Not yet run. The bundle at blind/2026-10-08-reviewed/ (112 MCQ, 4 typed, 1 FRQ, 1 multi-select; 118 held, 118 bundled) and its CHANGED_IDS.md (ids only) are ready for two cold readers; the eight recall-bank items are not bundled (held per the 8 Oct ruling until the course wires them in).

The science is right and the seed is kept observable, the reviewer found: the register's belief voiced once by Hana and answered the 7 October way; no banned word in any student line; every structure-job pair, migration, hibernation and instinct / learned sort true of the organism named. Six must-fixes (the rule cards; the L02 video standing alone — the split; "individual" coined; the practice stem answering itself; the chipmunk; the vole table) and fourteen smaller fixes, all applied — above. One accuracy class worth noting for later topics: a "year later" on an animal that lives a season turns an observable statement into one about offspring. Left as built: "pull through", the fox coat line, the contrast pairs, the typed recalled words. The full table with every check listed clear or acted: REVIEW_G5T29_2026-10-08.md.

All 31 renders opened at 2× and the built page read on every article and quiz route, the practice, the bank, the summary and the intro. The four rule cards for the video rule slides ran their first line past both edges of the frame (Topic 18's drawer prints one line per string); the topic now wraps lines by measured width and asserts each inside the card, L02's card says the lexicon's rule word for word, and L02b has its own. The behaviors table lost its "born knowing / learned" column and bear-cub row to L02b's new table. The six fox cards (two tall bumpy bands, four low bands, words not letters), the March ticks and crosses under their printed question, the timeline strip's counts, the ears strip and the two bank tables are right on the render; no placeholder's note names an item's answer; no id is written. Three new placeholders (Gus at the Montana campground; sea turtle hatchlings; a young orca) carry briefs for the picture pass.

53 items in content/G5T29_pp100_bank.py: 48 four-option MCQs, four typed words (instinct twice, migrate, huddle) and one two-of-four select. Per atom: LIF5-040 13 · LIF5-034 13 · LIF5-041 14 · LIF5-042 13. Easy 23, medium 17, hard 13, sorted easy to hard and rising with the atom's place. MCQ keys A 15 · B 16 · C 10 · D 7. Eighteen types on the type sheet, one exemplar each; the review's changes are its v2 note (a new "individual" term item; the night-hunting coyote and Farah's mice re-scened; the vole table re-asked; a key's "and" clause cut; two retags; two ⟂ pairs). Every bank claim is "X is wrong. What is the correct reason?" with four single-statement reasons; no option bundles a verdict with its reason.

Skill: review-lesson @ cca3af6 (Role B guidance review and Role C fix pass; Role A is the two cold readers' job after this) · review-questions @ cca3af6 · review-figures @ cca3af6 · review-accuracy @ cca3af6

Grade 5 Topic 29 — Who pulls through when a place changes — non-author review and fix pass (8 October 2026)

§1 The takeaway, in plain words

The topic teaches the natural-selection seed the way the register asks: the fox versions were there in November, the cold changed no coat, it only decided who pulled through; "natural selection", "evolve", "adapt" and anything about offspring are absent from every student line, and Hana's claim is the register's own SD-5 belief, answered the 7 October way. The science of the saguaro, the Gila monster, the jackrabbit, the kit fox, the geese, the groundhog, the quail and the bear cub is right. Six things had to change before a child meets it, and they are changed: (1) every rule card for the video rule slides ran its first line off both edges of the frame; (2) the L02 video never said what an instinct or a learned behavior is, and four of its quiz items tested exactly that, so a video-only student would have met them cold — L02 is now two lessons, L02 (the four behaviors) and L02b (born knowing or learned, both raise the chance), each video standing alone and each article under 400 words; (3) "individual" arrived bald in L03's rule (the author glossed it as Topic 28's word, but Topic 28 does not coin it), so L03 now coins it in the frame and the quiz, practice and bank recognise it; (4) a practice item gave its own answer in the stem; (5) the summary's chipmunk "living on stored fat" is wrong biology (chipmunks store food and wake to eat it) — it is a ground squirrel now; (6) a bank table item was answerable with the table covered. Fourteen smaller things are fixed too (below); the three long articles are 353, 397 and 395 words. Every gate is at its stated result; the blind bundle is re-cut at blind/2026-10-08-reviewed/ with CHANGED_IDS.md; the batch (now seven videos) is rebuilt and checked, not sent. One question for James, in card shape, in REVIEW_NEXT_G5T29.md: whether "function", TEKS 5.13A's word, should be said once beside "job".

