, temperature and amount of solid; pH-based items enter Unit 8 as retrieval once pH is defined.

Why: the CED names pH only as a "such as" example, and the EK stays covered by the other observables. No honest one-line pH gloss exists that does not smuggle Unit 8 content (your jargon-before-use rule).

Cost of changing it: Unit 8 content enters Unit 7 early; the dedup rule breaks.

My recommendation: Confirm the deferral.

Keep the inert-gas atom assessed, beyond the CED's literal stress list?

The course says: "adding an inert gas shifts the equilibrium" is taught AND assessed (it sits in gate G4's cold sample) — though the CED's own stress list omits inert-gas addition.

Why: it is a standard AP distractor for the compression heuristic, and your house bar tests every heuristic on the case where students think it switches off.

Same class, same card: S09 also carries one carbon-gasification check keyed no shift for adding pure solid — another beyond-the-literal-list case.

My recommendation: Keep it assessed.

Heat-as-reactant survives only as an as-if datum — keep it, or go K-only?

The course says: "treat heat AS IF it were a reactant/product when predicting the shift" — a stated bookkeeping device, prose and table only. Heat never acts, never travels, and no "+ heat" appears inside any chemical equation anywhere in the unit (your October heat rules bind every surface; the sweep enforces at build time).

Why: the device is the standard AP shortcut; the real story (temperature changes K itself) follows two lessons later, and the thermodynamic why is a documented Unit 9 deferral.

My recommendation: Keep the as-if device.

License the word "catalyst" at one lesson only?

The course says: the build engine bans "catalyst" unit-wide — except at the K-depends-only-on-temperature lesson (EQM-031), in the narrow does-not-change-K sense.

Why: "which change alters K" items list catalyst as the canonical no on the real exam (your 2026-08-06 bias-to-the-exam directive); the word is Unit 5 retrieval, not a new coin. Rate-law machinery stays banned everywhere.

My recommendation: Keep the licence.

Confirm the six registered misconceptions?

The course says: misconception claim-items run only on a registered, sourced misconception. Six were registered for Unit 7:

"equilibrium means the reaction went to completion" (EQM-006)
"at equilibrium the reactions stop" (EQM-008)
"equal rates means equal concentrations" (EQM-016)
"a large K means a fast reaction" (EQM-041)
"adding an inert gas shifts the equilibrium" (EQM-070)
"insoluble means nothing dissolves" (EQM-091)

Why: each is sourced to the standard chem-ed catalogue or the register's own evidence; claim items use reported speech, bare Yes/No; assessment pages never reuse the guaranteed-No shape (your claim-salt rule rode every gate and the test: 12 No / 4 Yes).

My recommendation: Confirm all six.

Coin the symbol as "the double arrow" first, full name three lessons later?

The course says: lesson 2 coins ⇌ as "the double arrow"; the full name "equilibrium double arrow" waits until "equilibrium" itself is coined (lesson 5).

Why: using the full name at lesson 2 would breach the unit's own forward fence — the word "equilibrium" would appear before it is taught.

My recommendation: Stands.

FYI: a register example value was wrong and is fixed

The course says: the register's bracketed example K for 2H₂+O₂⇌2H₂O at 25 °C read ≈10³⁴. Two independent recomputations (ΔG°f route) give ≈10⁵⁰. The register now reads 10⁵⁰.

Why it matters: register hygiene only — no lesson printed the old value.

The small-x validity rule lives inside one lesson — keep, or split it out?

The course says: EQM-090 (when the small-x shortcut is allowed) is taught, judged on cases AND checked inside S11's L11 — one lesson.

Why: your 7 Oct rule wants a judged rule to get its own example-sequence lesson; L11 IS that lesson (the validity condition runs its own positive-negative-positive case run). A split would need a new atom — your suffix insertion.

My recommendation: Keep the one lesson.

FYI: the H₂+I₂ workhorse value is now Kc 49 at 730 K

The course says: Kc 49 at 730 K, sourced — superseding the overtaken pilot spec's 50 at 700 K. One course, one value; every surface re-pinned and grep-proven.

