AP Chemistry Unit 5 — Kinetics — your whole-unit review

## S02. Blueprint adjudications — U5-S02 What changes a reaction's rate (2026-10-07)

Blueprint: BLUEPRINT_U5-S02_rate_factors.md (checker exit 0). The nine calls, answered:

- **S02-1 — Image-pack hooks: YES, all four** (L01 zinc in 1 M vs 6 M acid; L03 milk cartons;
  L05 marble chip vs powder; L07 H₂O₂ + MnO₂ froth) under James's same-day rule ("Code draws
  what must be exact; David's image pack renders what must look real"). Every particle diagram,
  comparison table and box diagram stays figlib. Each generated image passes
  science-image-gen-skills:qc-figure AND our review-figures + give-away audit; the sequence
  record lists pack figures. Engine gains a static `image` fig kind (base64-embedded so the
  page stays self-contained; validator requires alt + existing file; mutation test extended).
- **S02-2 — Capstone decrease cases: YES for the three continuous factors** (diluting,
  cooling, powder→lump): KIN-020's statement says "a described single change", not an
  increase. Catalyst cases are ADDITION-only (the statement names "addition of a catalyst").
- **S02-3 — Why lessons carry what→why→transfer with NO misconception items: YES.** The
  register documents none and the 7 Oct rule bars invented ones. The catalyst-consumed
  misconception is handled structurally by KIN-018's definition, not by an invented item.
- **S02-4 — Catalyst capstone reason: AMENDED.** The blueprint proposed a definition-level
  "reason" for catalyst cases; adopted instead: **catalyst capstone cases are VERDICT-ONLY**;
  the two reason items sample only the three factors whose particle-level reasons S02 teaches
  (concentration / temperature / surface area). A definition restated is not a reason, and the
  register's own deferral (KIN-019 → KIN-113/114) makes the honest line "why comes later".
  This resolves the one genuine register contradiction the blueprint found (KIN-020's
  "matching particle-level reason" vs the documented catalyst deferral) — logged for James.
- **S02-5 — L09 six checks (4 verdict + 2 reason): YES**, with S02-4's amendment; the
  validator's >4-checks warning is blueprint-licensed (S01 L09 precedent).
- **S02-6 — L08's one honest deferral line: YES** ("Why a catalyst speeds a reaction needs
  ideas that this unit teaches later" — no forward term).
- **S02-7 — Scene equations shown once as context: YES** (S01 L09 precedent: plain words
  carry the teaching, formulas shown once).
- **S02-8 — Fresh-species check pools RESERVED as named** (thiosulfate+HCl; Mg+ethanoic acid;
  glow stick; H₂+I₂; antacid; steel wool; sugar cube; catalytic converter Pt; Fe with N₂+H₂;
  bleach+food dye; egg; kindling/logs). Later sequences keep off them; the coordinator carries
  the pool ledger forward in every later brief.
- **S02-9 — The nine headache scenes stand as written**; James edits in place at review.
- **S02-A1 (accuracy, carried to review-accuracy):** MECH-CONC-COLLISION verbatim over the
  Zn-surface instance is the register's own application — keep verbatim; review-accuracy
  gives the particle–surface reading one explicit eye. **S02-A2:** L07/L08's uncatalysed
  H₂O₂ framing must stay consistent with S01's measured 0.0010 M/s (no "sits for weeks").

## S03. Blueprint adjudications — U5-S03 The rate law from initial-rate experiments (2026-10-07)

Blueprint: BLUEPRINT_U5-S03_rate_law.md (checker exit 0). The ten calls:

- **S03-1 — NO + H₂ thread carries the taught examples L01–L18: YES** (register-bound
  exemplar; B4 scene continuity; checks fresh).
- **S03-2 — Taught instances may reuse NH₃/N₂O₅/NO₂ although S01's CHECKS used them: YES.**
  B4's disjointness is one-directional: a CHECK must never reuse an instance the student was
  TAUGHT with; later TEACHING with a species an earlier check happened to use invalidates
  nothing (that check was already taken). Register binds these examples anyway. LOGGED FOR
  JAMES as the explicit reading.
- **S03-3 — Invented table values with TRUE orders: YES** (S04-A2's precedent: pedagogic
  values licensed; every qualitative/order claim real; stems supply data as given).
- **S03-4 — The three-way doubling-response sort sits on L04, last of the triplet: YES**
  (the vault's sort rule; S01 C5 precedent).
- **S03-5 — L02 photo-real hook (hot platinum in ammonia): YES, CONSTRAINED** — the image
  shows a hot platinum wire/foil in a flask of ammonia vapour; no flame/air in frame and no
  glow-from-oxidation claim (the classic glowing-wire demo is NH₃ OXIDATION; L02's instance
  is DECOMPOSITION on hot Pt). Caption/prose match the decomposition instance exactly;
  review-accuracy gives this image's chemistry one explicit eye.
- **S03-6 — k-UNITS answers key as MCQ; orders and k values as numeric entry: YES**
  (interaction matches behaviour; typed unit strings are an input fight, not chemistry).
- **S03-7 — One tripling comparison beside the doubling cases: YES** (the atom's behaviour
  is predict-the-factor, not recite-doubling).
- **S03-8 — L18 capstone chains orders → rate law → k in one item set: YES**, under H1
  (a carried value restated only where load-bearing).
- **S03-9 — S02 builds before S03: ADOPTED in practice** — the S03 spec wave launches only
  after specs/U5-S02.json exists, so L01's recap lines cite built lessons.
- **S03-10 — The eighteen headache scenes stand**; James edits at review.
- **S03-A1:** L06's N₂O₅-vs-H₂O₂ k comparison needs supplied same-temperature values — the
  spec author supplies honestly framed pedagogic values; review-accuracy verifies direction
  and framing before the blind read. **S03-A2:** KIN-032 stays fact-only (why deferred to
  KIN-077); no collision/energy language.

## S04. Blueprint adjudications — U5-S04 Concentration change over time (2026-10-07)

Blueprint: BLUEPRINT_U5-S04_integrated.md (checker exit 0; check_parallel run on frames).
The eleven calls, answered under option-(b) delegation:

- **S04-1 — "integrated rate law" coined ONCE at KIN-051's lesson** for all three families;
  no fluency row (B7 — a calculation lesson). EXECUTED with this record: register rev 1.4 adds
  the glossaryAudit entry (integrated rate law → KIN-051) so the validator's forward-term
  fence covers it; validate_design.py re-run to PASS required before spec authoring.
- **S04-2 — The DIFFERENCE form is the one standard equation** (ln[A]ₜ − ln[A]₀ = −kt;
  1/[A]ₜ − 1/[A]₀ = kt), as the register's statements and the AP equation sheet give it
  (meta-directive). The slope-intercept rearrangement is SHOWN once at each linear-plot
  lesson, explicitly mapped to y = mx + b — the same equation rearranged, never a second
  standard equation.
- **S04-3 — Calculator ln/eˣ: YES** (the AP exam supplies a calculator); value sets still key
  exactly (0.693-multiple logs, integer reciprocals) so wrong-maps stay computable.
- **S04-4 — One value set governs the N₂O₅ thread: YES** (k = 0.0231 s⁻¹, t½ = 30.0 s —
  exact 0.693/0.0231). The register's bracketed digits self-conflict (25 s vs 30 s vs "100 s");
  examples are scope fences, not binding digits — unified on the record, LOGGED FOR JAMES.
- **S04-5 — L05's draw item sketches all three order curves on one shared pair of axes: YES**
  (the contrast is the atom's point; the checklist names each curve's feature separately).
- **S04-6 — Fresh check species stand** (N₂O/Pt, cyclopropane, azomethane, HI, NOBr, I-131,
  P-32, Co-60) — review-accuracy explicitly confirms each claimed order at spec QA.
- **S04-7 — L20 contrasts first order's constant half-life against SECOND order only: YES**
  (no atom states zeroth-order halving; honest scope). No second-order half-life formula
  anywhere (register exclusion).
- **S04-8 — L22 derives t½ = 0.693/k in one worked line: YES** — shown once as teaching;
  one-step items show answers directly (Oct rule), no reveal chain.
- **S04-9 — L25 opens on a photo-real excavation scene: YES** (image pack, qc-figure +
  review-figures gated); the decay curve and all data stay figlib.
- **S04-10 — "linearize" stays OUT of the course; plain "comes out straight" wording: YES**
  (jargon-where-possible rule). Not added to the glossary; noted here as the ruling.
- **S04-11 — The twenty-seven headache scenes stand as written**; James edits at review.
- **S04-A1:** AST-008A (isotope) has NO built lesson (Unit 1 predates the pipeline) — L25
  carries an R177 recall MCQ + one inline recap line restating the definition. **S04-A2:**
  pedagogic k values (NO₂ 0.010 M⁻¹·s⁻¹; NH₃ 5.0×10⁻⁴ M/s) are licensed — qualitative order
  claims are real chemistry; review-accuracy confirms. **S04-A3:** "rise over run" is the
  slope term, not the banned experiment verb — stands (S01 precedent).

## B4(U5). Cross-sequence check-pool convention (coordinator, 2026-10-07 — logged for James)

S03 and S04 both reserved cyclopropane (and S04 azomethane, which S01's checks touched).
RULING: within a sequence, sibling option pools stay disjoint and no check ever reuses ANY
taught instance unit-wide (B4 verbatim). ACROSS sequences, check-pool reuse of a species is
LICENSED when the new item tests a different atom — a species met in an earlier CHECK hands
no answer to a later item about different behaviour. Cyclopropane stays in both S03's and
S04's check pools; azomethane stays in S04's.