§2 Role A

Not run here (two separate cold readers follow). The bundler's count line: bundled: 112 mcq, 4 text (0 numeric), 1 frq, 0 order, 1 multi — adversarial rows: 4 — held: 118 — every held item bundled. Note for the coordinator: the eight RECALL_BANK items are not bundled, not on the page and not read by the checker (the brief's ruling 11 holds them for later wiring), so they have had no blind read.

§3 Findings, by verdict (the Role B, questions, figures and accuracy tables, merged; every row's disposition)

Every other check in review-lesson B1–B7 ran clear: one atom per lesson; forest-trees-forest in every article; no staging imperative in any article (S36 and the eye); every scene noun introduced before it acts; plain word before term; no figure referred to that is not on the screen; fb ≤ 30 and why ≤ 60 words with one link per line (S5 silent); no meta language; the rendered glossary prints definitions only.

§3b Figures (review-figures; every one of the 31 renders opened at 2×, and the built page on every article and quiz route, plus practice, PP100, summary and intro)

The page forces one colour scheme, so it is shot once. No id is written by the module (check_svg_ids.py PASS).

§3c Accuracy (review-accuracy; every named organism and claim)

Deliberately omitted (the register's scope): the survivors' offspring and inheritance of the version; "natural selection", "adaptation", "evolution"; day length as the trigger of the fox's molt; cold-induced fur thickening in some mammals.

§4 Items changed (ids in blind/2026-10-08-reviewed/CHANGED_IDS.md; no key text there)

§5 Gates after the fixes

Key strings after the pass: quizzes two-option L01 A A B B A · L02 A A B · L02b A B B A A · L03 A B A B B · L04 B A A B; three-option L01 0 2 1 1 · L02 0 2 1 2 0 1 · L02b 1 0 2 · L03 0 2 1 1 · L04 0 2 2 0; practice A A A and 0 2 1 0 1 2 1 1 2 0; bank MCQ keys A 15 · B 16 · C 10 · D 7 (48 MCQs); bank 53 items, per atom 13 / 13 / 14 / 13, easy 23 · medium 17 · hard 13; the claim string in the quizzes is one No (L04-03).

§6 For the coordinator (general findings, not applied outside this topic)

• Topic 18's rule_card prints each passed string on one line with no width guard; every Grade 4 and 5 topic that imports it and passes a whole rule sentence has clipped rule slides. This topic now wraps by measured width; the drawer itself should gain the guard (vault review-figures check 17).
• RECALL_BANK items are neither bundled nor rendered nor swept; when ruling 11's wiring lands, they need a blind read.
• Topic 28 does not coin "individual"; if a later Life topic recalls it, it should recall Topic 29's coining.

Skill: course-build → write-lesson (write-example-sequence, design-check-sequence, sentence-atomise, write-explanation) → write-video-script, write-intro, write-summary, write-mcqs @ cca3af6

Grade 5 Topic 29 — Who pulls through when a place changes — build record (v1, 8 October 2026)

What this is, in plain words

Four short lessons that carry a Grade 5 child from "every body part does a job" (Grade 3) to the seed of natural selection, said in observable words only. First, in one place different kinds survive with different structures (the saguaro's thick stem, the Gila monster's fat tail, the jackrabbit's big ears, all on one desert trail). Second, the things an animal DOES can raise its chance of survival (hunting at night, migrating, hibernating, huddling; an instinct or a learned behavior). Third, the individuals of one kind are not all the same, and when their place changes, the ones whose version of a feature suits the changed place are more likely to pull through (six Minnesota foxes, two thick-furred, in the coldest winter in fifty years). Fourth, WHY those individuals pull through: the version that suited the change was already there before it; the change gave no fox a new coat, it only decided who pulled through — meeting, by name, the register's belief that individuals change their own features to fit a new place. The words "natural selection", "evolve", "adapt" and any mechanism (offspring, passing on) are never said. Every lesson has a video script and an article, a quiz of eight or nine items and one attached question; the topic has an intro, a recall bank of eight earlier-topic items, a practice set of thirteen, a summary with three attached questions and a PP100 bank of fifty-two items with its type sheet. Nothing has been sent, committed or published.

