Topic 30 "The Sun's path and the year's pattern": your read (2 October 2026)

Two readers answered 111 items cold and agreed every multiple-choice key with each other and with the bank. Both caught one key the topic's own rule contradicted: the practice item PR-09 gave April 13, May 14, June 15 hours and asked for July, keyed "about 16" (the pattern carried straight on), while L04 and the bank teach that after June the daylight grows shorter. The key rewarded what the topic calls wrong. It is now a temperature run (Fresno, 65, 75, 85 → about 95) that does not cross a turn. Both also found a bank item with two defensible numbers (Peoria, August from April to June: 14 by the turn, 17 by the straight pattern), now a direction item (more / the same / fewer than June); a quiz graph whose three shortest bars were December, January and November while the key said February; the month names handing over the season on every bar-graph item; three bank items that reused quiz datasets town for town; "the Sun's path" printed on the page before the item that asked for it; one bank item written twice; and Paige beside Parker. All applied in v2.

Reader C then read the whole v2 bundle, 117 items, and agreed every key; on every multiple-choice item exactly one option stood up, and every full-year key agreed with the bars described. Five things were worth a look, all applied in v3: the quiz's summer graph (15, 15, 14) gave a child who had just learned "falling = fall" a foothold, so it is flat at 15; a temperature prediction ran to July, one month past the June turn the bank also tests, so it now ends at June; two give-away pairs are written down as draw constraints (the for-ever claim beside the Peoria turn item; the Savannah stem that names December, January and February beside the two full-year winter items); and three typed forms a child might write ("the summer months", "the path of the Sun") that the grader would have refused are accepted. The one faint tempt left (Baltimore 14, 15, 15, read as summer) stays, because the flat shape now belongs to the quiz and a bank item may not be a quiz drawing renamed.

The teaching is sound and the science honest, the reviewer said: the four lessons say exactly what Grade 4 may say, the class's records round correctly from real daylight at 40 degrees north, and every figure opened is clean with east at the left. But the item sets were not yet testing the atom they claimed: most bar-graph season items could be answered from the month labels alone, three bank items reused quiz datasets, and the lesson text said "shorter each month after June" while its own graph showed June equal to July. The reviewer's ruling, adopted: three-bar items hide their month names and use mid-range values so the bars do the work; full-year items keep their months as an easy scaffold, on data that agrees with the calendar. Five essentials (the practice key above, the Peoria item, the quiz graph's bars, the dataset reuse, the month labels) and twelve worth-fixing lines (the plateau sentences, "looked up when the Sun came up", "the Sun is still high" at 6 in the evening, a claim item that repeated its neighbour, a TRUE claim mis-tagged, a one-step item tagged hard, the quiz accept lists, a feedback line that mentioned Earth's orbit) were all applied. The reviewer answered the author's three questions yes, yes and yes, and ruled a rename, as the page above shows.

What a student notices first is right, the reviewer said: east at the left, the Sun low over the parking lot, highest over the gym but a third of the frame down, June's orange arc higher and wider than November's blue one over the same landmarks. But at L03 the bars stopped doing the work. The teaching graph's bars did not match its brackets: the class's year of daylight had November at 10 hours under FALL and February at 11 under WINTER, so the three shortest bars were December, January and November while the text beside the graph called December, January and February "the shortest"; May tied August under the summer claim too. The fix pass moved the four cells the capstone does not lock (September 13, October 12, November 11) so both "taller a little each month to June" and "shorter a little each month after" are true of the drawing, cut the text back to what the bars show, and put every exact-triple item on a temperature graph, where the calendar fits. The honest fact underneath, which the topic cannot yet say, is that the calendar's seasons start a little after the daylight turns; that is question 4. Also fixed: two bank graphs with ties on the asked triple; three-bar items keyed by month names while the bars pointed the other way; figures that were the same drawing twice; an unplaced headache drawing; a chip over a bar; an arrowhead through the midday Sun; and a 560-wide graph family whose month labels fell to 6 px on a phone, now a 360-unit frame with 17-unit labels, measured at 16 px in a 360 px shot.

52 items in content/G4T30_pp100_bank.py: 39 four-option MCQs, one declared three-option direction item, one order item (three paths, shortest day to longest), 5 typed words (the term twice, a season three times) and 6 typed numbers. Per atom: SPA4-001 14 · SPA4-002 11 · SPA4-003 12 · SPA4-004 15. Easy 20, medium 15, hard 17, sorted easy to hard and rising with the atom's place in the topic. MCQ keys A 10 · B 10 · C 8 · D 11. Sixteen question types, one exemplar each on the type sheet: which direction, from where the Sun comes up or goes down; where the Sun is at a named time; which time was midday, from a shadow table; type the name of the curve; which month has the most or fewest daylight; which two months a high and a low path could be, and the order of three paths; how the daylight or the sunset time changes over a span; a child's claim is wrong, pick the reason; which three months are a season, on a full-year temperature graph; which season three unnamed bars are; type the season from three unnamed bars; predict the next number from a table or a graph, picked and typed; the turn after June; the month a falling pattern reaches a value; next year's repeat. Every scene is fresh against the lessons, quizzes and practice (the scene-noun list is in the build record), and the type sheet lists the draw constraints, including the two Reader C asked for.

Skill: course-build → write-lesson (L01 a concept introduction in the fact form, coining the Sun's path last; L02 a comparative fact lesson; L03 a graph READ lesson; L04 the capstone prediction routine) → write-example-sequence, design-check-sequence, sentence-atomise, write-explanation, write-mcqs (quizzes, attached questions, practice, the PP100 bank and its type sheet); write-video-script (intro, four lesson scripts, summary; the elementary micro-lesson anatomy); review-figures (author self-check on the renders and the built page; the non-author pass is owed) @ efd8fba

Grade 4 Topic 30 — The Sun's path and the seasons — build record (v1, 2 October 2026)

What this is, in plain words

Four short lessons, one per atom, and no cause anywhere. A child first watches the Sun through one school day over one schoolyard: low over the parking lot in the morning, highest over the gym at midday (but not straight overhead), down behind the trees at the end of the day — the shadows say the same thing — and only then meets the name the Sun's path (L01). Next the same schoolyard carries two paths, June's high and long, November's low and short, and the class's year of daylight records shows the days growing a little longer each month to June and a little shorter each month to December (L02). Then the child reads a season off a bar graph: the shortest bars are winter, the tallest summer, bars growing taller are spring, bars growing shorter are fall — worked on a temperature graph from Portland, Maine, then on the class's daylight graph (L03). Last, the capstone: four months of records, a pattern (one more hour each month), carried on one step to predict May, checked, then turned after June and repeated next year (L04). Why the days change is said once to be "a question for a later grade"; the Sun "seems to move", "comes up" and "goes down" (Grade 3's words).