FYI: CaCO₃ teaches here but is never a check; your U6 ship-gate is untouched

The course says: CaCO₃ decomposition is a Unit 7 teaching thread (the Kp = P_CO₂ worked lane) and is banned from every Unit 7 check. Your open Unit 6 ship-gate concerns CaCO₃ + HCl — a different process — and nothing here touches it.

PCl₅ appears in two sequences' items — keep the declared share, or force one owner?

The course says: one S01 check and one S07 lane both use PCl₅ decomposition — a declared share (different atoms, the Unit 6 precedent). The instance ledger carries a one-tap alternative: strict one-owner (S07 moves to SbCl₅).

Cost of flipping: one sealed S07 lane re-instances and re-reads.

My recommendation: The share stands.

Twelve photo-real hooks shipped — keep all, or strike any back to drawn figures?

The course says: each photo carries a felt scene and no data; everything exact stays a drawn figure (your 2 Oct decision). All twelve are QC'd, independently reviewed on the rendered page, and file-referenced (never embedded).

Notes: the brown-NO₂ tube is ONE shared file reused across four sequences (one depiction per thing). The barium-meal scene rendered only after the subject-expert confirmed the medical safety line (a barium meal is safe because BaSO₂ barely dissolves — 1.44 mg of Ba²⁺ per litre against grams-scale toxicity).

My recommendation: Keep all twelve.

Your open card: should sharpness alone fail a photo?

The course says: the photo-QC seat sometimes rejects takes purely on SHARP_NO_DOF (everything must be in focus). We now use a depth-composed prompt pattern instead of burning takes — carried from Unit 6, still open on your list.

The live example: the S06 city-haze photo passed QC with a structural sharpness residual — the subject IS a haze.

My recommendation: Relax it — a structural residual (haze, steam, depth) passes when the teaching subject is sharp.

Three small-phone rendering notes on the solubility figure families — accept, or re-render?

The course says: three family-level notes from the figure reviewers, none item-breaking, all measured at 480 px:

kspscale: tick-label font renders small at 480 px (readable, below house preference)
satbeaker: the solid lump draws at one size — lump size is not a data channel; prose carries the scale
satbeaker pairs: the panel pair-tags render ≈7 px at 480 px

Why accept: no item's answer depends on any of these; both blind reads ran clean on the drawn facts.

My recommendation: Accept the notes; fold any change into the next natural family touch.

Build a computed-residue beaker variant for the S12 ion-count panels?

The course says: the S12 author offered a saturated-beaker variant whose drawn solid residue is computed from the numbers rather than drawn by convention. The shipped panels are correct and sealed.

Cost of building it: a new figure-family member with its own refusal battery; the sealed panels re-render and re-read.

My recommendation: Leave as shipped.

Thousands separator on figure axes: your new SPACE rule, or the engine's COMMA law?

The two rules, side by side:

Your vault rule (9 Oct): 1 000 — thousands get a space
The engine law (U5, every U5–U7 figure): 1,000 — thousands get a comma
Recommended: keep the comma on figures until you rule at vault level

Why: the comma is baked into the shared tick-formatter; switching re-renders three units' figures. Cross-course and engine-level, so it is yours, not the operator's.

Related, already ruled on record: kelvin temperatures print solid ("730 K") — figure-text agreement outranks (the AF15 ruling).

Write acetic acid as CH₃COOH course-wide?

The two forms:

Shipped at one S01 instance: HC₂H₃O₂
Recommended course-wide: CH₃COOH, with one equivalence line at that instance

Why it is your call: the G1 reviewers recommended CH₃COOH (the structural form students meet everywhere else), but the change touches a shipped, sealed page.

My recommendation: Adopt CH₃COOH course-wide.

Add the one-line naming bridge for "particulate model" vs "particulate representation"?

The collision: an earlier unit coined particulate model; the AP exam says particulate representation; S08 teaches under the exam's term. Under your 9 Oct term rules this is a genuine rule collision (earlier coin stands vs the exam's term) — your call, not a defect.