Counts

• Lessons 4 (L01 LIF5-040, L02 LIF5-034, L03 LIF5-041, L04 LIF5-042 the capstone). Scripts 149 / 148 / 149 / 149 spoken words (the batch builder's own count: 145 / 144 / 142 / 144). Articles 376 / 434 / 423 / 425 words (20 / 24 / 23 / 22 idea lines; 6 / 7 / 5 / 3 figures; 6 / 6 / 6 / 6 subtitles — a subtitle or figure every ≤ 4 idea lines). Intro 117 words, 2 question marks. Summary 168 words.
• Quiz items 9 / 9 / 8 / 8 (34; no typed item: the one coined term is nine letters, RECALL_BY_MCQ). Attached questions 7 (four lessons, three summary). Recall bank 8. Practice 13 (3 Grade 3 / Grade 4 retrievals, 1 FRQ). PP100 52: 47 four-option MCQs, 4 typed, 1 multi-select; per atom 13 / 13 / 13 / 13; easy 20, medium 19, hard 13.
• Figures 26: 13 SVG by code, 9 photo placeholders, 4 cartoon placeholders.

Structural decisions (in place of a blueprint)

Lesson order and titles. L01 "Same place, different structures: how each kind survives there" (LIF5-040) · L02 "Behaviors that raise an animal's chance of survival" (LIF5-034) · L03 "Predict who pulls through when the place changes" (LIF5-041) · L04 "Why those individuals pull through" (LIF5-042, the capstone). The architecture row's order: structures, then behaviors, then the what of pulling through, then its why.

Headache scenes and who asks. Every opener is the scene with the NARRATOR asking its question; one named child per scene (8 Oct ruling), and no child voices a line except Hana in L04, whose claim IS the register's documented belief. L01: Farah on a desert trail near Tucson at noon in June, a saguaro, a Gila monster and a jackrabbit all living in the same hot, dry place — "How does each one survive here?" L02: Gus and his dad at a desert campsite after dark, a kit fox trotting through the flashlight beam after a day in its den — "Hunting at night is something the fox does. How does doing that help the fox survive?" L03: Hana at her grandma's farm in Minnesota in November, six red foxes, two with thick winter fur and four with thin, the coldest winter in fifty years forecast — "Which foxes are more likely to pull through?" L04: Hana at the same fence in March, three foxes left; Hana says "The foxes grew thicker fur because the winter turned so cold." — no question is asked of it; the article and script recap the claim and contradict it after the evidence (the 7 Oct correction). Openers vary: a noon trail, a night camp, a farm fence in two seasons.

Example sequences, written out.

• L01 (example-sequence classification, the atom judges cases): frame "Does this structure help the kind survive here?"; the place holds (the desert) while the kind changes one step at a time — the saguaro's thick stem (yes: stores water) → the Gila monster's fat tail (yes: stores food) → the jackrabbit's big thin ears (yes: heat the air, so the body stays cool) → the SAME jackrabbit's SAME ears in deep snow in Alaska (no: the ears heat the freezing air so fast the body gets cold; the one feature changed is the place) → back in the desert (yes). p p p n p; the n→p return lives in the article only (the script carries the three positives and the coining). "Structure" coined explain-then-label after the three cases: "A part of a living thing's body, like a stem, a tail or an ear, is called a structure." Rule last.
• L02 (fact form, examples only — no honest near miss exists for "a behavior that raises the chance of survival" that a child could believe): the one frame "How does this behavior raise the animal's chance of survival?" answered for the kit fox hunting at night (keeps it out of the heat) → Canada geese migrating (takes them to open water and food) → the groundhog hibernating (lives on stored fat through a winter with no plants) → bobwhite quail huddling (keeps the group warm); then Grade 4's two sentences on instinct and learned behavior word for word, and the black bear cub that learns where its mother finds food — so the payoff is said of a learned behavior too (TEKS 5.13B names both). Rule last; the kit fox returns.
• L03 (prediction, a Transformation atom): the object holds and one thing moves a step — the six foxes and a much colder winter (the thick-furred foxes) → the same foxes and a much hotter summer (the change moves: the thin-furred foxes) → the same foxes, the same hot summer, a white tail tip or not (the difference moves: about the same — a tail tip keeps no fox cool or warm). Then one fresh kind outside the run: twelve cottontail rabbits, some faster, a coyote moves in (the faster rabbits). The observed March count (both thick-furred alive, one of four thin-furred alive) is stated as a count. Rule last.
• L04 (the why, meeting the belief): before the winter the two versions were already there (the November cards) → a living thing does not change its features by trying or wanting; a fox cannot grow a thicker coat because the air turns colder; each fox kept the coat it had → thick fur held in the body's warmth, so the thick-furred foxes stayed warm and three thin-furred foxes died → the winter gave no fox a new feature, it only decided who pulled through → "So, when Hana said the foxes grew thicker fur because the winter turned so cold, she was not right." → the rabbits and the coyote as the fresh kind (the coyote made no rabbit faster). Rule last.