Counts. 4 lessons · 4 scripts (146, 147, 129, 150 spoken words) + intro (105 words, 3 question marks) + summary (143 words) · 4 attached questions + the intro's poll + 3 summary questions · 39 quiz items (10 + 10 + 9 + 10; 36 MCQ, 1 typed text, 2 typed numeric) · 13 practice items (11 MCQ, 1 typed, 1 written answer; three Grade 3 retrievals; "the Sun's path" recognised in PR-12 and typed in PR-13) · PP100 bank 52 items (40 four-option MCQs, 1 order item, 5 typed text, 6 typed numeric; SPA4-001 14, SPA4-002 11, SPA4-003 12, SPA4-004 15; easy 15, medium 19, hard 18; MCQ keys A 10 · B 11 · C 9 · D 10) · 70 figures (63 SVG by code, 3 photo placeholders, 4 cartoon placeholders) · checker 0 FAIL, 19 WARN (every WARN answered below) · check_g4_articles.py 0 FAIL (articles 287 / 376 / 257 / 267 words — the one WARN answered below) · names register exit 0 · page builds (/tmp/g4t30.html: 4 lessons, 13 practice, 52 pp100, 70 figures) and its eight routes were screenshotted and read · check_svg_ids.py PASS (0 ids) · blind bundle cut (94 MCQ, 15 typed of which 8 numeric, 1 written, 1 order; [Drawing: lines 42, season / size-word hits 0) · video batch built (6 videos), check_brief.py clean except the INTRO's standing R79 hit and the intro-questions WARN at the limit, grep -ci strand 0 · NOT sent.

Decisions made without a blueprint

One line each, with the reason:

• Four lessons, one per atom, in register order — the architecture row names four atoms and the capstone (SPA4-004) closes the topic; each payoff is one clause.
• PATTERNS ONLY: no cause anywhere — the register's JM-P23 scope and the task's rule; the daily motion's cause is Topic 31's (SPA4-009) and the seasons' cause is grades 6-8's; the single sentence "Why the days change through the year is a question for a later grade." sits in L02's article after the records and nowhere else (not in a script, so no render carries it).
• No item can invite the distance answer — because the topic offers no cause, no claim item is written on "the Sun is closer in summer" or "Earth is far from the Sun in winter" (a "no" key would itself assert a cause); the distance belief is avoided by wording, recorded in the misconception table below.
• The Sun "comes up", "goes down" and "seems to move across the sky" — Grade 3's convention, quoted word for word: "The Sun comes up over the sea." (Grade 3 Topic 29, L fig caption), "the Sun goes down behind the roofs" (Grade 3 Topic 30 intro), "the Sun is rising over the hill" (Topic 29 intro). This topic uses the first two forms and never "rises" / "sets" / "travels" / "sinks" as the Sun's own motion ("sunrise" and "sunset" appear only inside two item tables as the column heading "Sunset time", a time of day).
• East and west only; no north or south — the register's behaviour says "toward the east … toward the west"; the midday Sun's southern side would be a third direction word with no item to carry it; the arc drawing is a side view with EAST at the left and WEST at the right (the register's convention: the student looks south).
• "Not straight overhead" is said in L01 and tested twice — Plummer 2009's belief is confronted by wording in the article and the script, by the WHERE4 distractor, and by Oliver's claim (L01-q8).
• The sunrise spot shifts, said in one sentence — "So the Sun does not come up in exactly the same spot every day." in L02, drawn on the two-path figure (June's path starts left of the parking lot, November's right of it); no compass word for the shift and no item asks for it (the register's behaviour is day length, and the shift is one of the age's documented beliefs — Plummer 2009 — so one sentence, no more).
• L01's shadow table is in L01, not a lesson of its own — the Sun's height read from a shadow is the same fact as the path (TEKS G4-9A's "patterns of change in shadows"); Grade 3's shadow sentence is recalled, never re-taught.
• The daylight records are MADE UP, for one named real town — Miss Rasmussen's class in Springfield, Illinois (about 40 degrees north): Jan 10, Feb 11, Mar 12, Apr 13, May 14, Jun 15, Jul 15, Aug 14, Sep 12, Oct 11, Nov 10, Dec 9 (whole hours; real values at that latitude round to these within an hour); every item town's records are made up the same way and every graph says "records", never "data from the weather service".
• Whole hours, a plateau in June and July — the bar graph needs multiples of its grid; the equal June and July bars are real at this latitude and are read in L03 as "the tallest"; every numeric prediction is set on a run with a steady step (the rising half, Jan to June; the falling Sep to Dec), and the register's qualitative "longer still" carries the uneven stretches.
• Temperature is the second data kind, in L03 only — the register's behaviour names "monthly temperature or daylight-hours data"; the Portland temperature graph is the worked case because its seasons are read from the bars alone; the class's own daylight graph is the second run so the daylight routine is also seen.
• L03's rule says "the shortest bars", not "the shortest daylight bars" — the routine is run on temperature bars first; the rule must be true of both kinds (the shortest temperature bars are the coldest months).
• L04's calculation is its own line — "13 hours + 1 hour makes 14 hours." (the vault: never talk a calculation through in prose); "makes" is the verb the checker's S2 needs, and the fb / why lines use the bare "14 hours + 1 hour = 15 hours" form marked # fragment ok.
• The intro is one scene (Paige at the bus stop at 7 in the morning, May then December) — it does not reuse L01's recess or L02's soccer field, so both headaches stay fresh; it coins nothing ("path" is never said) and names only things a child can picture.
• Every lesson gets a video — L01 is a motion a child should watch; L02 shows two paths drawing over one horizon and the bars building; L03 and L04 are data videos and put the graph on screen with the bars lit as they are read.
• The order item on three paths is the bank's one non-MCQ, non-typed form — a pair of paths has only two honest options and a bank MCQ needs four; the order item asks shortest day → longest day and stores the key that way (L, K, J); no sibling asks the reverse. Convention stated here: the drawer lists paths as given; letters sit at each path's top.
• "Path" is the typed term (4 letters) and no term is in RECALL_BY_MCQ — the seasons (winter, summer, fall) are typed from three bars because the season is not in view on a month-labelled graph, with "autumn" accepted.

Grade 3 and Grade 4 doorways, recalled by one item and one line each, never re-taught: Grade 3 Topic 13 ("An object blocks the light. Its shadow is on the side away from the light source.", ENE3-012) — L01's shadow line, L01-q1, PR-2; Grade 3 Topic 23 ("To read a bar, find its top, follow it across, and read the number it lands on.", SR3-012; "If the same weather came every year, it is likely to come again next year.", EAR3-023) — L03's read line and L02-q7 / L03-q1 / PR-1; L04's repeat line and L04-q2 / PR-3; Topic 25 (SR4-010's two labels, split into two idea lines of ≤ 22 words: "A bar graph needs two labels." / "One label up the side says what the numbers count, and one along the bottom says what each bar stands for.") — L02's lines and L02-q1; Topic 8 ("A pattern is something that happens again and again in the same way.") — L04's recall line and L04-q1.