## S05. Blueprint adjudications — U5-S05 Elementary reactions (2026-10-07)

Blueprint: BLUEPRINT_U5-S05_elementary.md (checker exit 0). The nine calls:

- **S05-1 — The register's NO/ozone chemistry carries the taught examples: YES.**
- **S05-2 — The bare "2 NO₂" squared-exponent example completes as NO₂ + NO₂ → NO₃ + NO**
  (a real elementary step — the accepted slow step of the NO₂ + CO mechanism, reusable by
  S08/S09 — and it avoids colliding with S03's experimentally-taught 2 NO₂ → 2 NO + O₂).
- **S05-3 — L01 teaches "a single particle-level event": YES.** The register's "single
  collision event" phrasing stays for the bi/termolecular picture; one honest line says a
  unimolecular step is one particle changing on its own (an O₃ molecule splitting involves
  no collision). Truth over echo — review-accuracy eyes L02's one-particle line; LOGGED FOR
  JAMES as a deliberate scoping of the register sentence.
- **S05-4 — L05's example run crosses the plausibility boundary BOTH ways on the one
  NO + O₂ scene: YES** (7 Oct both-crossings rule; boundary pair keeps the same object).
- **S05-5 — Reserved check pools stand, including every declared B4(U5) reuse** (2 HI, I+I+Ar,
  2 SO₂+O₂, 2 H₂+O₂ — each tests a different atom than the earlier use; declared in-spec).
- **S05-6 — L06 retests KIN-027's coefficient-copying belief on a fresh overall reaction:
  YES** — the sequence's documented misconception, never an invented slot.
- **S05-7 — The S05 spec wave waits for specs/U5-S03.json and U5-S04.json: YES** (S03-9
  precedent; recap lines cite existing lessons).
- **S05-8 — figlib_u5_particles approved** as the spec wave's documented extension (exact
  particle-event drawings; eqstack/anneq/tables cover the rest; no image-pack figures in S05).
- **S05-9 — The six headache scenes stand**; James edits at review.
- **S05-A1 (binding item constraint):** "implausible" is NEVER keyed on a real termolecular
  step (I + I + Ar; O + O₂ + M); it sits only on proposed single steps whose real chemistry
  is multi-step. "Rare ≠ impossible" wording verified by review-accuracy.

## S06. Blueprint adjudications — U5-S06 The collision model (2026-10-07)

Blueprint: BLUEPRINT_U5-S06_collision.md (checker exit 0). The eleven calls:

- **S06-1 — NO₂ + CO carries the taught examples: YES** (register-bound; S07's thread is the
  same reaction — deliberate continuity).
- **S06-2 — L01 carries the honest scope line (this one-collision picture of NO₂ + CO is its
  high-temperature behaviour): YES** — it is what stops S08's two-step low-temperature
  mechanism contradicting S06 on the same reaction. Prose stays qualitative ("at high
  temperatures"); review-accuracy verifies before any specific kelvin figure is printed.
- **S06-3 — L03 teaches on H₂ + I₂ though an S02 check used it: YES** (S03-2 one-directional
  reading); H₂ + I₂ now leaves every future check pool.
- **S06-4 — L05 bridges the U3 speed-axis curve to the new energy-axis curve: YES** (PSM-070
  resolves to the SEEDED U3-S08 L08; the recap SHOWS the speed curve; energy axis is the
  new teaching).
- **S06-5 — L05 graph reads stay fully qualitative: YES** (the register excludes numeric
  fraction computation; identify-and-compare only).
- **S06-6 — L04 closes with the three-way diagnosis sort: YES** (sort sits on the pair's
  last lesson, C5 precedent).
- **S06-7 — L06 names the multiplying tail, not extra collisions, as the main cause: YES**
  (collision frequency grows only ∝ √T; the tail dominates — review-accuracy confirms the
  "few percent" magnitude line).
- **S06-8 — L07 image-option flaw variants ADOPTED as proposed** (wrong-way peak shift;
  taller warmer curve; equal tails with moved Ea) — each from a named misconception path;
  neutral alts.
- **S06-9 — L08 prediction items vary temperature and concentration only; orientation is
  explain-only: YES** (orientation is not a student-manipulable variable in described
  experiments).
- **S06-10 — The reserved pool stands; LEDGER GRAIN CONFIRMED as REACTION grain** — the
  instance is the reaction/scenario, not the species, as the ledger has run since S01.
  ICl + H₂ is fresh; argon/neon bare-curve reads licensed under B4(U5) (different unit and
  atom, named). LOGGED FOR JAMES.
- **S06-11 — The eight headache scenes stand**; James edits at review.
- **S06-A (carried to review-accuracy):** the H₂ + I₂ "collisions too gentle" line never
  calls the reaction elementary (historically it is not simple one-step); L02's
  "billion collisions each second" and L06's √T magnitude lines verified; KIN-077 closes
  BOTH deferrals (KIN-015 sharpening + KIN-032's why) — both recaps named. One image-pack
  candidate LICENSED: L03's sealed flask of purple iodine vapour with hydrogen, visibly
  unreacting at room temperature (qc-figure + review-figures gated).

## S07. Blueprint adjudications — U5-S07 Energy profiles and temperature dependence (2026-10-07)

Blueprint: BLUEPRINT_U5-S07_profiles.md (checker exit 0). The seven calls:

- **S07-1 — The symbolic form k = Ae^(−Ea/RT) appears NOWHERE: ADOPTED.** KIN-090's State
  behaviour is satisfied by naming the equation and what it ties together; the CED excludes
  Arrhenius calculations, and D6's engine sweep bans the machinery — showing the form would
  need a sweep exception for one decoration. LOGGED FOR JAMES (an AP student may meet the
  symbolic form elsewhere; his call to add it as a shown-not-used panel).
- **S07-2 — The NO₂ + CO thread keeps the register's 20 / 90 / −30 kJ/mol levels: YES**
  (the one consistent profile the KIN-084/085 brackets define; pedagogic magnitudes licensed,
  S04-A2; review-accuracy confirms direction claims only).
- **S07-3 — Generic "reaction A / reaction B" check profiles: YES** — the register's own
  bracket form; logged against the name-the-specific-thing rule (the register governs here).
- **S07-4 — L08 carries one image-option MCQ beside the pen-and-paper draw item: YES**
  (U5-G3 samples KIN-087 in the image-option format per the drawing-assessment ruling).
- **S07-5 — Overall-change items key the SIGNED value (signRetry); released/absorbed asked
  separately: YES** (KIN-002 convention parallel; verdict and value never bundled).
- **S07-6 — Zero image-pack photographs in S07: YES** (every scene is particulate/graphical;
  code draws what must be exact).
- **S07-7 — The twelve headache scenes stand**; James edits at review.
- **S07-A (carried to review-accuracy):** the "two otherwise similar reactions" hedge on
  KIN-088/089 stays and is never unpacked (frequency factor banned); the L12 proportional
  tail-growth figure gets one explicit accuracy eye; the coined term is the full "reaction
  energy profile". S06's spec wave builds before S07's (recaps cite built lessons).

## S08. Blueprint adjudications — U5-S08 Reaction mechanisms (2026-10-07)

Blueprint: BLUEPRINT_U5-S08_mechanisms.md (checker exit 0). The thirteen calls:

- **S08-1 — NO₂ + CO two-step thread via NO₃ for L01–L05/L08: YES.**
- **S08-2 — L01 carries the low-temperature scope line (S06-2's twin), qualitative only: YES.**
- **S08-3 — Cl + O₃ converts from check species to TAUGHT instance: YES** (S03-2 reading;
  it leaves every future check pool — ledger updated).
- **S08-4 — The taught chlorine mechanism uses the two-O₃ form (net 2 O₃ → 3 O₂), framed as
  "one mechanism by which chlorine destroys ozone": YES**; review-accuracy verifies the
  framing against the dominant ClO + O stratospheric cycle.
- **S08-5 — The capstone tests Cl as an INTERMEDIATE in fresh 2 NO₂Cl → 2 NO₂ + Cl₂,
  against its taught catalyst role: YES** — the role-follows-pattern discrimination is the
  capstone's point.
- **S08-6 — The NO-catalysed SO₂ oxidation stays in L07's pool despite its embedded taught
  step: YES, declared.**
- **S08-7 — The capstone's fresh catalysed mechanism is the OH-catalysed 2 O₃ → 3 O₂: YES**
  — same net equation as the taught Cl case is a feature (different catalyst, same net),
  not leakage; the item names its candidate species plainly. Logged.
- **S08-8 — L05 may judge an order-reversed case "not an intermediate" before the catalyst
  lesson: YES** (verdict by the order test; the catalyst LABEL waits for L06).
- **S08-9 — Generic species letters (Q, X, Z — never element symbols): YES, one item.**
- **S08-10 — 2 N₂O serves L01 and L08: YES, declared** (the second item hands nothing over).
- **S08-11 — L06 opens on the photo-real ozone-hole image: YES** (image pack; qc-figure +
  review-figures gated; the "tens of thousands" magnitude verified by review-accuracy
  before it prints).
- **S08-12 — All steps stay speed-unlabelled; fast/slow is entirely S09's: YES.**
- **S08-13 — The nine headache scenes stand**; James edits at review.
- **S08-A (carried to review-accuracy):** L01's lower-temperature energy claim (no kelvin
  figure without verification); NO₃ spectroscopic detectability (L08 leans on it); 2 N₂O's
  simplified two-step framed honestly (real O + N₂O branching exists); L02's hypothetical
  rejected proposal read as the device it is; KIN-097's "net concentration is constant" =
  amount-unchanged-overall, no item implying instant-by-instant constancy.