Check types. L01 type (a) fluency (a part and its job; yes / no in its place; the term recognised at position 3 before any stem prints it; the boundary near miss; two kinds at one beach). L02 type (a) fluency (behavior → payoff; the behavior named; instinct or learned with the evidence in the stem; a learned behavior's payoff). L03 type (a) fluency on predictions (two or three options; "about the same" keyed once where the difference does not suit the change). L04 type (c) what / why / misconception / transfer, the misconception item on the register's SD-5 belief.

Videos. All four lessons get a video: L01's three structures and the swelling stem; L02's day-to-night den scene and the four behaviors; L03's snow falling over the cards and the ticks and crosses appearing; L04's cards holding their fur through the winter. None "could do without a video".

Recall bank (ruling 11). Eight items, none in a quiz: before L01 — LIF3-020 (every body part does a job), LIF4-002 (a plant part with the right shape); before L02 — LIF4-033 (instinct or learned behavior), LIF3-062 (migrate / hibernate); before L03 — LIF3-041 (alike, but not the same), LIF5-032 (a different mix of instructions, Topic 28), LIF3-063 (the same change helps some and harms others); before L04 — LIF4-052 (a living thing does not change its features by trying or wanting). L04's own prerequisite is L03, the lesson immediately before, so no recall of it.

Terms. One coined term, "structure" (TEKS 5.13A's word; "function" is said as "job", Grade 3 and Grade 4's word — see judgement call 1). Recalled, never re-coined, each with the earlier lesson's sentence word for word on first reuse: behavior / instinct / learned behavior (Grade 4 Topic 36), migrate / hibernate (Grade 3 Topic 42), offspring (Grade 3 Topic 40), individual (Topic 28's word, glossed "each single fox, each individual"). "Pull through" is the title's plain phrase, explained in the lesson's own sentences ("lives through the winter"), not a coined term.

Questions for James

None. Every decision here is one his answer would not change what a student learns by (names, the title phrase, the choice of desert kinds, the fox counts), or applies a ruling he has already given (the natural-selection seed's observable-only scope is the register's; the belief met by name is the register's SD-5).

Per-lesson table

Design-check-sequence records

`

Skill: design-check-sequence @ cca3af6

Part: L01 (LIF5-040)

Type: (a) one-step categorisation — given a kind in a place and a job, the student picks the structure, or says whether a named structure helps there.

Plan: the prairie dog's claws (3 opt, 0) → the bison's thick coat in January (yes) → the term recognised (3 opt, 2) → the pronghorn's long legs (2 opt, 0) → do the three prairie kinds rely on one structure (no) → the prairie grass's deep roots (3 opt, 1) → the musk ox's coat in a Florida summer (no, the boundary) → which of these is a structure (3 opt, 1) → the pelican and the crab (yes). Two-option string A A B B A; three-option 0 2 1 1. The term is recognised before any stem prints it.

Belongs in practice/test instead: the four-structure pick; the four-sentence two-kinds item; the same-structure-elsewhere four-sentence item; the jobs set (PP types 1–4).