The proposed line, verbatim (inserted in S08 L01 after the definition):

"We called this kind of drawing a particulate model in an earlier unit — a particulate representation is the same drawing, under the name the AP exam uses."

My recommendation: Add the bridge line.

Add a one-line isomer gloss in S05?

The course says: S05's near-1 example is butane ⇌ isobutane (K ≈ 2.5) with no gloss on "isomer". Optional — neither cold reader needed it to answer.

My recommendation: Leave it out.

Drop the name "water-gas shift" before S09 coins "shift"?

The course says: the reaction appears early (S03, S06) as its formula only; the industrial name "water-gas shift" is used freely once S09 has coined "shift".

Why: it is a proper reaction name, not the S09 concept — but keeping pre-coin surfaces clean was cheap, so we did.

My recommendation: Keep the drop. One tap restores the name everywhere.

A mis-attributed phrase was swapped — veto open

The two lines:

Was: "at the plant's operating temperature" (S06 L04, three surfaces)
Now: "at the temperature of the table's study"

Why: the subject-expert showed the quoted temperature is not the sulfur plant's operating temperature — the attribution was wrong; the data and keys never moved.

FYI: twice, the generated lexicon outranked a reviewer's exact wording

The course says: "holds" became "stays" (the agency lexicon errors "holds" as a quantity acting) and one adjudicated string shipped as "the relationship between" (the hide-link lexicon). Both one-word, both on the record, meaning unchanged.

"Ester flask" names an uncoined word — leave, or take the one-tap swap?

The course says: S02's q14→q15 pair says "ester flask". "Ester" is never coined — but the string is an adjudicated exact rewrite (it killed a banned catalyst token), the word is a flask label rather than a tested term, and both cold readers answered without it.

The one-tap swap on record: "the acetic-acid flask" in q15's stem and "product and water" in its feedback.

My recommendation: Leave it.

S12 L02's quiz keeps a letter run because the example must come first — agree?

The course says: the quiz's answer letters run same-letter for a stretch. The only swap that breaks the run would open the quiz on the temperature NON-example — and example-before-non-example outranks a letter pattern.

Why it is safe: two sealed blind reads answered every item without using the run as a tell.

My recommendation: Leave the order.

The standing feedback-opener tension: house form vs vault rule

The two forms:

Every shipped U5–U7 surface: "Correct." / "This option is wrong — you …"
The vault rule (R156): no commentary opener — go straight to the reason

Why it rides: the course contract followed its own form for consistency; a change now would desync every shipped gate and sequence across three units. Cross-course, so it is yours.

My recommendation: Keep the house form; rule once at vault level for future courses.

Five student-facing lesson retitles cleared the agency rules — keep them?

The course says: five lesson titles were reworded so no quantity acts. S10's three, verbatim:

"Most stresses leave K unchanged"
"The one stress that changes K itself"
"Which way K changes with temperature"

S02 retitled its two convergence lessons (EQM-011/016) the same way. Register title cells keep their own wording — only what the student sees changed.

My recommendation: Keep the retitles.

Capstone lessons run longer check sets than the 2–4 default — licences stand?

The course says: eight capstone/synthesis lessons carry six-to-nine-check sets (S02 ×2, S03, S09, S10's L02 nine and L03 eight, S11, S12) — the Unit 5/6 precedent: a multi-step synthesis needs the scaffold, then the whole performance.

My recommendation: Licences stand.

Order the "spoonful" micro-pass on the AgCl saturation lesson?

The physics: S11 L05's scene stirs a "spoonful" of AgCl into water — but AgCl saturates at about 2 mg per litre. A spoonful is thousands of times past saturation; even "pinch" exceeds it.

Why it is not a one-word patch: the noun threads through the article, the figure tags AND sealed item surfaces (items 29–31, including an adjudicated stem clause). Patching one surface would desync surfaces the QA just sealed. The honest fix is a themed micro-pass — a trace-scale noun or a re-scaled scene — across article, tags and items together, then a sequestered re-read.