## S09. Blueprint adjudications — U5-S09 Mechanisms and rate laws (2026-10-07)

Blueprint: BLUEPRINT_U5-S09_mech_ratelaws.md (checker exit 0). The nine calls:

- **S09-1 — The register's two threads stand** (NO₂ + CO for L01–L04; 2 NO + O₂ via N₂O₂
  for L05–L08, explicitly closing S05's rejected one-step story).
- **S09-2 — ⇌ first appears in L06 with its one-line gloss; L05 keeps plain arrows: YES**
  (the gloss and the reversibility arrive together).
- **S09-3 — Exactly ONE lettered AP-style mechanism carries the zero-order check: YES,
  flagged** — no real fresh slow-first mechanism with a fast-step-only reactant and
  whole-number orders exists (the honest 2 O₃ pre-equilibrium gives a negative order,
  excluded); AP released items use lettered mechanisms; S07-3 precedent.
- **S09-4 — Every judgment check splits verdict from reason; zero bundles in S09: YES.**
  AP-shaped bundles, if wanted, live in U5-G4/test items only.
- **S09-5 — The worked substitution collects constants into k_obs; items accept k or
  k_obs: YES** (D4 executed as adopted).
- **S09-6 — Declared B4(U5) reuses licensed: ICl + H₂, NOBr, NOCl** (each tests a different
  atom, named in the rows).
- **S09-7 — The five named fresh mechanisms reserve for S09's checks: YES** (NO₂+F₂ via F;
  NO₂+O₃ via NO₃; NO₂Cl via Cl; 2 NO+Br₂ via NOBr₂; 2 NO+Cl₂ via NOCl₂ — ledger updated).
- **S09-8 — One honest qualitative line names the low-temperature scope of the measured
  NO₂ + CO rate law: YES** (no kelvin figure; review-accuracy eyes it — this is what keeps
  S06/S08/S09 telling one story about one reaction).
- **S09-9 — The eight headache scenes stand**; James edits at review.
- **S09-A (carried to review-accuracy):** L06's "stops changing" wording stays clear of
  steady-state vocabulary (grep-verified at blueprint level — hold it at spec level); the
  literature's consistent fast-first route for 2 NO + O₂ means L08's slow-first rival is
  judged as a PROPOSAL ("supported, never proven" sits in L04's forest); negative orders
  never assessed.

## S10. Blueprint adjudications — U5-S10 Multistep energy profiles (2026-10-07)

Blueprint: BLUEPRINT_U5-S10_multistep.md (checker exit 0). The eight calls, all ADOPTED as
recommended:

- **S10-1 — Thread profile keeps S07's 20/−30 end levels, new gridline interiors (100/40/90).**
- **S10-2 — The valley is labelled NO₃**, closing the loop with S08's intermediate.
- **S10-3 — Generic taught figures for L02/L03: YES** — extends S07-3's generic licence from
  checks to TAUGHT figures where the thread cannot carry the teaching job (the thread agrees
  on tallest-peak-vs-largest-climb, so it cannot teach the contrast). Logged.
- **S10-4 — One L02 check reads the real H₂ + ICl two-step profile** (B4(U5) declared;
  review-accuracy verifies the mechanism H₂ + ICl → HI + HCl, HI + ICl → I₂ + HCl before
  the spec ships).
- **S10-5 — Sketch items two-step only; three-step confined to the reading lessons.**
- **S10-6 — L04 carries one image-option MCQ beside the draw item** (G5 samples KIN-112 in
  that format).
- **S10-7 — The tallest-peak misconception is tested by the disagree-case MCQ**, not a
  separate claim item (the contrast is the atom's own wording; S02-3's bar respected).
- **S10-8 — The four headache scenes stand**; James edits at review.
- **S10-A (carried):** thread energies pedagogic, direction claims only (S04-A2/S07-2); the
  slow-step assignment must stay consistent with S08/S09's slow-first-step story —
  review-accuracy gives it one explicit eye. S10's leans-on cells cite register-position
  lesson numbers — the coordinator re-points them if S08/S09 blueprints renumber; S08/S09
  spec waves build before S10's.

## S11. Blueprint adjudications — U5-S11 Catalysis (2026-10-07)

Blueprint: BLUEPRINT_U5-S11_catalysis.md (checker exit 0). The seven calls:

- **S11-1 — H₂O₂ carries the taught thread with three catalysts (MnO₂ → iodide → catalase):
  YES** (the one-substrate-many-catalysts contrast is the sequence's quiet spine).
- **S11-2 — L02 teaches on Pt + H₂ + C₂H₄ (KIN-120's register instance) with the mode label
  deferred to L08: YES** (deliberate early use of the bracket instance; fence kept).
- **S11-3 — The catalytic converter converts from S02's check ledger to TAUGHT (L08 headache
  + photo-real cutaway): YES** (S03-2 reading; it leaves every future check pool; image
  licensed, qc-figure + review-figures gated).
- **S11-4 — "acid–base catalysis" and "surface catalysis" get glossary entries and R233
  fluency rows at L07/L08: YES — EXECUTED with this record**: register rev 1.5 adds both
  glossaryAudit entries (→ KIN-119, → KIN-120) so the forward-term fence protects the mode
  labels; validate_design.py re-run required to PASS before the S11 spec wave.
- **S11-5 — Classification checks use single-mode cases, overlap named in prose, plus ONE
  "which classifications apply" item keying both on the platinum case: YES.**
- **S11-6 — Fe³⁺-catalysed persulfate–iodide reserves as the capstone's fresh check: YES**
  (review-accuracy verifies the mechanism's honest two-step account).
- **S11-7 — The ten headache scenes stand**; James edits at review.
- **S11-A (carried to review-accuracy):** iodide/H₂O₂ classified as the binds-the-reactant
  mode (IO⁻ as catalyst-bound intermediate) gets an explicit eye — it anchors L09's boundary
  pair; catalase turnover magnitude verified before "millions every second" prints; KIN-117
  keyed only in the softened conditional form; MECH-CAT-PATH's because+so chain exempt from
  the two-clause split (declared). Commit-line practice CONFIRMED: a record names the commit
  actually read (S11's 68ac518 stands; ade1f47-pinned earlier records stand as written).

## S04-S. Spec-stage adjudications — U5-S04 author's flags (2026-10-07)

- **"spontaneous" (KIN-063): the course wording stands** ("the breakdown of unstable atomic
  nuclei, happening on its own with no outside trigger") — the validator's U9 exclusion
  battery governs STUDENT text; the register's internal statement wording needs no carve-out.
  Logged for James/register keeper.
- **S01 pool enumeration corrected**: S01's built checks DO use cyclopropane, azomethane and
  HI — S04's reuse of all three is licensed under B4(U5) (different atoms: tangent reads
  there, integrated-rate-law work here) and now DECLARED rather than accidental.
- **HI carries two orders in two conditions** (gas-phase second order; zeroth on hot gold):
  licensed — real chemistry, each stem self-contained, review-accuracy verifies both claims.
- **Cyclopropane k = 2.31×10⁻³ s⁻¹ echoing N₂O₅'s 0.0231 ×10⁻¹, and L19's reuse of L02's
  curve window: carried to review-questions** for a deliberate second eye.
- **Linear Unicode rendering of the integrated laws and t½ = 0.693/k matches the AP sheet's
  own linear forms — stands**; stacked-fraction upgrade noted as a James option.

## S03-S. Spec-stage adjudications — U5-S03 author's flags (2026-10-07)

- **The L06 k-pair direction (k(H₂O₂) > k(N₂O₅) "at 25 °C") is COURSE-WORLD, not
  literature-world** — it follows from S01's own taught flask (recorded Leave: the generous
  0.0010 M/s) and L03's approved N₂O₅ data. Both compared reactions are first-order, so the
  comparison is dimensionally honest. STANDS as supplied, pending the S03 review-accuracy
  ruling on framing; ON JAMES'S CARD LIST (the alternative — renumbering S01's thread — is
  his call, not ours).
- **One-running-scene rule across sequences: temperature-tag the datasets.** The same species
  may not show two k values at one implied temperature. S03's N₂O₅ keeps "at 25 °C"
  (5.0×10⁻⁴ s⁻¹). S04's N₂O₅ dataset (0.0231 s⁻¹) GAINS "at 65 °C" and S04's NO₂ dataset
  (0.010 M⁻¹·s⁻¹) GAINS "at 300 °C" — pedagogic values, named regimes (S04-A2 licence).
  EXECUTION: bundled into S04's next touch-round (after its in-flight QA passes land), with
  a cold re-read of every touched stem.
- **L18's H1 chain stands** (the restated rate law is load-bearing; the gate's
  whole-question archetype is the reason) — noted for the S03 blind reader.
- **Printed-then-determined pools (BrO₃⁻, CH₃Br) stand** — the capstone demands the
  derivation and k; carried to review-questions for a deliberate eye.
- **The S117 warning on KIN-025 is licensed** ("a proportionality constant" is the correct
  pre-coin naming; "rate constant" is KIN-026's coin).