`

`

Skill: design-check-sequence @ cca3af6

Part: L02 (LIF5-034)

Type: (a) one-step — given a described behavior, the student names how it raises the chance of survival, names the behavior, or sorts it as instinct or learned on the stated evidence.

Plan: the Indiana bat (3 opt, 0) → the behavior named (3 opt, 2) → the Colorado hummingbirds (3 opt, 1) → the Michigan deer mice huddling (yes) → the Texas bobcat (3 opt, 2) → the Virginia fawn, instinct (2 opt, B) → the Washington otter, learned (2 opt, B) → the otter's payoff (3 opt, 0) → can a learned behavior raise the chance (yes). Two-option string A B B A; three-option 0 2 1 2 0. No claim item.

Belongs in practice/test instead: the four-outcome payoff; the four-behavior name set; the four-kind sort; typed instinct / migrate / huddle (PP types 5–8).

`

`

Skill: design-check-sequence @ cca3af6

Part: L03 (LIF5-041)

Type: (a) one-step prediction — a described difference and a change → which individuals are more likely to pull through.

Plan: Ohio sparrows' beaks and hard seeds (3 opt, 0) → Wisconsin deer's legs in deep snow (2 opt, A) → Texas lizards on soot-black rock (2 opt, B) → Idaho trout's spots in warmer water (3 opt, about the same, 2) → New Mexico cactus skin in a drought (2 opt, A) → Utah hares in a snowless winter (2 opt, B) → Vermont bees in a cold spring (3 opt, 1) → are all individuals equally likely (no). Two-option string A B A B B; three-option 0 2 1. No claim item (the belief is L04's).

Belongs in practice/test instead: the four-option who-pulls-through set with "none of them"; the same kind under two changes; the table read; which change suits a version; two differences weighed (PP types 9–12).

`

`

Skill: design-check-sequence @ cca3af6

Part: L04 (LIF5-042)

Type: (c) concept understanding — the why, with the register's documented belief (SD-5) addressed in the lesson.

Plan: WHAT: the Oregon gray moths before the bark turned dark (3 opt, 0) → WHY: why the gray moths pulled through (3 opt, 2) → MISCONCEPTION: Gus's claim that the moths turned gray because the bark turned dark (no) → WHAT: were the Maine snails' thick shells already there (yes) → TRANSFER WHY: why most snails left have thick shells (3 opt, 2) → WHAT: did each snail keep its shell (yes) → TRANSFER: the salty Florida lake's fish (3 opt, 0) → the sparrow trap as a plain item: did the small-beaked sparrows grow bigger beaks (no). Two-option string B A A B; three-option 0 2 2 0.

Belongs in practice/test instead: the FRQ why on a fresh yard (PR-10); the claim-by-reason form; the set item; the evidence item (PP types 13–18).

`

The key strings

• Attached questions in video order (L01–L04, summary 1–3): B A A B B A A — A 4 of 7.
• Lesson quizzes, two-option: L01 A A B B A · L02 A B B A · L03 A B A B B · L04 B A A B; three-option: L01 0 2 1 1 · L02 0 2 1 2 0 · L03 0 2 1 · L04 0 2 2 0. Practice two-option A A A, three-option 0 2 1 0 1 2 1 1 2. Recall bank two-option A B B B, three-option 0 1 2 0.
• PP100 MCQ keys: A 14 · B 16 · C 10 · D 7 (47 MCQs); the four typed items and the multi carry no letter. S7 silent; S116 silent after the order pass (two L04 keys and one practice key moved; PP-28..33 re-spread).
• Claims: one in the quizzes (L04-q3, Gus, keyed No; the register's SD-5 belief); the TRUE twin of this belief is the lesson's own rule, so no Yes-keyed claim exists — recorded. Bank claims: four, all "X says … X is wrong. What is the correct reason?" with four single-statement reasons (Farah once, Gus once, "a student" twice); James's 8 Oct four-option verdict-and-reason form is not used, so no exception is named.
• Order items: none. Lettered drawings: none (the fox cards are named by their fur; the tables by their row words).