Neighbouring topics as built: Topics 29 and 31 are not yet on disk (other agents); the latest Earth and Space topics read as built were Topic 22 (the schema and record shape) and Topic 25 (the bar graph, record table and two-label conventions: graph(), table(), item_alt, bars_desc, imported from topic25_build/content/figures.py).

The simplification, named: whole-hour daylight records for a town at about 40 degrees north; "about fifteen hours" and "about nine hours" in prose; the sunrise spot's shift stated without a direction; the midday Sun's height stated as "not straight overhead" with no angle. No note is printed in the article beyond "about".

Questions for James (the build does not wait on them)

• Should the article say the midday Sun is "toward the south", or stay with "high above the gym"? Context: in the US the midday Sun is always in the southern half of the sky; the register asks only for east → highest → west, and the figures are a side view. Recommendation: stay with east and west only — a third direction word would need its own drawing and items. Built without "south".
• Is the daylight plateau (June 15, July 15) right to keep, or should July be 14? Context: whole-hour records at 40 degrees north genuinely round to 15 and 15; the equal bars are read as "the tallest" in L03 and no prediction item sits on them. Recommendation: keep — honest data, and a child should meet a pattern that flattens before it turns. Built with the plateau.
• Does L02 (376 words) stand at its length, or should the records section be a short lesson of its own? Context: the brief allows an article above 300 words with a subtitle or figure every eight lines and a reason; L02's longest run of idea lines without a subtitle or figure is 5; it carries two faces of one atom (the two paths and the year's records) and both feed items. Recommendation: keep as one article — a split would make a fifth lesson for a fact atom. Built as one.

Design-check-sequence records

`

Skill: design-check-sequence @ efd8fba

Part: L01 (SPA4-001)

Type: (a) one-step fact — where the Sun is at a time of day; which side is east or west; the answer is one label.

Plan: recall the shadow side (2 opt, A) → the Oregon hills (2 opt, A) → the Colorado mountain (2 opt, B) → the lettered midday spot (3 opt, 0) → midday in Georgia with the overhead distractor (4 opt, 1) → the mailbox shadow table (4 opt, 2) → the Ohio window (3 opt, 1) → Oliver's overhead claim (no) → the Minnesota fisherman (3 opt, 2) → the term among three (1). Two-option string A A B B; three- and four-option 0 1 2 1 2 1.

Belongs in practice/test instead: the four-way direction and where-is-the-Sun sets; the typed term (PP types 1, 2, 3, 5).

`

`

Skill: design-check-sequence @ efd8fba

Part: L02 (SPA4-002)

Type: (a) one-step fact with two faces — which month is longest / shortest; which drawn path is summer; which way the daylight changes over a span.

Plan: recall the side label (2 opt, A) → the Michigan lifeguard (3 opt, 0) → paths J K (2 opt, A) → paths L M (2 opt, A) → Denver's fewest hours (3 opt, 2) → the Georgia porch light (3 opt, 0) → the Boise graph read (3 opt, 1) → Paige's same-length claim (no) → the Illinois flagpole's midday shadow (2 opt, B) → Omaha June to December (3 opt, 1). Two-option string A A A B B; three-option 0 2 0 1 1.

Belongs in practice/test instead: the four-way month and change sets; the month-pairing path items; the order item; the claims as reason items (PP types 6–9).

`

`

Skill: design-check-sequence @ efd8fba

Part: L03 (SPA4-003)

Type: (b) multi-step routine — find the shortest / tallest bars, read the direction between, name the season.

Plan: recall the bar read on Tucson's temperatures (3 opt, 1) → step 1, Salt Lake City's shortest three months (3 opt, 0) → step 2, El Paso's tallest bars' season (3 opt, 1) → Richmond's three rising bars (3 opt, 0) → Nashville's three falling bars (3 opt, 1) → bare, Boston described (3 opt, 2) → Paige's TRUE claim (yes) → the recognition item, which season has the shortest bars (4 opt, 1) → Columbus's three bars, the season typed. Two-option string A; three- and four-option 1 0 1 0 1 2 1. The why FRQ form is not used here (a routine); the typed item is the production step.

Belongs in practice/test instead: full-year graphs with four-way triples; typed seasons from three bars (PP types 10–12).

`

`

Skill: design-check-sequence @ efd8fba

Part: L04 (SPA4-004)

Type: (b) multi-step routine — read how the records change, carry the pattern on one step, say the next value; the turn and the year's repeat as the bare transfers.

Plan: recall pattern (3 opt, 1) → recall came-every-year (2 opt, A) → step 1, Denver's change (3 opt, 0) → step 2, Omaha's direction (3 opt, 2) → bare, Reno's number (3 opt, 0) → Albany's sunset transfer (3 opt, 2) → Parker's for-ever claim (no; tagged apply, not misconception — no documented belief, see below) → Wichita's June typed (15) → next year's December (3 opt, 0) → Toledo's June temperature typed (75). Two-option string A B; three-option 1 0 2 0 2 0. Typed items sit at positions 8 and 10 (never in the first third).

Belongs in practice/test instead: four-way numeric predictions; typed predictions on drawn graphs; the two-step items; the claims as reason items (PP types 13–16).

`

The key strings

• Attached questions in video order (L01, L02, L03, L04, summary 1–3): A A A C A C B — A 4 of 7 (57 %), under the 60 % line.
• Lesson quizzes, two-option items: L01 A A B B · L02 A A A B B · L03 A · L04 A B — no four in a row alternate; no position above 60 % within a lesson.
• Three- and four-option items: L01 0 1 2 1 2 1; L02 0 2 0 1 1; L03 1 0 1 0 1 2 1; L04 1 0 2 0 2 0; practice two-option A A A B A B, three-option 1 0 1 2 1.
• PP100 MCQ keys: A 10 · B 11 · C 9 · D 10 (40 MCQs); the order item and the eleven typed items carry no letter.
• Key never the longest option: the three bank items the length check flags (PP-7, PP-26) sit on the fixed sets WHERE4 and TRIPLES, whitelisted by name (the same words sit under every item of the set); PP-37's distractor was lengthened so the key is not the longest. In the quizzes the flagged items are two-option (one is always longer) or differ by one or two letters ("December" / "September"; "a pattern" / "a season").
• Claims: four bare yes / no claims in the quizzes (L01-q8 no, L02-q8 no, L03-q7 YES, L04-q7 no) plus PR-11 (no); No carries 4 of 5 — above the 70 % line, recorded here: L03's TRUE claim is the one honest true twin (the tallest bars are summer); the other beliefs have no true twin that is not the lesson's own rule.

The example sequences, written out

L01 (one frame, examples only; a fact true of every day, so no non-example — the Forces log's rule):

• Early in the morning, the Sun comes up over the parking lot. That side is the east.
• Around midday, the Sun is at its highest, over the gym, but not straight overhead.
• Through the afternoon, the Sun seems to get lower, toward the trees.
• At the end of the day, the Sun goes down behind the trees. That side is the west.
• → the name: the curve the Sun seems to follow across the sky each day is called the Sun's path. → the rule.