## S06-S. Spec-stage adjudications — U5-S06 author's flags (2026-10-07)

- **"collision model" is TAUGHT, not title-only: register rev 1.6** adds the glossaryAudit
  coin (→ KIN-072); the S06 QA fix round adds the one coining line in L01's forest ("This
  picture is called the collision model") + one term-recall item appended to L01's fluency
  run. Meta-directive governs (the AP exam uses the term). LOGGED FOR JAMES.
- **L07's missing R177 recall (deviation from the usual pattern): STANDS** — the adjudicated
  blueprint plan carries none; L06's recall drills the same rules one lesson earlier and
  L07's forest restates them. Flagged, not silently patched.
- **The drawn M–B energy form f(E) ∝ (E/ep)·e^(1−E/ep)** (2-D form, the standard AP
  schematic — honest ordinal claims, machine-asserted; the 3-D √E form would make a true-Ea
  tail invisible): carried to review-accuracy for the explicit blessing.
- **Review-accuracy eyes ordered:** the H-to-Cl contact stem datum; L02's "billion
  collisions each second"; L06's "few percent more collisions"; the one deliberate
  check_parallel FLAG (register-statement sentences left unwarped).

## E. QA-round adjudications (2026-10-07, after the accuracy + question passes)

Both non-author passes are on file (qa/REVIEW_ACCURACY_S01.md; qa/REVIEW_QUESTIONS_S01.md).
Cold-solve: 42/42 keys matched. One Essential, found independently by BOTH reviewers.

- **E1 — L04 line J/K (ESSENTIAL, both reviewers): FIX.** The drawn secant (20 s → 60 s) is
  parallel to the tangent — option K is numerically defensible and its feedback false as
  drawn. Adopted fix: the accuracy reviewer's smaller change — secant.p2 → [80, 0.35] (still a
  knot; slope −0.0033 M/s ≠ −0.0040), feedback clause "…at 20 s and again at 80 s", label
  nudged if it collides. Changed item gets a fresh blind read (H4 rule).
- **E2 — L09 iodine + propanone "no ions form" (false fact) + the check-count/duplicate
  findings: FIX by CUTTING.** The iodine item is cut (the real reaction ionizes — the dead
  handle is live); if two of the remaining method items are the same design move with nouns
  swapped, the weaker duplicate is also cut. L09 keeps its 2 recall MCQs + one method item per
  instrument (+ the transfer case if distinct); the validator's >4-checks warning stays
  acceptable (blueprint-licensed).
- **E3 — "measured" wording (L06/L07, 3 spots + column head): FIX** to non-measurement words
  ("the two species the equation links"; column head "Δ[X]/Δt"). The Δ[O₂]/Δt formalism itself
  stands (register's own bracketed examples).
- **E4 — L03 secant slope vs sign: FIX** — "the SIZE of that joining line's slope is the
  average rate"; figure label "slope size = average rate over 0–60 s". Keeps the sign step
  honest without teaching L04's content early.
- **E5 — L04 NO₂ magnitudes: FIX** — both NO₂ figures' y-values ÷10 (knots, ticks, tangent,
  secant), nq key 0.0040 → 4.0 × 10⁻⁴ M/s with wrong-map entries rescaled. Changed item gets
  a fresh blind read.
- **E6 — KIN-004 nq inside a State atom (W1): KEEP, logged for James.** The blueprint's
  adjudicated check column mandates the find-the-rate nq (the 7 Oct rule: a fact naming a
  quantity is tested by finding it); the item reads a SUPPLIED drawn tangent — KIN-011 owns
  choosing read points and constructing the slope from a bare curve. Deliberate overlap,
  James may veto at review.
- **E7 — W3 taught-instance halogen swap (L09 2-bromo-2-methylpropane): FIX** — re-instance
  to a visibly different ions-forming reaction (fix agent's choice under write-mcqs; chemistry
  must be true, timescale followable, no reuse of any sibling pool; NOT a halide swap of the
  taught tert-butyl chloride).
- **E8 — W4 stem/key head-noun echo: FIX** (reword the stem so the key's noun is not handed
  over).
- **E9 — W11 all four claim items key No: FIX** by converting ONE claim item to a TRUE claim
  (the author's own suggestion) — never flip a key over an unchanged stem; fix agent picks the
  lesson where a true claim is most natural.
- **E10 — Small wording/contract fixes, all adopted:** the self-contradicting fb opener
  (KIN-005/mcq3); the "again" lean-back in KIN-008/nq2 (stems self-contained); KIN-010/nq1's
  missing whole-table wrong-map value −1.61e-8 added (the low-plausibility mapped path may be
  dropped); scientific-notation entry hints made consistent across the three e-key items; the
  two failed option-count challenges get their genuine third options (KIN-004/mcq3
  height-for-slope; KIN-009/mcq0b electrons-flow-through-solution) with recomputed feedback.
- **E11 — Figure label overlaps (coordinator's browser sweep): FIX** — KIN-009 calline
  ("0.00"/"0.0" and "0.6"/"A = 0.60") and KIN-011 (axis tick 0.80 vs point label "(0 s,
  0.80 M)") — offset or suppress the colliding label in the figlib params/renderer.
- **E12 — Every Leave row stands as left** (both reports' Leave tables; the accuracy report's
  deliberately-omitted list binds the blind reader too).

## F. Figure-pass adjudications (2026-10-07, after review-figures on the rendered page)

Report: qa/REVIEW_FIGURES_S01.md (20 figures judged as rendered; give-away audit 7/7 clean).

- **F1 — L01 flaskpanels bubble-over-label (ESSENTIAL): FIX at template level.** A bubble sits
  on the "7" of "0.74 M" so the render reads "0.04 M" — the sequence's anchor datum. The
  renderer keeps a bubble-exclusion zone around the concentration label (any future bubbled
  panel inherits the fix), rather than a one-off nudge.
- **F2 — L09 calline strike-throughs (ESSENTIAL): FIX.** "A = 0.60" moves off the x = 0.04
  gridline; "0.050 M" moves off the dashed read line; the read line gets a visibly distinct
  style from the grid (guide colour / heavier dash) since its horizontal leg legitimately
  coincides with the y = 0.60 gridline.
- **F3 — L04 "tangent at t = 50 s" label floats far from its line (item figure): FIX** —
  anchor the label beside the tangent's upper end.
- **F4 — L06 two-curve steepness nearly invisible (Worth fixing): FIX as suggested** — draw
  rise/drop brackets over ONE shared interval (e.g. 0–40 s: −0.20 M vs +0.10 M), no rescale;
  the honest shared axis stands.
- **F5 — Blind (screen-reader) answerability of item figures: NOT fixed here** — an
  engine-level course-wide question (U4 has the same property); logged on James's list in
  BUILD_STATE. Item alts stay neutral per the 2026-08-06 ruling.
- **F6 — The four Leave rows stand** (incl. fig 10's header: its cells show the sign step and
  the lesson teaches it; the header names the division, which is the column's single job).

## G. Blind-read adjudications + final figure fixes (2026-10-07, closing the QA round)

Blind read (qa/BLIND_READ_S01.md, fresh reader, key-sequestered): **41/41 keys matched, zero
clashes, zero second-defensible options, zero Essentials; verdict READY.**

- **G1 — The seven Worth-fixing prose edits (B-1…B-7): ADOPTED verbatim** from the record's
  rewrite column and applied by the coordinator (staging imperatives given a named actor or
  recast as descriptions; the two double-idea reveal lines split/trimmed; L06's back-reference
  names lesson one; L09's next-lesson preview cut). B-8/B-9 (unit-level motivation lines):
  LEFT, as the reader judged.
- **G2 — Figure fixes F1–F4 executed** (coordinator): F1 flaskpanels bubbles structurally
  confined between the liquid surface and the concentration label; F2 gridlines made faint
  (--line hairline) in conccurve AND calline so guide-class dashed reads stand apart, calline
  read labels given a paper halo (paint-order stroke); F3 the L04 "tangent at t = 50 s" label
  anchored beside its line's lower end; F4 a new optional conccurve "drops" bracket (serif-
  capped change bracket + leader) added and used on L06 ("falls 0.06 M in 60 s" / "rises
  0.03 M in 60 s"), honest shared axis kept.
- **G3 — Engine bug found while verifying G2: the course CSS (.dg .flabel{text-anchor:middle})
  silently overrides a renderer's text-anchor ATTRIBUTE**, re-centring every start-anchored
  label over its bracket/point. Fixed in figlib_u5_curves.py by inline style (style beats
  class). NOTE FOR THE VAULT + U4: any figlib module using text-anchor attributes under a
  class that sets text-anchor has the same latent fault — vault-sync candidate.
- **G4 — Post-fix verification, all clean:** spec validator 0 errors; design validator 0
  errors; mutation suite 31/31; figure sweep 20 figures 0 problems; clickthrough 122 pages /
  41 checks / 0 failures / 0 console errors; changed figures re-screenshotted and eyeballed by
  the coordinator (flask label legible, J/K discriminating, calline clean, brackets clear);
  review page regenerated (blind-read strip row filled).