The rules of 7 and 8 October, applied (rulings 1–13 of the brief)

Other rules applied

Misconceptions used, by source, and the beliefs deliberately not stated

Checker results, every WARN answered

• build_tools_check_topic.py → 0 FAIL, 56 WARN; S46-COUNTS B=0 C=0 D=0; S30-COUNTS typed=4 fail=0 warn=0 (two S30-INFO lines: the typed "instinct" items are closed by the author's note, Grade 4's term recalled in L02). The 56 WARN: 6 × S8 "beat carries 3–4 sentences" — kept: every beat is one scene state, as in the approved Grade 3 and Grade 4 scripts; 1 × S4 L04-q1 (46 words, 3 sentences) — kept: the what item needs the scene, the change and the outcome before its question, and it was cut from 53; 49 × S36 unusual words — salty, saguaro, Gila monster, burrow, kit (fox), pronghorn, pouch, trail, mussels, voles, waxy, salamanders, student, musk (ox), quail, hooves, chickadees, trots, fawn, soot, drought, split, cling, bumpy, wolves, walrus, talons, spiky, pika, shrunk, command, forecast, perfect, furry, salmon, drier, monarch, bushy, tusks, bald, guinea, antlers, mate, shiver, trash, ledge — all kept: each is a US animal or plant a ten-year-old knows or a word described in its own sentence (the Gila monster "a big slow lizard"; the pika "a small round animal like a guinea pig"; voles "small mouse-like animals"); "lichen", "pry" and "burnt" were swapped during the build for "soot", "pull" and "burned".
• grade5/check_g5_articles.py → 0 FAIL; four WARN: articles 376 / 434 / 423 / 425 words (> 300). Answered: L02 carries the four behaviors TEKS 5.13B names plus the instinct / learned recall; L03 and L04 carry the fox run and a fresh kind each; every article has a subtitle or figure every ≤ 4 idea lines and sits inside Grade 3's approved range (81–505, median 308); the scripts carry the whole lesson in 148–149 words.
• build_tools_check_terms.py → hits: {} (no A, B, C or D).
• check_key_tells.py → KEY TELLS: 0 of 109 MCQs flagged (a: 0, b: 0, c: 0).
• check_svg_ids.py → PASS (26 figures, 0 ids).
• check_lessons.py --adapter tools/check_adapter_g5.json → 0 FAIL, 72 WARN (S118 0 FAIL; S116 silent). The WARNs: S8 beats (as above); 43 × S11 "fb[key] is not empty" — kept: the Grades 3–5 page prints an fb line for every option; 7 × S22 — L01 q4–q9 carry "structure" after q3's key by design (the term is recognised at position 3 and used from then on, the 2 Oct "after a term is coined the scaffold noun is retired" ruling), L02-q7's stem names both options of the sort and L02-q9's stem states the definition the item then applies; 8 × S4 long stems — the hard bank items (allowed hard), PR-9 (the beetle wings, four sentences; kept because the "nothing is harder for shorter wings" datum is what makes "about the same" honest), the summary's third question (27 words); S17 silent after the attached-stem trim; S18 "34 live checks" — the adapter counts all four quizzes; S26 "no typed term-recall item" — "structure" is RECALL_BY_MCQ; S30 "no article form 'the migrate'" — a verb takes no article; the accept list carries its verb forms.
• build_tools_names.py Farah Gus Hana Cardoso → Farah (topic17), Gus (topic11), Hana (topic01) recur in other topics — information only (James, 6 Oct); Cardoso free; one child per name inside this topic.
• build_tools_build_topic_site.py → builds (4 lessons, 13 practice, 52 pp100, 26 figures). Screenshotted at 800 px: the L01, L03 and L04 article routes, the L02 quiz route and the summary route; every figure present and in frame (the ears strip was widened from 420 to 500 units after its second note ran past the frame on the first render).
• build_tools_blind_bundle.py → bundled: 100 mcq, 4 text (0 numeric), 1 frq, 0 order, 1 multi — adversarial rows: 4 — held: 106 at blind/2026-10-08/ (re-cut after the bank's re-scening pass).
• build_tools_video_batch.py → 6 videos, handoff/G5T29/videos/batch_G5T29.json + READY.md; attached key string B A A B B A A. check_brief.py → INTRO ✓, L01 ✓, L02 ✓, L03 ✓, L04 ✓, SUMMARY ✓ with the standing summary/protected note only (three closing bullets were shortened after the first run). grep -ci strand = 0. Not sent.