L02 (two frames, lined up, the comparison in identical structure):

• The June path is high and long, so the daylight lasts a long time.
• The November path is low and short, so the daylight lasts a short time.
• The same two frames on the shadows: at midday in winter the Sun is low, so even the midday shadow is long; at midday in summer the Sun is high, so the midday shadow is short.
• The records: from December to June the bars grow taller a little each month; from June to December the bars grow shorter a little each month. → the rule.

L03 (a routine, worked once then run again): the shortest bars are January and February: winter · the tallest bars are July and August: summer · between winter and summer the bars grow taller each month: spring · between summer and winter the bars grow shorter each month: fall · the same four reads on the daylight graph · the season table · the rule.

L04 (a routine, worked then checked): the records in order (10, 11, 12, 13) · one more hour each month, again and again: a pattern · carry it on: 13 hours + 1 hour makes 14 hours · May counted: 14, right · the turn after June · the year's repeat (the two-year table) · the rule.

The rules applied

Misconceptions, by name, and what was done

The vault's elementary catalogue is a physics catalogue and is silent on the sky beliefs; the sources below are the literature on children's astronomy.

Checker result, every WARN answered

python3 build_tools_check_topic.py grade4/topic30_build/content/G4T30.py → 0 FAIL, 19 WARN.

• S34 "bar graph" has no recognition or recall item; "shadow" and "pattern" have no recall item — kept: all three are earlier topics' words (Grade 3 Topic 23, Grade 3 Topic 13, Topic 8), listed in terms for the glossary and recalled by one line and one item each; coined_at names only "the Sun's path", which has both items. The checker reads every terms key.
• S36 unusual words (16), each answered: gym (×6) fine — the everyday US school word, the schoolyard's landmark; soccer (×3) fine — the sport every US 9-year-old names; prediction (×3) fine — the noun of the lesson's verb "predict", explained by the routine before it is used; shrinks (×2, the blurb and the intro) fine — the everyday word for "gets smaller", of the daylight; rink (×2) fine — "skating rink" is whole in the stem; ferry, hiker, crew fine — plain US words, each in a stem whose science does not depend on knowing it; zoo fine; Tucson, Reno, Fresno, Cheyenne, Louisville, Cleveland, Des Moines — real US place names, one each, as the brief asks (real US places). Swapped before this run: "plus" → the calculation line; "unfrozen" → "free of ice"; "horizon" (fb) → "is longer than"; "construction crew" → "road crew"; "casts a shadow" → "has a shadow".

python3 grade4/check_g4_articles.py → 0 FAIL; one WARN: L02's article is 376 words (> 300). Kept: it has a subtitle or figure every ≤ 5 idea lines (eight figures and five subtitles), the two faces (the paths; the records) each feed items, and the brief allows the length with a reason — put to James above.

check_brief.py → the INTRO's standing R79 hit (accepted, every topic) and intro/questions WARN "3 question marks — at the limit": the two headache questions plus the ready line; kept, because the two questions ("Where did the morning daylight go? And will it come back?") are the topic's two payoffs (L02 and L04).

Bank freshness, re-performed

Scene nouns of every lesson, quiz, attached, intro and practice item: bus stop, streetlight, schoolyard, parking lot, gym, trees, recess, basketball post, soccer, field lights, cousin, Portland, chart, calendar, barn / Iowa, highway sign / Texas, lighthouse / Oregon, camper / Colorado, mailbox / Kansas, apartment window / Ohio, fisherman / Minnesota, lifeguard / Michigan, Denver, porch light / Georgia, Boise, flagpole / Illinois, Omaha, Vermont, Topeka, Tucson, Salt Lake, El Paso, Richmond, Nashville, Boston, Columbus, Maine, Albany, Reno, Wichita, Toledo, Idaho, mail carrier / Kentucky, rink / Wisconsin, Arkansas, Sacramento, baker, Madison, Fresno, Utah. Grep of the bank's stems, options and whys: every scene noun 0 hits; the only hits are STATE names (Iowa 0 after the Evansville swap; Kansas 1 — Kansas City, Missouri; Ohio 3 — Cleveland, Dayton, Cincinnati and the fence post; Michigan 2 — Detroit, Lansing; Georgia 1 — Savannah; Illinois 1 — Peoria; Utah 1 — the tent pole) and "flagpole" 1, re-scened to a telephone pole before this record. A state is not a scene; every bank scene (a ferry, a water tower, a zoo, a drive-in, a crossing guard, a hiker, a road crew, a fence post, a tent pole, a park tree, a marching band, a newspaper's sunset times, a dairy farm, two farms with lettered paths, three paths in order) is one the student meets nowhere else in the topic.

Blind-text grep

grep "\[Drawing:" BLIND_BUNDLE_G4T30.md | grep -i -E "summer|winter|spring|fall|autumn|longer|shorter|higher|lower" → 0 hits on 42 drawing lines. Arc items give each path's top in px above the horizon and each end's distance from the house and the tree; graph items give each bar's top on the axis as a number (bars_desc); tables give their cells.

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

Every one of the 70 figures was rendered to /tmp/g4t30_figs/*.png in a headless browser; opened and read: path_day, path_anim (still), yard_two_suns, arcs_two, graph_daylight_marked, graph_daylight_seasons, graph_portland, graph_portland_seasons, graph_jan_may_pred, table_shadow, table_seasons, table_two_years, q_arcs_jk, q_sun_three, q_graph_boise, q_table_albany, pp_arcs_np, pp_arcs_jkl, yard_cartoon (placeholder). Fixed on the render, not the source: the trees moved clear of the day path's end and the afternoon Sun (first render ran the path through the canopy); the item house moved left and the item arcs' starts moved right so no path starts inside the house; the season graph's "Month" label moved below the season chips (it sat on the SUMMER chip); the prediction chip "OUR GUESS: 14" kept inside the frame; the three-path item's letter chips moved beside each Sun (they had stacked over the Sun below). The built page's eight routes were screenshotted at 480 px: every figure present, no empty box, item graphs readable at phone width. Arrows along every path run one way, left (east) to right (west). Colours in captions match the renders (orange June path, blue November path, yellow daylight bars, red temperature bars).

Names run

python3 build_tools_names.py Rasmussen Oliver Paige Parker → all four "free", exit 0 (Topic 30's row in NAMES_POOL.md). No other name appears in any student-facing string; "Miss", never "Mr." or "Ms.".