## E-S02. QA-round adjudications — U5-S02 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S02.md (0 Essential, 5 Worth fixing, 4 Leave; both ordered eyes
S02-A1/A2 cleared; all keys confirmed); qa/REVIEW_QUESTIONS_S02.md (cold-solve 32/32, 1
set-level Essential, 8 Worth fixing, 3 Leave). One fix round, all by write-mcqs fix agent:

- **E-S02-1 (ESSENTIAL, questions E1): FIX as the reviewer's primary** — the five why-item
  keys lose their ", so …" consequence clause (the only two-clause, longest option marked
  the key five times); the causal chain stays in the fb; the transfer items already perform
  the consequence link. The reviewer's five exact key wordings adopted verbatim. The
  all-options-same-frame alternative is noted for James, not taken (cut beats pad).
- **E-S02-2: FIX — "ethanoic acid" → "acetic acid"** (US/AP naming; 3 swaps + pool-ledger
  rename; accuracy A-1).
- **E-S02-3: FIX — item 2/W1 boundary leak**: "twice the volume of each solution" + the
  clearer fb(a) (accuracy A-2 = questions W1, found independently by both).
- **E-S02-4: FIX — drop "of every kind"** from L03's warmer-faster line (accuracy A-3; the
  absolutes-soften rule).
- **E-S02-5: FIX — L07 top-up clause** → "stirs concentrated hydrogen peroxide into the
  flask, raising its concentration" (accuracy A-5; keeps the example run's one-feature
  discipline: speeds AND is consumed).
- **E-S02-6: FIX — milk image regenerated** (left glass overstates: set curds vs "smells
  sour" prose) through the image pack + qc-figure; prose untouched per S02-9 (accuracy A-4).
- **E-S02-7: FIX — W2** steel-wool beaker: add "The acid still covers the steel wool."
- **E-S02-8: FIX — W3/W4/W5** stems made self-contained (reviewer's restatements; S01-E10
  precedent).
- **E-S02-9: FIX — W6** item 11's fb(b)/fb(c) recomputed from the stem's actual error paths
  (reviewer's rewrites); **W7** "They collide…" → "The particles collide…"; **W8** fb(a)
  one-word fix.
- **E-S02-10: Every Leave row stands as left** (both reports; the deliberately-omitted lists
  bind the blind reader). The MnO₂-froth drama and the three image residual-softness rows
  stay on James's card list.

## E-S03a. Accuracy-round adjudications — U5-S03 (2026-10-07; fix round waits for the

questions pass)

Report: qa/REVIEW_ACCURACY_S03.md (0 Essential, 4 Worth fixing, 4 Leave; all 79 keys +
datasets recompute; ten orders confirmed against literature; photo passes S03-5).

- **E-S03-1 — L06 k pair: ADOPT the reviewer's fix (1).** The comparison partner swaps from
  H₂O₂ to **tert-butyl chloride + water** (rate = k[(CH₃)₃CCl], k = 1.5 × 10⁻³ s⁻¹ at 25 °C
  — inside the real range in water-rich solvents; an S01 TAUGHT instance with no printed
  rate, so no one-scene clash; S03-2 licenses taught-side reuse). Every number and the
  "three times" sentence survive; edits are the species name, "reacts with water", "the two
  reactions". This makes the row literature-honest AND removes the 15% S01-flask misfit the
  reviewer sighted ahead of S04's integrated law. The James card updates to record the swap
  and the reasoning (his veto can restore the course-world H₂O₂ row with fix (2)'s scene
  line).
- **E-S03-2 — L16 "Most of the six pairs" → "Two of the six pairs": FIX.** False on the
  page's own table. The sentence sits in an approved blueprint scene — accuracy overrides
  the scene (A1 precedent); LOGGED FOR JAMES.
- **E-S03-3 — Overall-order definition: FIX** — drop "reactant" from KIN-029's definition
  lines; add one licence line where [H⁺] first enters an order item ("every concentration
  the rate law names counts — here that includes the catalyst's"). Truth over echo of the
  register statement (S05-3 precedent); LOGGED FOR JAMES/register keeper.
- **E-S03-4 — L01 recap MCQ pinned to the heated NO + H₂ flask: FIX** (an absolute assessed
  one lesson before its correction is a planted error; the pinned instance keys true
  forever). VAULT-SYNC CANDIDATES (close-out list): pin the instance in recall MCQs that
  restate earlier absolutes; count-check every "most/few" claim against its visible table.

## E-S03b. QA-round adjudications — U5-S03 (2026-10-07, questions pass + combined fix round)

Report: qa/REVIEW_QUESTIONS_S03.md (cold-solve 79/79; 4 Essentials, 8 Worth fixing, 5 Leave;
both ordered rulings delivered). The combined fix round = E-S03a rows 1–4 PLUS:

- **E-S03-5 (Q-E1, set-level key tells): FIX as specified** — reorder items 26 (key→A) and
  34 (key→B); swap item 2's table rows 2↔3 (key→A, fb remapped); trim item 24's key option.
- **E-S03-6 (Q-E2, alt-text leaks): FIX** — items 63 and 74 take the record's plain-cell
  alt rewrites BEFORE any blind bundle is generated.
- **E-S03-7 (Q-E3, imaginary exam paper in item 28's stem): FIX** with the record's rewrite
  (the Class-3 fault; key No untouched).
- **E-S03-8 (Q-E4): FIX** the two M·s⁻¹ unit-distractor fb lines ("gave k the rate's own
  units" — item 68's correct naming copied).
- **E-S03-9 (W1–W8): ALL ADOPTED** — item 24 re-instanced to cyclopropane; item 77 gains its
  own table AND reorders (key→C, W7); item 78 gains the H1 restatement (the chain's honest
  summing step); items 6/8 self-contained; 21/33 pronoun options named; 72's hedged map
  clause cut; 58 re-asked in the L04 frame; 44 gains "with the temperature steady".
- **E-S03-10: Leaves stand** (five rows; both deliberately-omitted lists bind the blind
  reader). **GATE REQUIREMENT recorded for U5-G1:** the gate's rate-law archetype item must
  use a species whose rate law S03 never printed (the printed-pool trade's one casualty).

## E-S03c. Review-lesson round adjudications — U5-S03 (2026-10-07)

Report: qa/REVIEW_LESSON_S03.md (Role A 22/22; F1–F9). Rulings: **F1–F4, F6–F9 ADOPTED**
verbatim (F6 takes the inline rate-law form; the optional eqstack stays parked for James).
**F5 ADOPTED** — part title → "Six pairs, two needed" (accuracy beats the approved scene's
title, E-S03-2's own precedent; logged for James). F8/F9 touch fb only → items 3 and 27 get
the S02 §6-style cold re-read after application. VAULT-SYNC CANDIDATES (close-out list):
bare-"run" imperative added to the lexicon pattern; count-check extended to lesson-distance
claims and part titles.

## E-S04. QA-round adjudications — U5-S04 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S04.md (0 Essential, 3 Worth fixing, 4 Leave; 70 keys + 47
figures recomputed clean; 11-species order table confirmed); qa/REVIEW_QUESTIONS_S04.md
(cold-solve 70/70; 2 Essentials, both wrong-map mismatches; 8 Worth fixing). One fix round:

- **E-S04-1 (ESSENTIAL, both reviewers on item 43):** wrong-map "0.2" → value 0.13 with its
  stated inversion fb, PLUS a new "0.2" map naming the first-order-style halving error (the
  fuller of the two convergent fixes).
- **E-S04-2 (ESSENTIAL, item 34):** the "0.0002" map is replaced by 0.0000032 with
  "dividing k by the time" fb (no computable path gives 2×10⁻⁴).
- **E-S04-3:** L22/L23 — DELETE the "at a higher temperature" clauses (the course must not
  show temperature changing k by nothing; accuracy W1), AND re-instance ONE L23 item with
  the questions reviewer's suggested fresh case (kills the L22→L23 numeric round-trip, W8).
- **E-S04-4:** L19 headache line takes the "often… rather than" hedge (accuracy W3).
- **E-S04-5:** item 3 fb "stand for" rephrased (live S51 hit). ENGINE NOTE owed: verify the
  validator's S51 sweep reads mcq fb (the reviewer believes it does not) — next OPS task.
- **E-S04-6:** L11 order-ID triplet re-instanced to NOBr/N₂O/azomethane order (kills the
  L04-echoed key string; counts unchanged; reviewer's full string adopted).
- **E-S04-7:** item 7 stem names TIME as what the options compare.
- **E-S04-8:** item 63's order-signature riders cut to bare orders (test-wise matching).
- **E-S04-9:** item 47's circular map fb and item 1's map-0.16 fb recomputed from the
  likelier named paths (reviewer's rewrites).
- **E-S04-10:** L27 capstone re-instanced HI → NOBr (blueprint said azomethane or NOBr; the
  third 0.0050-on-HI keying falls away; keys and maps recomputed from the one NOBr dataset).
- **E-S04-11:** every Leave row stands (both reports; deliberately-omitted lists bind the
  blind reader). The L25 photo-alt "brushes" two-word fix IS taken (the block is being
  touched this round anyway).