Bank freshness, re-performed

Scene nouns of every lesson, quiz, attached, summary, recall-bank and practice item (71 nouns: saguaro, Gila, jackrabbit, Tucson, trail, kit fox, campsite, flashlight, Canada geese, groundhog, bobwhite quail, bear cub, red fox / foxes, Minnesota, cottontail, coyote, beaver, heron, monarch, tadpole, Georgia, prairie dog, bison, pronghorn, Nebraska, musk ox, wolf, pelican, Indiana, hummingbird, deer mouse, bobcat, fawn, sea otter, clam, sparrow, Wisconsin, soot, Idaho, trout, waxy, hare, Vermont, bee, moth, snail, Maine, Florida lake, chipmunk, barn, owl, Utah, kitten, robin, sister, Iowa, woodpecker, spider, puppy, sea star, Tennessee, tree hole, Louisiana, wing, cricket, sunflower, Nevada, giraffe, pond) grepped against the bank. Five items were re-scened on the first grep: the Washington clams (→ an Oregon beach), the dry Oklahoma pond (→ a stream that dried up), the Utah lizards after a fire (→ snakes), the Georgia beetles after a wildfire (→ Oklahoma grasshoppers; the attached L04 question owns the Georgia beetles), the Arizona crickets on dark soil (→ New Mexico spiders on bark chips; the practice FRQ owns the crickets). Remaining hits are place names or nouns shared with a DIFFERENT move (Minnesota PP-21's pond plants; the Nevada coyote hunting at night; "clam" in the Massachusetts shell items; "moth" in the typed instinct item; "trout" in Gus's fins claim; "hare" in the snowshoe hare's feet; "bee" matching "beetles"; "wing" as an option word; "pond" in six items whose moves differ from the intro's drought and the attached pond items) — read and kept.

Blind-text grep

The bundle's two [Drawing: …] lines (the beetle-legs table and the vole table), grepped for the keys' words — "suited", "already", "pulled through", "more likely", "tick", "cross" — 0 hits (each line prints the four counts as data).

Figures (author self-check on the renders; the non-author review-figures pass is owed)

Eight code-drawn figures rendered to PNG by headless Chrome and read, plus the three article routes on the built page: the six fox cards show two tall bumpy bands and four low bands, labelled in words, no letters or numbers; the March form carries ticks on both thick-fur cards and on one thin-fur card, crosses on three, under the printed question; the timeline strip shows six cards, the arrow with its "the coldest winter" chip, then three cards, with the counts under each group; the ears strip prints its question above a tick and a cross with each place's reason (widened to 500 units so the second reason sits inside the frame); the three tables' cells fit their columns (Topic 18's guard). The rule cards carry each lexicon sentence. Placeholders print their notes; none names an item's answer.

Sentence-atomise self-check (one table per paragraph, collapsed to the rows that changed)

Every other row: yes (ideas counted; cause before effect — what the fox did before why; named actors — "the ears", "the stem", "the thick-furred foxes", "the change"; verbs carry the cause — "heats", "holds in", "stores", "decided"; explicit links; plain word before the term — "a part of a living thing's body … a structure"; comparisons in one frame; no fluff frames).

check_parallel.py on the topic's contrast pairs — "Thick fur keeps a fox warmer in the cold." / "Thin fur keeps a fox cooler in the heat."; "A structure is a part of the body." / "A behavior is something the animal does."; Grade 4's instinct / learned pair reused word for word — reports every differing slot and asks for one frame on all three. Read and kept: each pair already shares one sentence frame; the slots that differ are the deciding variable and the outcomes it forces (thick → warmer → cold; thin → cooler → heat), which cannot be made one word, and the third pair is Grade 4's own approved wording. For the non-author reviewer to read.

Banned-word grep (natural selection, evolve, adapt, selected, pass on, gravity, leak, lift, snap, non-living, strand, cord, outlet) over the three content modules: 0 hits in student text; the only matches are the modules' own comments and two literal uses — "lift a trash can lid" (PP type 7, the raccoon) and "geese lifting off" in a screen cue — both the real action, which the figurative-verb rule allows.