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

Judgement calls

• The schoolyard horizon is a side view with EAST at the left — the student looks south, as every US child does to watch the Sun; the register names this once; the item horizon keeps the same left-to-right convention with a house and a tree so a child never confuses it with the teaching scene.
• "Comes up" / "goes down" over "rises" / "sets" — Grade 3's convention; the one place "sunset" appears is a table heading ("Sunset time"), a time of day, not a motion.
• The recess headache asks "where will the Sun be at lunchtime?" rather than "did the Sun move?" — the second question has no honest answer at this grade (the spin is Topic 31's); the first is answered by the path itself.
• L03's headache is a mailed graph from a cousin's class — a thing classes really do; the teacher asks the question a teacher would ask; no sign or notice is invented.
• No claim item on the distance belief — see the misconception table; a "no" key would assert a cause the topic does not teach.
• Parker's for-ever claim is tagged apply — the belief is not documented in the literature as a held belief; the item tests what the lesson covered (the turn after June).
• The order item is the bank's one non-MCQ form apart from typing; its three paths are drawn on their own figure and the key runs the stem's way.
• Place names count as S36 hits — kept, because the brief asks for real US places and a name is not a reading-load word.

Owed

• Non-author passes: review-figures over the 70 renders; review-questions over the 39 quiz, 7 attached, 13 practice and 52 bank items; the blind read of blind/2026-10-02/BLIND_BUNDLE_G4T30.md (keys sequestered in KEYS_SEQUESTERED_G4T30.json; the order item's reader answer compared as a list with check_order_keys.py).
• The picture pass (3 photos, 4 cartoons) through David's pack — AFTER James's sign-off, because this record carries three questions for him.
• Videos: the batch at handoff/G4T30/videos/batch_G4T30.json with its READY.md is built and checked, NOT sent (James's word is needed for Grade 4's first batch).
• The knowledge-graph repo: no register change proposed; a note that SPA4-001's behaviour is taught with east and west only.

v2 — reviews applied (2 October 2026)

Skill: course-build → write-lesson (lesson lines; write-video-script for the beats) → write-mcqs (every item change; check_key_tells.py) → write-explanation (fb and why lines) → review-figures (author re-render after the drawer changes) @ fb37d14

In plain words. The blind read, the guidance review and the figure review of 2 October are applied in place. The class's year of daylight now matches what the lesson says about it (July to December falls a little each month; the plateau after June is said out loud); every season item on three bars hides its month names, so the bars do the work; the full-year "which three months" items sit on temperature graphs, where the calendar's seasons genuinely fit; two keys the course's own rule contradicted are gone; Parker is Wade; and every graph is drawn in a phone-readable frame. Nothing was committed and no batch was sent.

#### Finding → change

#### The arrays, before → after

• DAYLIGHT (the class, Springfield): Jan 10, Feb 11, Mar 12, Apr 13, May 14, Jun 15, Jul 15, Aug 14, Sep 12 → 13, Oct 11 → 12, Nov 10 → 11, Dec 9. L04's chain (10, 11, 12, 13 → 14) and the two-year table are untouched; "December to June taller a little each month" and "July to December shorter a little each month" are both true of the drawing; the one remaining tie (May 14 = Aug 14) is not named by any sentence or bracket.
• q_graph_slc: the teaching daylight data → temperatures 35, 45, 55, 60, 70, 85, 90, 90, 80, 65, 50, 40. pp_graph_milwaukee: daylight 9, 10, 12, 13, 15, 15, 15, 14, 12, 11, 10, 9 → temperatures 30, 35, 45, 55, 65, 75, 80, 80, 70, 60, 45, 35. pp_graph_albuquerque: Nov 10 → 11. pr_graph_sacramento: Nov 10, Dec 9 → Nov 10, Dec 10 (its own records). pp_table_kc: Mar 12, Apr 13, May 14 → Feb 11, Mar 12, Apr 13. pr_table_fresno: daylight Apr 13, May 14, Jun 15 → temperatures Mar 65, Apr 75, May 85. Lansing: Sep 12, Oct 11, Nov 10 → Jul 15, Aug 14, Sep 13. Three-bar values: above.

#### Judged differently from the reviews, and why

• The review's array (Feb 10, Nov 11, Aug 13) was not applied as written. Feb 10 breaks L04's capstone (the class's own Jan–Apr records 10, 11, 12, 13 and "one more hour each month"), stalls L02's "taller a little each month", and would give the one running scene two datasets; Aug 13 puts May (14) above August, so "the tallest bars, June, July and August" becomes false; Nov 11 beside Oct 11 stalls the fall read. The honest facts: a daylight curve is symmetric about the solstices while the calendar's seasons are not, so December–January–February can never be the three shortest bars (November is nearer the solstice than February) and June–July–August never the three tallest (May is nearer than August). So the four cells L04 does not lock were moved (Sep 13, Oct 12, Nov 11), L03's daylight lines now claim only what the bars show ("The shortest bars are December and January… The tallest bars are June and July…"; the brackets name the calendar seasons under their months), and every exact-triple item sits on a temperature graph, where the calendar fits (Portland's worked case already did). One line for James: the daylight brackets on graph_daylight_seasons are the calendar's; a sharp child may ask why November (11 hours) is "fall" and February (11 hours) "winter" — the answer the topic cannot yet give is that the seasons lag the daylight.
• PR-09 is a temperature prediction, not the review's March–May daylight run (that run is L04-08 Wichita's; whole-hour daylight runs are nearly used up across the topic).
• Lincoln (PP-30) reads spring, not summer. With month names hidden there are only two honest summer shapes on whole hours (15-15-14 and 14-15-15); Columbus (quiz) and Baltimore (bank) hold them, so a third summer item would be a twin. Cincinnati's typed winter is 10-9-10 (a dip) rather than the ruling's 9-10-11, which rises and so fits spring's rule too.
• L03-02/03 "retagged easy": quiz items carry check-type tags (scaffold), not tiers; a "d": "easy" field records the ruling beside the tag. The type 10 bank items are retagged easy and re-sorted.
• Claims keyed No are now 5 of 6 (L01-08, L02-04, L02-08, L04-07, PR-11 no; L03-07 yes): the new L02-04 must be keyed no because it is the documented belief; the only honest true twin remains L03-07.
• "⟂ PP-32 ⟂ PP-45" read as Cheyenne ⟂ Lansing (the items sharing the summary's dataset); Lansing was given its own run and the Cheyenne ⟂ summary Q3 line recorded.