## E-S04b. Review-lesson round adjudications — U5-S04 (2026-10-07)

Report: qa/REVIEW_LESSON_S04.md (Role A 14/14; F1–F12). Rulings:

- **F1/F2 ADOPTED:** L06 and L13's headache questions get their graph ON SCREEN — the
  part-2 linplot inserted as the headache figure with points/rise-run layers OFF (the
  impossibility/need stays the headache; S04-11 scenes otherwise untouched; James may veto).
- **F3–F5, F6, F7, F12 ADOPTED** (exact rewrites from the record).
- **F8 ADOPTED:** the seven ≤30-word prose trims; contract opener kept (standing card).
- **F9: LEAVE** the six numeric-route maps (item 36's own reasoning governs).
- **F10: LEAVE, LICENSED** — halving fractions ((1/2)², 1/4, 1/16) are VALUES in worked
  arithmetic, inside the vault rule's own carve-out; on James's card as the convention call.
- **F11 ADOPTED as a RETITLE** — L10's title avoids the raw expression (sibling-consistent
  wording, e.g. "Sketching the reciprocal concentration plot"); engine keeps skipping titles.
- **BUNDLED (§S03-S execution):** S04's N₂O₅ dataset lines gain "at 65 °C"; S04's NO₂
  dataset gains "at 300 °C" (one-running-scene rule; pedagogic values, named regimes).
- Touched fb/stem items get a fresh cold re-read (S02 §6 pattern).

## E-S05. QA-round adjudications — U5-S05 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S05.md (1 Essential, 2 Worth fixing; keys 28/28; S05-3 + S05-A1
eyes PASS); qa/REVIEW_QUESTIONS_S05.md (cold-solve 28/28 — C01's cold status disclosed as
compromised, re-covered below; 1 Essential; C28 genre-tell ruling). One fix round:

- **E-S05-1 (acc. ESSENTIAL): FIX** — L06's every-collision-reacts line → "The reaction
  happens through those collisions, so doubling [NO] doubles the rate."
- **E-S05-2 (q. E1): FIX** — C12's termolecular-distractor fb takes the reviewer's rewrite
  (the named error must produce the option's count: CH₃ + H split, two particles stated).
- **E-S05-3 (C28/C26 genre tell): ADOPT the reviewer's set-level fix** — C26 (the
  named-elementary 2 HI twin) gains "The rate law cannot be stated from the equation alone"
  as a FOURTH, WRONG option with the supplied fb (the inverse of KIN-027's copying); C28
  stays exactly as written (exam-faithful). The genre is disarmed set-wide, not per-item.
- **E-S05-4 (q. W2): FIX** — C01's key rewritten to one clause ("A single collision between
  one Cl atom and one O₃ molecule forms the products."). Key-text change + C01's compromised
  cold status ⇒ C01 is named for the blind reader's explicit attention.
- **E-S05-5 (q. W3): FIX** — C15's correct-fb trimmed so it no longer hands C16 its key
  verbatim.
- **E-S05-6 (acc. W1/W2): FIX** — "Initial-rate experiments only" → experiments-only wording
  (comparetable cell + MCQ6 fb); "A few elementary reactions" → "Some".
- **E-S05-7 (minors): M2 ADOPTED** (both C15 arms carry "at one instant"); **M3/M4
  declarations added** to b4Reuse (L06's Cl+CH₄/2 HI reuse; L01's fresh Cl+O₃, Br+H₂,
  4 Fe+3 O₂); **M1 LEFT** (C26/C28 contrast is the blueprint's design; James's call noted).
- **E-S05-8: Leaves stand**; both omitted lists bind the blind reader. Register note for
  James carried (KIN-067 "collide" vs its own unimolecular bracket — spec's "react" form
  recommended upstream).

## E-S05b. Review-lesson round adjudications — U5-S05 (2026-10-07)

Report: qa/REVIEW_LESSON_S05.md (Role A 5/5; F1–F7). **F1–F6 ADOPTED** verbatim (F1/F2 keep
the S05-4 boundary pair's object/frame/crossings — noted for James's veto; F6's eight fb
trims trigger the §6-style cold fb re-read). **F7 LEAVE** (C12's 36-word fb is E-S05-2's
adopted wording — optional trim on James's card).

## E-S06. QA-round adjudications — U5-S06 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S06.md (0 Essential, 3 Worth fixing; all 37 keys correct; the
M–B energy form BLESSED on the collision-flux identity — stronger ground than the author's;
magnitude lines verified) and qa/REVIEW_QUESTIONS_S06.md (cold-solve 37/37; 2 Essentials;
key-string ruling). One fix round:

- **E-S06-1 (q. E1): C09 re-instanced onto N₂ + O₂** (the reviewer's Fix A — pool-honest;
  the energy-exactly-equal-to-Ea read keys Yes); changed item → fresh blind read (the full
  S06 blind read runs post-fix anyway).
- **E-S06-2 (q. E2): C30's image-option panels re-parameterised** so the correct render
  takes tag J (ok=0, fb reordered) — the unit's which-sketch family becomes K, L, K, J and
  the never-first-panel habit dies. FORWARD NOTE for S07's image-option item: pick its
  position against this family string.
- **E-S06-3 (q. W1): L03's check run reorders to A A B B** (reviewer's full string) — the
  three-lessons-open-on-No habit dies at its third repeat; L01/L02 stand as narrative-forced.
- **E-S06-4 (q. W2/W3/W4 + polish): ADOPTED** — C31 key trimmed to the verbatim canonical
  sentence (fb keeps the full chain); C37 re-instanced onto CH₃CHO vapor with the factor-words
  key; C04/C14 wrong-side fb differentiated; C02's fb error clause reworded.
- **E-S06-5 (acc. W1/W2/W3): ADOPTED** — L06's two lines anchored to the real mixture ("In
  the real mixture, …"); L04 mcq3's fb overclaim cut at "which is all the requirement
  names."; the figlib_u5_mb.py line-34 DOCSTRING corrected to the drawn collision-flux form
  (comment-only, no render change).
- **E-S06-6 (§S06-S execution): the "collision model" coin line enters L01's forest** ("This
  picture is called the collision model.") **+ one term-recall item appended last to L01's
  run** (run string re-checked for tells after the append; register rev 1.6 already fences
  the term).
- **E-S06-7: Leaves stand**; both omitted lists bind the blind reader; the S08 forward watch
  item (NO₂Cl's first step carries one collision-energised line) is RECORDED for the S08
  spec wave; L03's drawn one-collision H₂+I₂ event stays (word-only fence, S06-A) — on
  James's cards.

## E-S06b. Review-lesson round adjudications — U5-S06 (2026-10-07)

Report: qa/REVIEW_LESSON_S06.md (Role A 8/8; F1–F5). **F1–F5 ADOPTED** verbatim (F1/F2
class-1 rewrites; F3/F4 line splits; F5's eight fb trims → §-style cold fb re-read after
application). The L04 cross-run letter pattern and the two content-carrying forest lines
stay LEAVE on James's cards, as the reviewer recommended.

## E-G12. Gate QA adjudications — U5-G1/G2 (2026-10-07)

Reports: qa/REVIEW_GATES_G1G2.md (cold-solve 13/13; accuracy 13/13 real systems; 0 Essential,
5 Worth fixing) and qa/BLIND_READ_G1G2.md (13/13, zero clashes; two Leave watch-items). One
fix round, all five adopted:

- **E-G12-1:** KIN-020's key loses its only-second-clause tell → "The rate increases,
  because oxygen molecules now collide with the acetylene more often."
- **E-G12-2:** KIN-005's stem names the construct → "…gives the reaction's initial rate —
  the instantaneous rate at the moment the reaction begins?"
- **E-G12-3:** KIN-042's stem → "From the graph, what is the order…" (the shape alone
  cannot split first from second order; the marked halvings do).
- **E-G12-4:** KIN-065's stem → "…during one experiment in a heated flask…" (C₂F₄
  dimerisation is not a 25 °C reaction; sources in the record).
- **E-G12-5 (template):** figlib_u5_curves.py `_fmt` prints true minus signs on computed
  tick labels (one line); every negative-axis instance re-rendered (S04's ln-plots share it).
- **E-G12-6:** the four text-changed gate items get a fresh cold read; the builder's
  mis-stated G1 key-string record line is noted (items themselves correct). Blind-read
  watch-items stay Leave.

## E-S07. QA-round adjudications — U5-S07 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S07.md (0 Essential; the 1 Confirmed error was the register's own
KIN-083 bracket → rev 1.7 EXECUTED by the coordinator; inset ruled chemically right; L12
tail-growth honest) and qa/REVIEW_QUESTIONS_S07.md (cold-solve 53/53; 1 Essential; 3 W; 2
polish). One fix round:

- **E-S07-1 (q. E1): C35's image-option flaws redistribute** so the correct render takes tag
  M (J ea-to-products / K deltae-to-peak / L products-wrong-side / M null; ok=3; fb
  reordered) — the unit family becomes K, L, K, J, M. Changed item → fresh blind read (the
  full S07 blind read runs post-fix).
- **E-S07-2 (q. W1): C42's key cut to one clause** ("Its cutoff sits farther right on the
  same energy distribution."); the chain stays in fb.
- **E-S07-3 (q. W2): C29's reactants level moves to e:10** (map 60=forward-Ea path, 70=
  TS-alone, 110/−30 as now) — the wrong-map collision dies; stem/key unchanged.
- **E-S07-4 (q. W3): C48 (the exam-meta trivia item) is CUT** — its subject is the exam, not
  the atom (items-do-the-atom); L11's run stays spread (1 0 0 1 2). LOGGED FOR JAMES.
- **E-S07-5 (polish ×2 ADOPTED):** L11 q1 fb "a run of concentrations" → "concentrations
  changing through time"; C27's fb "Neither end-to-end gap" → "That gap does not reach the
  peak…".
- **E-S07-6: Leaves stand**; the accuracy watch-word (KIN-089 fb "collide at similar rates"
  = an entailment of the hedge) is named for the blind reader; both omitted lists bind.