My recommendation: Run the micro-pass (small, self-contained, with its re-read).

Accept 2021 Q6(c)'s draw→describe conversion?

The released part, verbatim:

"Draw a particulate representation of PbSO₄(s) and the ions dissolved in the solution in the beaker on the right in the diagram. Draw the particles to look like those shown to the right of the beaker. Draw an appropriate number of dissolved ions relative to the number of dissolved ions in the beaker on the left."

Ours: a describe task scoring the same three features for the same one point — the platform has no drawing surface (the same conversion you accepted on U5-P2 and U6-T1). Everything else in Q6 is text-verbatim. The rubric mirror rides this card.

My recommendation: Accept the conversion.

2023 Q2 used in part: accept the split, and how should the ledger count parts?

The course says: practice P2 carries parts (a)/(b)/(e)/(f) of the released 10-point long — 5 points, text-verbatim; parts (c)/(d) (5 points of pure Unit-2 drawing retrieval) are NOT consumed, because no Unit 7 figure family draws PE curves or Lewis diagrams and a prose conversion would change the skill. The page tells the student plainly which five of the ten points it carries.

The open policy question (your CARD-ASSESS-1 family): should the usage ledger track released questions at PART level, so a future Unit-2 refresh could legitimately spend (c)/(d)?

My recommendation: Accept the split; track at part level.

A real FRQ uses a hydroxide the course fences off — accept the tension?

The course says: hydroxides are kept OFF authored Unit 7 items (they belong beside Unit 8's acid–base work). Practice P1's 2023 Q7 is Sr(OH)₂ — but it is a reserved REAL released FRQ, a declared retrieval, not an authored instance. The fence binds what we write, not what the College Board wrote.

Recorded honestly: no licence was invented; the tension comes to you as the builder flagged it.

My recommendation: Accept — the real question stays.

The test is solubility-heavy (13 of 32 points) — keep, or commission a gas long?

The course says: 3 MCQs + the authored 10-point TlCl long put 13/32 points (41%) on solubility territory that holds 17% of the unit's atoms.

Why it shipped: licence-forced — TlCl was the only fresh-instance grant left and the letter-system roster was empty; AND exam-faithful — the released unit-7 FRQ record is itself solubility-heavy (2023 Q7 and 2021 Q6 are both Ksp; your bias-to-the-exam directive carries it).

Cost of the alternative: a fresh gas-system grant, re-adjudication, and a new 10-point build with its full QA chain.

My recommendation: Keep as built.

Incept question generation was skipped for U7 practice/test — ratify?

The course says: the standing practice (since 2026-09-25) tries Incept candidates for practice and test items. For Unit 7 there was no legal lane: the assessment round allowed only real released contexts, licensed reuses, instance-free forms and named grants — transplanting a generated item onto those would discard the candidate's substance. The deviation is recorded.

My recommendation: Ratify the skip; Incept resumes where the licence space allows.

Sequence S01 of 12

Reversible processes and the equilibrium state

8 lessons · 42 checks (all MCQ) · 15 figures + 2 photos

• Cold solve + blind read42/42 and 42/42, zero clashes, zero second-defensibles.
• What the fixes wereTwo check systems swapped onto fresh instances (your H₂+I₂ bar); one "reversible" example swapped for a modest-K pair (the old one is quantitatively one-way); a family answer-tell ("back" only in keys) killed.

Sequence S02 of 12

Approach to equilibrium: rates and graphs

11 lessons · 49 checks (43 MCQ + 4 numeric + 2 sketch) · 50 figures + photo

• Cold solve + blind read49/49 blind with both sketch tasks performed 8/8 and 8/8, zero clashes.
• What the fixes wereThe tube dataset re-drawn to the pinned canonical composition (an expert temperature-consistency catch); an impossible drift-table row corrected; every concentration-time plateau then re-measured pixel-exact on all 56 graph surfaces.