#### Gates

• python3 build_tools_check_topic.py grade4/topic30_build/content/G4T30.py → 0 FAIL, 23 WARN (v1: 19). New WARNs answered: S10 ×2 — L03-07's claim-maker is Paige, who also appears in L03's headache; the item was exempt while tagged misconception and the retag to apply exposes it; kept (the claim-maker is the topic's child, the rule's concern is a scene noun); S11 — the declared three-option set (Peoria), accepted by the checker and recorded here; S36 'wade' ×13 — a name, like the place names. The S34 ×3 and S36 place-name WARNs stand as answered in v1.
• python3 grade4/check_g4_articles.py … → 0 FAIL; scripts 146 / 149 / 129 / 150 words; L02's article 385 words (James: keep as one).
• python3 build_tools_names.py Rasmussen Oliver Paige Wade → free ×4, exit 0. Old-name grep over content/, the batch and the briefs → 0.
• check_svg_ids.py → figures read 70, ids 0, PASS.
• check_key_tells.py → KEY TELLS: 0 of 94 MCQs flagged (a: 0, b: 0, c: 0).
• Blind bundle re-cut: blind/2026-10-02-v2/BLIND_BUNDLE_G4T30.md + KEYS_SEQUESTERED_G4T30.json — 94 MCQ, 15 typed (8 numeric), 1 FRQ, 1 order, 7 adversarial rows; [Drawing: lines 42, grep for summer|winter|spring|fall|autumn|longer|shorter|higher|lower → 0; the order item prints "put these in order, first to last: Path J; Path K; Path L" (the letters carried; the key stays L, K, J, the stem's way).
• Video batch rebuilt (handoff/G4T30/videos/batch_G4T30.json, 6 videos; attached key string A A A C A C B unchanged); check_brief.py: every lesson and the summary clean, the INTRO's standing R79 hit and the 3-question-mark WARN as before; "Parker" 0, "strand" 0. Not sent.

#### Key strings (v2)

• Quizzes, two-option: L01 A A B B · L02 A A B B B (q4 is the new claim; B 3 of 5, at the 60 % line) · L03 A · L04 A B. Three- and four-option: L01 0 1 2 1 2 1; L02 0 2 0 1 1; L03 1 0 1 0 1 2 1; L04 1 0 2 0 2 0. Practice two-option A A A B A B, three-option 1 0 1 2 1. Attached A A A C A C B.
• PP100: 39 four-option MCQs A 10 · B 10 · C 8 · D 11, one three-option (Peoria, C), 1 order, 5 text, 6 numeric; tiers easy 20 · medium 15 · hard 17; per atom 001 14 · 002 11 · 003 12 · 004 15.
• Renumbering (old → new): PP-22–26 → PP-16–20; PP-16–21 → PP-21–26; PP-50 → PP-35; PP-35–49 → PP-36–50; PP-27–34, 51, 52 unchanged.

#### Renders opened (PNG at 1000 px, then the built page)

arcs_two, arcs_headache, path_day, graph_daylight_marked, graph_daylight_seasons, graph_jan_may_pred, q_graph_slc, q_graph_columbus_three, pp_graph_three_t1, pp_graph_three_t2, pr_arcs_mn, pp_sun_curve, pp_table_kc, pr_table_fresno; the mechanical pass (/tmp/g4t30_render.py: getBBox, text overlaps, frame fill) reports no overlap and no text outside its frame on any of the 70 (the static path_anim export's Sun at the origin is the known harmless artefact). Built page /tmp/g4t30_v2.html: the four article routes, the L03 quiz route, the practice and PP100 routes screenshotted at 1000 px and in a true 360 px iframe (/tmp/g4t30_shots/), tiled and read: every figure present, the two-row month axis and the season chips clean on both, the three-bar items showing 1st / 2nd / 3rd, the plateau lines in place.

#### Changed ids

L01-04 · L02 headache, article (a2, a7–a13 figure order, a19, a24–25), script beats 1, 3, 5, 8–9 · L02-04 · L03 article (a15–a16, caption a20) · L03-02, L03-03 (tag field), L03-04, L03-05 (why), L03-07 (tag), L03-09 · L04 article (a17–a19), script beats 1, 2, 6 · L04-07 fb, L04-08 and L04-10 accepts · PR-05, PR-07, PR-09 · PP-16–20 (retag, move), PP-21 (options, stem), PP-27–30 (stems, values; Lincoln → spring), PP-31, PP-35 (retag, move), PP-36 (wrong_fb), PP-37 (re-scened), PP-40–42 (stems, values), PP-43–48 (accepts), PP-44 (run), PP-49 (direction form), PP-51 (Wade) · figures: DAYLIGHT, arc(), W_H, arcs_two, arcs_headache, g30_graph and every graph, arcs_item label / noun, pr_arcs_mn, q_sun_three, pp_sun_curve (new), q_arcs_lm (retired), pp_graph_three_summer → pp_graph_three_spring2, pp_table_kc, pr_table_fresno, briefs · G4T30.py: PP100_THREE_OPTION_SETS, PRACTICE · NAMES_POOL.md row 30 · TYPES_G4T30.md.

#### Owed after v2

The blind read of blind/2026-10-02-v2/ by two cold readers (the type 12 items and PP-49 especially); James's one line on the daylight brackets above; the picture pass (unchanged: three photos, four cartoons) after his sign-off; the videos (batch built, not sent).

v3 — polish (2 October 2026)

Skill: course-build → write-mcqs (every item change; check_key_tells.py at 0) → write-explanation (the changed fb and why lines) → write-like-james (the Needs-you page) @ e23b67f

In plain words. Reader C read the whole v2 bundle cold, 117 items, and agreed every key; no second option stood up anywhere. What was left was five things worth a look, all shape: a quiz summer graph whose last bar dipped one hour, so a child who had just learned "falling = fall" had a foothold; a temperature prediction that reached one month past the June turn the course names; two give-away pairs to write down as draw constraints; and the typed accept lists checked against three adversarial answers per item. Nothing in a script changed, so the video batch stands as built and is not sent.

#### Flag → change

#### Gates after v3

• python3 ~/Documents/CourseBuilding-Vault/skills/write-mcqs/check_key_tells.py grade4/topic30_build/content/G4T30.py → KEY TELLS: 0 of 94 MCQs flagged (a: 0, b: 0, c: 0); KEY LONGEST (any margin) 16 of 94, fixed-set exempt 32.
• python3 build_tools_check_topic.py grade4/topic30_build/content/G4T30.py → 0 FAIL, 23 WARN (the same 23 as v2, answered there; the S36 'cleveland' line now quotes the March-to-May stem).
• python3 grade4/check_g4_articles.py grade4/topic30_build/content/G4T30.py → 0 FAIL; articles 287 / 385 / 265 / 277, scripts 146 / 149 / 129 / 150, intro 105, summary 143, opener ok; L02's 385 WARN stands (James's question 3).
• python3 build_tools_build_topic_site.py grade4/topic30_build/content/G4T30.py /tmp/g4t30_p.html → 4 lessons, 13 practice, 52 pp100, 70 figures.
• python3 ~/Documents/CourseBuilding-Vault/skills/review-figures/check_svg_ids.py grade4/topic30_build/content/figures.py → figures read 70, ids 0, PASS.
• python3 build_tools_names.py Rasmussen Oliver Paige Wade → free ×4, exit 0.
• Renders opened at 1000 px: q_graph_columbus_three (three bars on the 15 line, 1st / 2nd / 3rd) and pp_graph_temp_rising (Mar 50, Apr 60, May 70, orange).
• Blind bundle re-cut: blind/2026-10-02-v3/BLIND_BUNDLE_G4T30.md + KEYS_SEQUESTERED_G4T30.json — 94 MCQ, 15 typed (8 numeric), 1 FRQ, 1 order, 7 adversarial rows, 111 held; [Drawing: lines grep for summer | winter | spring | fall | autumn | longer | shorter | higher | lower → 0; L03-09's line reads "1st, 2nd, 3rd: top on the middle line between 14 and 16"; PP-45's "Mar: top on the 50 line; Apr: 60; May: 70".
• Video batch: no script line changed, so not rebuilt and not sent (handoff/G4T30/videos/batch_G4T30.json as in v2).