## E-S08. QA-round adjudications — U5-S08 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S08.md (0 Essential, 2 Worth fixing; ozone framing + EPA
magnitude BLESSED; 11 mechanisms verified; keys 33/33) and qa/REVIEW_QUESTIONS_S08.md
(cold-solve 33/33; 1 set-level Essential; 3+1 Worth fixing; both hardest-call rulings
delivered — L05's run-instance licence holds with L09 carrying the discrimination; the
shared OH distractor stays). One fix round:

- **E-S08-1 (q. E1, set-level positional tell): ADOPT the reviewer's exact param
  reassignment** — L05-q1 → [NOBr₂, NO, NOBr] ok 0; L07-q3 → [NO, NO₂, O₂] ok 1; L09-q4 →
  [O₃, O₂, OH] ok 2 (fb arrays reorder with options). The catalyst-middle/intermediate-first
  template dies; letters A5 B7 C7.
- **E-S08-2 (acc. W1): FIX** — "Repeating steps like these are how one chlorine atom can
  destroy more than one hundred thousand ozone molecules."
- **E-S08-3 (acc. W2): FIX** — the photo alt gains "standing for the invisible ozone layer".
- **E-S08-4 (q. W1–W4): ALL ADOPTED** — the S77-family fb rewrite; "the reacting mixture"
  stem; "which reacts as shown"; L08-q2's shape-parallel option set.
- **E-S08-5: Leaves stand**; both omitted lists bind the blind reader; the engine-reveal
  conditional note (recognition→recall ladders safe under one-at-a-time reveal) is RECORDED
  as a standing engine assumption.

## E-S08b. Review-lesson round adjudications — U5-S08 (2026-10-07)

Report: qa/REVIEW_LESSON_S08.md (Role A 8/8; F1–F7). **ALL SEVEN ADOPTED** verbatim (F-class
staging imperatives ×4; the L09 semicolon split; five fb trims ≤30 w; the "in this unit"
meta-drop). The fb trims get the §6-style cold re-read after application. The L04 compound
headache question stays under S08-13 (James's card).

## E-S09. QA-round adjudications — U5-S09 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S09.md (0 Essential, 1 Worth fixing; boundary HELD exactly; L05
wording ruled ONE STORY; register rev 1.8 EXECUTED — KIN-105 "measurable" → "settable") and
qa/REVIEW_QUESTIONS_S09.md (cold-solve 38/38; 1 set-level Essential; W1–W5; both-pass
capstone ruled FAIR). One fix round:

- **E-S09-1 (q. E1): the L04 q2/q3 swap** (openers/followers exchanged; strings become L04
  y n y, L08 unchanged — ONE lesson only, as the reviewer specified; sequences loc labels
  updated).
- **E-S09-2 (acc. W1): "at ordinary laboratory temperatures" → "at lower temperatures"** in
  the 4 idea-line spots (S08's own licensed phrase; accuracy beats the blueprint wording,
  A1 precedent).
- **E-S09-3 (q. W1/W2): the two key rewrites** (NOBr₂ equal-rates key; the bottleneck key
  shortened with the full sentence kept in fb).
- **E-S09-4 (q. W3): "An intermediate" short form ADOPTED** (post-coin short form; kills the
  length flag; the full coin stays in the teaching text).
- **E-S09-5 (q. W4): the two option reorders** (L08 q3, L07 q1 → keys land at a; 3-option
  spread a6/b7/c7).
- **E-S09-6 (q. W5): L05 q5's stem restates its mechanism as data** (stands alone).
- **E-S09-7: Leaves stand**; the L02 q3 strict-kinetics caveat and the KIN-105 hygiene are
  on James's cards; both omitted lists bind the blind reader.

## E-S09b. Blind-read clash ruling — U5-S09 KIN-102-item8 (2026-10-07)

The blind reader's one clash (37/38) is the strict-kinetics reading of the no-reverse-steps
guarantee — the EXACT simplification the accuracy pass recorded as Leave (CED EK 5.8.A.1's
own clause) and the questions pass bound. **RULING: the key stands** — the item tests the
lesson's taught condition, item 9 confirms the design by breaking both conditions, and the
steady-state refinement is excluded by the register. The strict-kinetics caveat remains ON
JAMES'S CARDS, now with the blind reader's independent arrival at it as supporting evidence
(his veto = reword item 8's stem to name the taught guarantee explicitly).

## E-S09c. Review-lesson round adjudications — U5-S09 (2026-10-07)

Report: qa/REVIEW_LESSON_S09.md (Role A 13/13; F1–F8). **ALL ADOPTED**: F1 staging drop; F2/F3
line splits; F4/F5 the two forward-preview lines CUT (S01's meta-class treatment); F6 part
title; F7 the eight fb trims (≤30 w, §-style re-read after); **F8 — CHANGED_IDS.md's E1 prose
rewritten so no key-derivable pattern survives** (the executed blind read was protocol-clean,
on record). The items-35–38 four-a run stays LEAVE (verdict-fixed positions; unit sign-off
re-count notes it). VAULT-SYNC CANDIDATE (close-out list): bundle-adjacent prose
(CHANGED_IDS, READMEs) never states key-derivable patterns — the mechanical letter-scan
cannot catch this class.

## E-S10. QA-round adjudications — U5-S10 (2026-10-07, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S10.md (0/0/0 — both ordered rulings PASS, "peak" blessed) and
qa/REVIEW_QUESTIONS_S10.md (cold-solve 20/20; 1 set-level Essential; W1–W7). One fix round,
all adopted:

- **E-S10-1 (q. E1): reaction G flips to the reverse disagree case** (30 / 100 / 60 / 110 /
  50 → key Step 1, climbs 70 vs 50, tallest peak step 2; fb rewritten with full units) — the
  always-step-2 RDS tell dies and all three shortcut routes land on distractors.
- **E-S10-2 (W1): the SEVEN "is drawn below" stems → "The drawing shows…"** (the reviewer
  verified seven, not the relayed two).
- **E-S10-3 (W2/W3): the three leaking fb clauses cut; the two reading-bearing alts
  neutralised** (path description / cut).
- **E-S10-4 (W4/W5): L01 q1's options recast in the one frame (key → c) + the L02 q1
  reorder** — letters a2/b5/c4, firsts c,c,a,a; residual b geometry noted for the unit
  sign-off count.
- **E-S10-5 (W6): the b4Reuse block added** (the supplied text; ICl+H₂ different-atom claim
  verified).
- **E-S10-6 (W7): L03's RDS fb arithmetic carries units.**
- **E-S10-7: BINDING CONSTRAINT FOR U5-G5 recorded — the KIN-112 gate sample keys tag M**
  (the family string K,L,K,J,M,L,J has spent the J margin; M is the only tag at 1).
- **E-S10-8: Leaves stand**; both omitted lists bind the blind reader.

## E-S10b. Review-lesson round adjudications — U5-S10 (2026-10-07)

Report: qa/REVIEW_LESSON_S10.md (Role A 10/10; F1–F4; figures record exists — the gate line
was stale). **ALL ADOPTED**: F1 split; F2's twelve fb trims (item17's adjudicated 43-word fb
LEAVE, on James's card); **F3 — the five key-derivable passages in CHANGED_IDS.md scrubbed**
(the E-S09c class, second instance — the vault-sync candidate hardens: bundle-adjacent prose
never prints family strings, run-break notes, or omitted-list lines that derive keys);
**F4 — "Label both axes," enters BOTH draw stems** (S10 item19 AND the identical S07 L08
twin — same house-pattern mismatch, same two-word fix; S07's CHANGED_IDS notes it). The fb
trims get the §-style cold re-read.

## E-S11. QA-round adjudications — U5-S11 (2026-10-08, after the accuracy + question passes)

Reports: qa/REVIEW_ACCURACY_S11.md (0 Essential, 5 Worth fixing; iodide-mode DEFENSIBLE;
propanone why-key Confirmed-at-grain) and qa/REVIEW_QUESTIONS_S11.md (cold-solve 40/40;
3 Essentials; W1–W9). One fix round:

- **E-S11-1 (q. E1): ONE why-item reorders** (item 39 → key a preferred; the fix agent may
  pick 7 or 17 instead, no new ≥3 run/alternation, letters re-counted).
- **E-S11-2 (q. E2): the key-tells dispositions ADOPTED** — one-line trims on items 2, 4, 7,
  17, 40; item 15's set recast in one frame; item 12's content-driven key DECLARED in
  b4Reuse; the tool re-run to a clean pasted KEY TELLS line in the fix report.
- **E-S11-3 (q. E3): THE FAMILY-STRING CORRECTION + G5 RULING.** The true string is
  K,L,K,J,M,J (+S11's M = K,L,K,J,M,J,M). S11 item 14 stays M (compliant). **E-S10-7 is
  AMENDED: the U5-G5 KIN-112 sample keys L** → final family K,L,K,J,M,J,M,L (2-2-2-2). The
  two stale record strings (S10's CHANGED_IDS already scrubbed; S11's b4Reuse sentence)
  are corrected by the fix agent.
- **E-S11-4 (acc. rows, all adopted):** the propanone why-item GENERICISED to the Suppose-
  case (rate = k[R][H⁺]; key wording/index unchanged; fb(a) recomputed — the autocatalysis
  fb falsehood dies with it); converter "metal" → "ceramic" (prose + alt; the headache
  wording stays James's per S11-7 — the one-word accuracy fix is within the A1 precedent);
  L10's second catalysed peak 80 → 70 (climbs now order with the slow/fast labels; renderer
  asserts recomputed); the L09 mode-table cell → "the catalyst itself binds the reactant";
  ethyl ethanoate/propanone → ethyl acetate/acetone everywhere student-facing (E-S02-2).
- **E-S11-5 (q. W1–W9): ALL ADOPTED** — the two frame-question inserts; the five drawn-below
  stems + one shown-below line (third sequence with the class); item 11's alt neutralised;
  option d's full mode name; "Label both axes," into the draw stem; item 40's stem noun;
  "marks"; item 30's "at the same temperature"; item 23's fb trim.
- **E-S11-6: Leaves stand** (incl. the IO⁻ phrase Simplification + the KIN-121 register
  scope-note recommendation, BOTH on James's cards); omitted lists bind the blind reader.