Sequence S03 of 12

Reaction quotient and equilibrium constant

12 lessons · 61 items (60 MCQ + 1 FRQ flow) · 16 figures + shared photo

• Cold solve + blind read61/61 and 61/61 (FRQ 2/2), zero clashes; every numeric recomputed from the stimulus alone.
• What the fixes wereA real chemistry catch ("deeper brown ⇒ larger Q" is false without the equal-fill line — one sentence fixed it); a sensor-pair value aligned to the pinned N₂O₄ determination; two template figure bugs found here and fixed engine-wide (dark-mode shade order; a subscript-paren mis-draw that touched two drawn keys).

Sequence S04 of 12

Calculating K and the ICE table

7 lessons · 42 items (27 MCQ + 14 numeric + 1 FRQ) · 39 figures

• Cold solve + blind read42/42 + FRQ 4/4 cold; blind 41/41 + 4/4, zero clashes; all 48 datasets recomputed by script.
• What the fixes wereOne feedback forward-leak re-instanced; numeric tolerances settled empirically (the shipped grader coaches unrounded correct entries to the right significant figures — proven by driving it).

Sequence S05 of 12

Magnitude of K

4 lessons · 21 checks (all MCQ) · 12 figures + photo

• Cold solve + blind read21/21 and 21/21, zero clashes; every estimated K independently re-derived (weakest class margin 18 powers of ten).
• What the fixes wereA grammar answer-tell (singular agreement) killed; two complex-question stems recast ("Which best describes the equilibrium mixture?").

Sequence S06 of 12

Algebra of equilibrium constants

8 lessons · 49 checks (27 MCQ + 22 numeric) · 24 figures incl. photo

• Cold solve + blind read49/49 and 49/49, zero clashes; all route arithmetic exact-Decimal, 96 values recomputed independently.
• What the fixes wereFour stems re-pinned to the correct P₄ temperature (1550 K, not 1800); the mis-attributed sulfur-plant phrase swapped (your veto card above); the WGS teaching value settled at Kc 4.0 at 800 K after an independent literature check.

Sequence S07 of 12

Q, direction and equilibrium concentrations

12 lessons · 60 items (44 MCQ + 15 numeric + 1 FRQ) · 23 figures + photo

• Cold solve + blind readCold 59/59 + FRQ 4/4; blind 60/60 with all four FRQ points, zero clashes. The blind reader's own verdict on your algebra boundary: no item made them reach for the quadratic.
• The big catchThis pass found an engine bug unit-wide (an internal FRQ source note rendering student-visible) — fixed at the renderer, swept everywhere, two mutation cases added.

Sequence S08 of 12

Particulate representations

4 lessons · 24 checks (17 MCQ + 6 numeric + 1 draw) · 30 figures

• Cold solve + blind read24/24 + draw 5/5 both passes, zero clashes; every particle count proven by the renderer itself (21 count/K pairs) and re-counted by eye in both colour schemes.
• What the fixes wereA stem that answered its own figure; a scatter-seed fix so same-total boxes draw genuinely different clouds; a typed key leaking into the next stem re-counted away.

Sequence S09 of 12

Le Châtelier's principle

7 lessons · 50 checks incl. 2 FRQ flows · 16 figures + 3 photos

• Cold solve + blind read50/50 both passes (FRQs 6/6 points), zero clashes; the heat as-if device read CLEAN cold. All 34 derivations and all eight ΔH values confirmed.
• What the fixes were"Concentrated" → "saturated" on the CaCl₂ demo (and the demo confirmed workable as written); one dilution-seam feedback line; a figure-text hook clash killed on dissent review.

Sequence S10 of 12

Justifying shifts with Q and K

6 lessons · 42 checks (38 MCQ + 3 numeric + 1 FRQ) · 14 figures + photo

• Cold solve + blind read42/42 + FRQ 3/3 both passes, zero clashes; the canonical Q-vs-K sentence appears exactly twice (its licensed sites) and in none of the 115 options.
• What the fixes wereA dry-reforming Kp corrected to 18 at 1000 K (the 3.3 was a 298 K-anchored shortcut — confirmed error); a rote-typing numeric coincidence killed by moving one scaffold system; the two-temperature N₂O₄ pair sourced by two converging routes before specs.