#### Counts and key strings after v3

Unchanged from v2: 4 lessons · 39 quiz items (10 + 10 + 9 + 10) · 4 attached + 3 summary questions · 13 practice · bank 52 (39 four-option MCQs A 10 · B 10 · C 8 · D 11, one declared three-option, 1 order, 5 text, 6 numeric); tiers easy 20 · medium 15 · hard 17; per atom 001 14 · 002 11 · 003 12 · 004 15. Every quiz, practice and attached key string as in v2.

#### Changed ids in v3 (for a mini re-read)

L03-09 (bars, wrong_fb, why, accept) · PP-45 (stem, figure months and values, wrong_fb, why) · PR-13, PP-36, PP-37 (accept list only) · PP-40, PP-41, PP-42 (accept list only) · TYPES_G4T30.md (⟂ list, type 14 scene line) · figures: q_graph_columbus_three, pp_graph_temp_rising.

#### Judgement calls, v3

1. PP-41 stays at 14, 15, 15. The flat shape now belongs to L03-09; a bank item may not be a quiz drawing renamed, and the hard tier's rule needs a run that is tallest without being flat.
2. PP-45 moved to March–May → June rather than taking a new word in its stem. The stem already said "afternoon temperatures" twice; the honest fix was the span, not the wording. It now shares its span with PR-09 Fresno (practice, a table, 65 / 75 / 85 → 95); practice and bank are never served together, and the form and data differ.
3. The accept lists were widened, never the grader. Three forms a reader wrote that the shipped grader would refuse are cheap to accept and cost nothing; the grader's rules are course-wide and were not touched.

#### Owed after v3

• James's six questions on the Needs-you page (records/REVIEW_NEXT_G4T30.md → course_preview/review/REVIEW_NEXT_G4T30.html): east and west only; the June/July plateau; L02 as one article; the calendar's season brackets under the class's daylight bars; the three-bar items with months hidden; Wade.
• A mini re-read of L03-09 and PP-45 (blind/2026-10-02-v3/) by a reader who has not seen the keys, if the coordinator wants one; every other key is Reader C's.
• The picture pass (three photos, four cartoons) after James's sign-off.
• The videos: the batch is built and checked; James's notes on the scripts are the cue to send it.

Nothing committed.

Skill: course-build → review-lesson (Role A blind key read, two readers; Role C adjudication by the coordinator) @ fb37d14

Topic 30 — The Sun's path and the year's pattern — blind read, 2 October 2026

One key the topic's own rule contradicts, and both readers caught it. PR-09 gives April 13, May 14, June 15 hours and asks July; the key is "about 16" (straight continuation), but L04-07 and PP-51 teach that daylight turns shorter after June — the key rewards what the course calls wrong. PP-48 (same run, asks August) is keyed "about 14" by the turn rule, so the two items contradict each other. Otherwise every MCQ key agreed by both readers (111 rows each; MCQ disagreements 0); PP-18's order L, K, J matches the stored key (stem's way, shortest to longest); numeric typed answers all on list; no key tells (the checker reads 0 of 94). The intro poll is not bundled by design.

Flags carried into the fix pass

Readers' files: blind/2026-10-02/READER_A.md, READER_B.md.

Skill: course-build → review-lesson (Roles B, B2) · review-questions · review-accuracy @ fb37d14 — non-author review, filed by the coordinator from the reviewer's report

Topic 30 — guidance, questions and accuracy review, 2 October 2026

The teaching is sound and the science honest; the item sets are not yet testing the atom they claim. The four lessons say exactly what Grade 4 may say; the Springfield records round correctly from real 40°N daylight (Jun 14.9 h, Jul 14.7 h → 15 / 15); every figure opened is clean with east at the left. But most bar-graph season items are answerable from the month labels alone, three bank items reuse quiz datasets town-for-town, and the lesson text says "shorter each month after June" while its own graph shows June = July. Two record claims off: L02's longest run is 6 (≤8); "seems to move across the sky" is not Grade 3's wording (Grade 3 says "comes up" / "goes down" — fine).

Key-tell checker: KEY TELLS: 0 of 94 MCQs flagged (a: 0, b: 0, c: 0).

Ruling on the bar-graph season items (coordinator adopts it): three-bar items (types 11 and 12; L03-04/05/09; PR-07; PP-27–30, 39–41) hide the month names in every instance — stem "three months in a row", bars labelled 1st, 2nd, 3rd — with mid-range values (11-12-13 rising = spring; 13-12-11 falling = fall; 9-9-10 = winter; 15-15-14 = summer). Full-year items (type 10; L03-02/03) keep their months as exam-fidelity scaffolds, retagged easy, with data that agrees with the calendar.

Essentials

1. PR-09 key contradicts the turn rule → March–April–May → June. 2. PP-48 → a direction item (longer / the same / shorter than June). 3. L03-02's bars (Feb 10, Nov 11). 4. Bank reuses quiz datasets: PP-31 = L04-08; PP-40 = L03-09; PP-41 = L03-05. 5. Type 11/12 month labels (ruling above).

Worth fixing

The June/July plateau sentences; "looked up when the Sun came up"; "still high" at 6 pm; "standing at its highest"; PR-05 prints "Sun paths" above typed PR-13; no claim item on "same path every day" (replace L02-04); L03-07 tagged misconception on a TRUE claim; PP-50 hard for one step; quiz accept lists vs num_accept; PP-35's wrong_fb mentions Earth's orbit; ⟂ PP-32 ⟂ PP-45 ⟂ summary Q3.

Leave

Paige/Parker — BUT the coordinator rules a rename anyway (Parker → Casey, free course-wide), as in Topics 23, 25, 28, 29, 34. "A question for a later grade". L01-05 ↔ L01-08. PP-21. Hawaii omitted. Topic 25's labels sentence. "The unit in the heading, written once".

Author's questions: (1) east and west only — yes. (2) keep the plateau — yes, and fix the three sentences that contradict it. (3) L02 one article — yes.

The sentence Topic 31 L05 recalls, word for word: "Every day, the Sun seems to move across the sky on a curved path. It comes up in the east, is highest around midday, and goes down in the west."

Skill: review-figures @ fb37d14 — non-author figure review with the page pass, filed by the coordinator from the reviewer's report

Topic 30 — figure review, 2 October 2026

What a student notices first: the yard pictures are right — east left, west right, the Sun low over the parking lot, highest over the gym but a third of the frame down, low over the trees; June's orange arc higher and wider than November's blue one over the same landmarks. But at L03 the bars stop doing the work: every season item carries the month names, and the class's own daylight data does not match its brackets.