## E-S11b. Review-lesson round adjudications — U5-S11 (2026-10-08)

Report: qa/REVIEW_LESSON_S11.md (Role A 16/16; F1–F3). **ALL ADOPTED**: F1 the colliding-
species rewrite; F2 four fb trims (§-style re-read after); F3 the L08 coin un-led (explain-
then-label; the item-31 sentence reused). The stale letter-count record line is noted, items
right; the two pre-existing alternating runs stay LEAVE for the unit sign-off re-count; the
species letter Y noted to the figures pass.

## E-G3. Gate QA adjudications — U5-G3 (2026-10-07)

Reports: qa/REVIEW_GATES_G3.md (cold-solve 8/8; accuracy all confirmed; 1 Essential — the
COORDINATOR'S bundler leaked the `assert` block into the blind bundle, voiding KIN-085's
blind verdict; 2 template-level W) and qa/BLIND_READ_G3.md (8/8 on the flawed bundle; one
wording flag). Rulings:

- **E-G3-1 (ESSENTIAL, bundler): asserts are BUILD METADATA, never bundle content.** The
  bundler strips `assert`/`eaForward`/`eaReverse`/`deltaE` keys and the leak-grep gains them;
  the G3 bundle regenerates; **KIN-085 gets a fresh single-item cold read** from a new
  reader. Standing bundler rule for every future bundle.
- **E-G3-2 (W1): "is drawn below" is FALSE of the rendered page** (figures render above
  stems) — the gate stem takes "The drawing shows…"; the SAME frame in ten closed U5-S07
  lesson stems is fixed in the same round (stem wording only, keys untouched; closed-not-
  blessed content takes rules in force; their CHANGED_IDS note it).
- **E-G3-3 (W2): figlib_u5_profiles' drawer prints true minus on negative ticks** (the
  G1/G2 W5 fix reached figlib_u5_curves only); every profiles instance re-rendered (S07
  shares the drawer).
- **E-G3-4 (blind flag): KIN-076's stem verb → "shows, qualitatively,"** (a feature shows;
  a reader estimates — the blind reader's point adopted).
- **E-G3-5: Leaves stand** (panel-K label; "you drew"; the termolecular key's length).

## E-PRAC. Practice QA adjudications + the G4/P3 collision (2026-10-08)

Report: qa/REVIEW_PRACTICE.md (cold-solve 15/15; 2024 Q6 fidelity faithful with two drawing
departures; 1 ESSENTIAL). Rulings:

- **E-PRAC-1 (ESSENTIAL): the G4 KIN-100 item re-instances; P3 KEEPS Ce/Tl/Mn** (the
  5-point production FRQ needs the canonical two-intermediate cycle; the gate item's
  classification ask works on any real cycle). CONSTRAINT the reviewer's suggestion missed:
  Cl-catalysed ozone is TAUGHT (S08-3) — barred. The fix agent finds a REAL catalytic cycle
  fresh across ALL specs + gates + practice + _assessment_prep (candidate direction: the
  documented iodide-catalysed Ce⁴⁺ + As(III) reaction, or another verifiable classic),
  verifies the chemistry, keeps the Ce⁴⁺-style appears-twice trap class if honest. The
  changed gate item gets a FRESH blind read; the practice record's stale grep line re-run
  and re-dated.
- **E-PRAC-2 (W1): linpanels gains a released-scales mode** (numeric ticks + axis units)
  used by U5-P2 so the sighted student sees what the alt says (exam fidelity for a REAL
  question's stimulus; the qualitative family stays default elsewhere).
- **E-PRAC-3 (W2): the NO₂ Lewis renders FLAT (O–N–O in a row) as the released exam draws
  it**; the shape word leaves the alt; the departure list updates (now zero undeclared).
- **E-PRAC-4 (W3): the (c)(i) rubric drops "linear"** (the part asks bonds + lone pairs).
- **E-PRAC-5: Leaves stand** (the ~120° released framing; the rubric inversion; phone wrap).

## E-G5. Gate QA adjudications — U5-G5 (2026-10-08)

Reports: qa/REVIEW_GATES_G5.md (cold-solve 10/10; 2 Essentials; constant-pH ruled honest)
and qa/BLIND_READ_G5.md (10/10). Rulings:

- **E-G5-1 (E1): KIN-065's table takes the short headers** ([peroxide] (M) / ln[peroxide] /
  1/[peroxide] (M⁻¹)) so all four columns sit inside the 408 px wrapper; PLUS the engine
  gains the reviewer's assertion (a gate/test table drawer refuses a natural width over the
  480 px wrapper). Re-render; bundle regen; fresh blind read of the changed item.
- **E-G5-2 (E2): LEDGER-ONLY fix ADOPTED** — the which-sketch family truly has NINE members
  (G3's KIN-087 gate item was real; E-S11-3's "no seventh item exists" was the error):
  **K,L,K,J,M,L,J,M,L = K2 L3 J2 M2 (33% < 35%, no adjacent repeats)**. E-S11-3 is hereby
  CORRECTED on the record (this entry amends it; records are amended, never rewritten);
  gates.json's _skill "2-2-2-2" line and BUILD_STATE's stale lines are fixed. NO RE-KEY:
  all three re-key alternatives cost more than the 3-of-9 they cure (40% letter-share /
  3-run / adjacent both-M) — the three costs sit on James's card with his veto.
- **E-G5-3 (W1–W4 ADOPTED):** the KIN-115 alt neutralised (gridline-data form); KIN-079's
  key trimmed to the cause clause; KIN-121 re-scened to lactase drops/lactose-free milk
  (genuinely dissolved — the honest fix; key untouched); KIN-122's mechanism moved to
  display-line equations (KIN-104's cure).
- **E-G5-4: Leaves stand**; changed items (KIN-065, 079, 112-adjacent none, 115-alt, 121,
  122) get re-render + bundle regen + a fresh blind read.

## E-T1. Unit-test QA adjudications — U5-T1 (2026-10-08)

Reports: qa/REVIEW_TEST_T1.md (examiner: cold-solve 20/20, 0 Essential, 3 W; topic purity
exact; 2022 Q5 DATA EXACT/stem not blessed — the released-paper check stays a James card)
and qa/BLIND_READ_T1.md (18/18 + 4/4 + 10/10; the T1-17 convention flag). Rulings:

- **E-T1-1 (ex. W1): "fast and reversible"** (one-name rule; the exam's own form).
- **E-T1-2 (ex. W2): the M4 ↔ FRQ-2(a) echo FIX ADOPTED** — experiment 3's rate →
  3.2 × 10⁻⁴ (order 2 in Y; key → b; law k[X][Y]², still never-printed; fb remapped; string
  a3 b5 c5 d5) — a test must not pre-walk its own FRQ's rubric points mid-test. Fresh cold
  read of M4 after.
- **E-T1-3 (ex. W3): the record's ethane clearance line amended** (conclusion stood).
- **E-T1-4 (blind T1-17): the FIGURE FIX ADOPTED — peak 2 lowers to 75 (step-2 Ea 55)** so
  the taught own-valley convention and the rigorous effective-barrier reading AGREE on
  step 1; the trap option (tallest absolute peak) still exists as the misconception path
  where it now points at step 1's peak... NO — recompute: with peaks 80/75, step 1 is both
  tallest AND largest-climb; the highest-peak reader now lands RIGHT for the wrong reason.
  AMENDED RULING: keep the item's discrimination by lowering peak 2 to 75 AND checking the
  distractor set still carries a computed per-step trap value (the fix agent recomputes and
  reports; if the item loses its discrimination entirely, the alternative is James's
  convention adjudication card — implement the geometry fix, verify by recompute, and state
  plainly which property survives).
- **E-T1-5: the engine fb-reveal question** (withhold feedback until test submission closes
  every mid-test method-leak channel; U4-T1 shipped the same behaviour) — JAMES'S CARD, not
  fixed here.
- **E-T1-6: every Leave stands**; the tangent-convention exam-form note joins James's cards.

## E-T1b. M4 coefficient-coincidence fix (coordinator, 2026-10-08)

The E-T1-2 fix left M4's orders (1,2) matching the stem's X + 2 Y coefficients — the
KIN-038 documented misconception (copy the coefficients) landed on the key. The system is
GENERIC, so the declared stoichiometry is ours: the equation becomes **X(g) + Y(g) →
products** (coefficients 1,1 ≠ orders 1,2); the coefficient-copier now computes k[X][Y] and
lands on wrong option (a). Data, options, key, fb untouched. Applied by the coordinator;
M4's fresh cold read covers it.