Sequence S11 of 12

Solubility equilibria

12 lessons · 76 checks (65 MCQ + 10 numeric + 1 FRQ) · 18 figures + 2 photos

• Cold solve + blind read76/76 and 76/76 (numerics to the digit), zero mismatches, timestamp-sealed; all 16 solubility lanes recomputed forward from literature Ksp; the 24-salt table verified value-by-value.
• What the fixes wereCaF₂ → SrF₂ re-instance; "at 25 °C" added where stems leaned on it; two impossible-panel figure fixes; half-last-digit tolerances proven through the shipped grader (12/12 drives). The "spoonful" scale finding is deliberately NOT patched — it is your card above.

Sequence S12 of 12

The common-ion effect

4 lessons · 33 items (27 MCQ + 5 numeric + 1 FRQ) · 13 figures

• Cold solve + blind read33/33 and 33/33 (FRQ 3/3 both ways), zero mismatches, sealed; all four suppression lanes recomputed exact (the 7 500× payoff checked).
• What the fixes wereThe coining line gained its before-the-salt criterion (an item's key rested on it); three option reorders; PbBr₂ pinned at the course value. The quiz letter-run leave is your card above.

Gate G1 of 5 · after S02

The equilibrium-state gate

7 exam-form items · capstones EQM-011/019

• Cold + blind7/7 and 7/7, zero clashes, zero second-defensibles.

Gate G2 of 5 · after S06

The K-algebra gate

10 items (6 MCQ + the unit's first 4 numeric gate lanes)

• Cold + blind10/10 and 10/10, zero clashes; the one raised second-defensible dismissed (outside the AP boundary, never armed).

Gate G3 of 5 · after S08

The ICE and particulate gate

7 items (4 MCQ + 3 numeric) · the perfect-square capstone

• Cold + blind7/7 and 7/7, zero clashes — closed with no fix round at all.

Gate G4 of 5 · after S10

The Le Châtelier and justification gate

7 items · capstones EQM-073/076 · the dilution corner defended

• Cold + blind7/7 and 7/7, zero clashes; one feedback reword discharged the single flag.

Gate G5 of 5 · after S12

The whole-unit interleave gate

10 items (6 MCQ + 4 numeric) · fresh granted salts (PbCrO₄, CaC₂O₄) · interleaves S03/S05/S07/S11/S12

• Cold + blind10/10 cold from the key-free bundle, zero clashes; all three expert flags signed off clean (saturation margins, spectator check, pin literature check).

Practice · 2 pages · 13 points

Three real released FRQs, all text-verbatim

P1: 2023 Q7 + 2021 Q6 (8 points) · P2: 2023 Q2 parts (a)/(b)/(e)/(f) (5 points)

• Verbatim verificationExaminer-verified line-by-line against the archived booklets; figures redrawn at the released counts and expert-checked; zero reconstructed stems.
• Blind read13/13 points (P1 8/8, P2 5/5), zero mismatches, sealed.
• Your three cardsQ6(c) draw→describe · Q2 split-use granularity · Sr(OH)₂ fence tension — all above.

Unit test · 32 points

18 MCQ + the real 2024 Q5 verbatim + an authored 10-point long

FRQ1 = 2024 Q5, text-verbatim (4 points) · FRQ2 = authored TlCl solubility long with its disclosure note (10 points) — no pure-U7 released long exists

• Blind read18/18 MCQ + 4/4 + 10/10 = 32/32 points, zero clashes, sealed; both singled-out figures solved from drawn geometry.
• Examiner pass2024 Q5 verified line-by-line; every value recomputed (18 + 15 + 14 lanes); zero key changes anywhere in the test's QA.
• Your two cardsSolubility point-weight (with the gas-long alternative) · the Incept skip — both above.