Must fix before a cold read

1. The teaching dataset contradicts the taught rule. DAYLIGHT = Jan 10, Feb 11 … Nov 10, Dec 9; May 14 = Aug 14. On graph_daylight_seasons the WINTER bracket sits under Jan/Feb/Dec but the three shortest bars are Dec, Jan, Nov; SUMMER's third month ties May. The L03 lines "The shortest daylight bars, December, January and February" / "The tallest … June, July and August" are false of the drawing beside them. Reused as q_graph_slc (L03-02) and pr_graph_sacramento. Fix the array once (Feb 10, Nov 11; Aug 13).
2. pp_graph_milwaukee (PP-22) Feb 10 = Nov 10 — lower Feb to 9 or raise Nov.
3. pp_graph_albuquerque (PP-25) Nov = Dec = Jan = 10 — make Nov 11.
4. Three-bar season items keyed by month names while the bars point the other way. pp_graph_three_t1 (Dec 9 / Jan 10 / Feb 11, typed "winter"): by the direction rule rising bars = spring; pp_graph_three_winter (9, 9, 10) rises; q_graph_columbus_three, pp_graph_three_t2, pp_graph_three_summer (15, 15, 14) fall = fall by the rule. See the guidance review's ruling (months hidden on three-bar items; mid-range values).
5. Check-figure twins. q_graph_slc = pr_graph_sacramento = teaching graph_daylight_plain; q_graph_columbus_three = pp_graph_three_t2 = pp_graph_three_summer; q_graph_nash_three = pp_graph_three_t3. Fresh instances.
6. arcs_headache unplaced; arcs_two (with SUMMER/WINTER chips) sits before the frames teach it. Place arcs_headache first; move arcs_two after.

Page pass

Desktop passes (month labels 17.9 px). Phone (forced 360 px, 278 px column): a 560-wide graph scales ×0.496 → month labels 6.0 px; three-bar graphs 7.3–8.8; season chips / "house" / "parking lot" 5.0 px. The 560-wide graph family needs a phone variant (fewer gridlines, 16 px fonts). check_svg_ids.py: 70 figures, 0 ids, PASS. Bar count: all 33 graphs parsed; every bar height matches its data array — the flaws are in the data. path_anim shows the Sun at the origin in a static export (harmless live).

Month axis

Hands over the season on every season item. Hiding the names makes the bars do the work on the 12-bar graphs; on three-bar items it does not unless the values are mid-range (guidance ruling: 11-12-13 rising = spring; 13-12-11 = fall; 9-9-10 = winter; 15-15-14 = summer, bars labelled 1st, 2nd, 3rd).

Briefs, unplaced, minor

All seven briefs complete bar "the eye must see first"; portland_cartoon says "red bars" — TEMP renders orange. Unplaced: arcs_headache. Minor: arcs_two WINTER chip covers the arc's descent and the gym's corner (W_H ~62); graph_daylight_marked SHORTEST chip over Nov's bar; path_day middle arrowhead pokes out of the MIDDAY Sun. pp_arcs_jkl one-list passes. Item alts and blind comments: numbers only.

Best: path_day; arcs_two; graph_jan_may_pred. Weakest: graph_daylight_seasons; pp_graph_three_t1; q_graph_slc.

Skill: review-lesson @ e23b67f (Role A — blind key reader, Reader C, cold read of BLIND_BUNDLE_G4T30.md v2 only)

Takeaway. Every item could be answered from the stem and the described drawing alone, and on every MCQ exactly one option stood up. The three-bar season items now read cleanly from the bars' direction and level: rising through 12 is spring, falling is fall, flat near the bottom of the 0-16 axis is winter, flat near the top is summer. Two of the flat-high items carry a one-hour wobble (L03-09 falls 15, 15, 14; PP-41 rises 14, 15, 15); I still read both as summer because the bars sit at the top of the axis, but a child who has just learned "falling = fall" has a foothold on L03-09. Every full-year key agrees with the bars described. Every prediction item has one honest next number; the only daylight run that crosses June is PP-49, which is built to test the June turn and is keyed by the rule the course teaches. No item asks why the path or the daylight changes. No term arrives before its coining item. Oliver, Paige and Wade are three different shapes on the page; none can be read as another. Nothing fails; five items are worth a look.

Adversarial answers (typed items)

The grader should accept "autumn" for fall and "path" on its own for the Sun's path, with or without the apostrophe.

Cross-item notes

• Three-bar items (L03-04, L03-05, L03-09, PR-07, PP-27 to PP-30, PP-40 to PP-42): the season reads from direction and level alone; month names are not needed. Eleven items, eleven single defensible seasons. The two flat-high items wobble one hour (L03-09 ends with a fall, PP-41 opens with a rise); PP-29 and PP-40 wobble at the bottom but read as winter without doubt because nothing else is that low. If one of the two flat-high items is to be hardened, L03-09 is the one: 15, 15, 15 or 14, 15, 15 (mirroring PP-41) removes the "falling = fall" foothold.
• Full-year items (L03-01 to L03-03, PP-16 to PP-20): every key agrees with the bars described; the shortest/tallest three are always separated from the fourth by at least 5 units.
• Prediction items: each has one honest next value. No daylight run crosses June except PP-49, which tests the turn deliberately. PP-45 is a temperature run from June to July; the course's full-year graphs make July warmer, so the key is honest, but it sits one month past the daylight turn the course names, and PP-51 is in the same bank.
• Giveaways within a draw: PP-51's key states the June turn that PP-49 tests; PP-35's stem states that Dec, Jan, Feb are the shortest bars, which answers PP-16 and PP-19's "which season" the same way; L03-07's claim states the rule L03-03 tests. None hands over a specific key letter, but a draw constraint (not in the same PP100 draw: PP-49 ⟂ PP-51; PP-35 ⟂ PP-16, PP-19) would be tidy.
• Why items: none asks why the path or the daylight changes. PP-39's "correct reason" is the taught link (lower winter Sun, longer shadow), not a cause of the path's change.
• Terms: "the Sun's path" coined in L01-10 before PR-12/13 and PP-36/37 use it; "pattern" coined in L04-01; "season", "daylight", "midday", "fall" are everyday words for a US nine-year-old. Nothing unmet.
• Names: Oliver, Paige, Wade — three initial letters, three lengths; no pair could be read as one child.
• Exam-like wording: the midday "where is the Sun" items (L01-05, PP-07, PP-21) always carry "straight overhead" as the tempting-wrong; a student who meets it three times in one draw learns to avoid it by habit, but that is the intended lesson.

Would not ship as is: none. Worth a look before shipping: L03-09 (wobble tempts fall), PP-41 (wobble tempts spring, faint), PP-45 (temperature prediction one month past the taught June turn), PP-49 ⟂ PP-51 and PP-35 ⟂ PP-16/PP-19 as draw constraints.