Names run

python3 build_tools_names.py Farah Gus Hana Cardoso → Farah INFO (topic17), Gus INFO (topic11), Hana INFO (topic01), Cardoso free; exit 0. All from Topic 29's row of grade5/NAMES_POOL.md; one child per name in this topic; no other child named anywhere ("his dad", "her grandma", "a student").

Judgement calls and rule collisions

1. "Function" is said as "job". TEKS 5.13A names "structures and functions"; Grade 3 and Grade 4 taught the idea as "a job" ("every body part does a job"), and the lesson carries one coin, "structure". A second coin in the same lesson would cost more than it buys at this grade; "job" is the word every earlier lesson and the recall bank use. Decided and recorded, not asked.
2. "Individual" is glossed, not coined (ruling 13 collision). NGSS and Florida say "individuals"; Topic 28 ("different instructions, different individuals") is being written in parallel and may coin it. Coining it here too would give the course two coinings; so L03 glosses it in plain words on first use ("each single fox, each individual"), lists it in recalled_terms as Topic 28's word, uses it in every rule line and never tests it as a term. For the coordinator: if Topic 28 does not coin "individual", this topic's gloss stands on its own and nothing breaks.
3. The articles run 376–434 words against the brief's "above 300, trim or split" (ruling 13 collision). L02 carries four named behaviors plus the instinct / learned recall; L03 and L04 carry the fox run and a fresh kind. Splitting would make five or six lessons of four atoms, against the architecture row; so each article keeps a subtitle or figure every ≤ 4 idea lines and the WARN is answered above. The scripts, the default path, carry each lesson in 148–149 words.
4. The winter coats were already grown. A real red fox grows its winter coat in the fall, so the register's trap ("the foxes grew thicker fur because the winter turned cold") is met on its true ground: "Every fox has its winter coat by now… A fox cannot grow a thicker coat because the air turns colder: its winter coat is already grown." The difference the lesson uses is between the foxes' winter coats, which is where the individuals differ.
5. The heat language. "The warm ears heat the air" and "thick fur holds in the body's warmth" follow the 7 Oct heat ruling; "a thin-furred fox got too cold" and "died of the cold" say the positive word.
6. Hana's claim is the one voiced line. The register's SD-5 belief is the lesson's point, so L04's headache may quote a child (ruling 1's exception); no question is asked of the claim, and the verdict recaps and contradicts it after the evidence (the 7 Oct correction).
7. No Yes-keyed claim in the quiz. The only honest TRUE twin of the SD-5 belief is the lesson's own rule, which the plain items already test; so the one quiz claim keys No and the two-option strings are balanced by plain yes-keyed facts instead.
8. "About the same" is keyed once in the quiz and once in the bank (the trout's spots; the turtles' shell color in salty water), so a child learns that a described difference matters only when it suits the change — the sorting the standards' "differences between individuals" asks for.
9. The typed hard items are Grade 3 and Grade 4 words carried in this topic's rule line ("migrate", "huddle") or recalled in L02 ("instinct", closed by the author's note), because this topic's one coin is nine letters; the hard tier is otherwise well-made MCQs (claim-by-reason, two differences weighed, the evidence item, the set).
10. Three-sentence script beats kept (S8 WARN): one scene state per beat, as in the approved Grade 3 scripts.

Owed

• The non-author passes: review-figures (every rendered figure, the fox cards' counts, the ears strip), review-questions (the 34 quiz items, 13 practice, 8 recall-bank, 52 bank), review-accuracy on the real organisms named (the Gila monster's fat tail, the jackrabbit's ears, the kit fox's night hunting, the groundhog, bobwhite covey roosting, the black bear cub's learned foraging, the red fox's winter coat).
• The blind read of blind/2026-10-08/BLIND_BUNDLE_G5T29.md against KEYS_SEQUESTERED_G5T29.json.
• Pictures: the four cartoon and nine photo placeholders through David's pack (generate-figure, qc-figure) after the text is reviewed; no question for James holds them.
• Videos: the batch at handoff/G5T29/videos/ waits for James's go after the Life sample (Topics 24 and 25); no Grade 5 batch leaves the machine without his word.
• The topic page on the review site, the scripts page and the Needs-you page: the coordinator's steps, not this agent's (nothing outside grade5/topic29_build and handoff/G5T29 was touched).