Skill: course-build → atomise (architecture steps 2 and 6c; standing audits; the no-standard sweep) → write-like-james → review-page @ 73121d6
MS Scientific Reasoning — course architecture v1 (15 October 2026)
Status: FOR YOUR READ. Every card below carries its default in force; the build runs on those defaults (your 9 Oct ruling: open rulings never block the build). The three sample topics are being authored from this page as it stands; nothing else is authored until you have read them.
Read this in five minutes. It arranges the knowledge graph's grade 6–8 Scientific Reasoning strand into one course. Sixty atoms are in the strand; twenty-four are taught inside MS Physics and MS Earth and Space and never here (your 13 Oct ruling: one atom, one home); the thirty-six that remain become seven topics of four to six lessons, one test at the end. Nothing below writes a lesson sentence. The atoms stay in the knowledge graph; one register edit the sweep supported is made and logged there.
The takeaways.
- Nothing is cut. Every one of the thirty-six atoms is named by Texas or Florida at grades 6–8; the two that carried only an NGSS practice code (what you observed; observation or interpretation) now also carry the Florida benchmark that names them, verified against its text.
- One course, seven topics, thirty-six video lessons, video-led, in the flow you fixed with David today: intro → lessons with their quizzes → one summary video → PP100. A short article only where a results table or graph must stay on screen.
- The course follows the shape of an investigation: what you observed and the question you can test; the three kinds of variable, in the student's words before their names; fair test or not; one result is not enough; from cases to a rule; together is not the same as caused; how science checks itself.
- Your investigation rulings from Grades 4 and 5 bind every lesson (applied, not asked): a fair test is taught as a rule that judges cases, by an example sequence with near-miss non-examples; the thing you change and the thing you measure are named in the student's own words before "independent variable" and "dependent variable" are coined; the control group keeps Grade 5's words.
- Six cards: the course's name and shape; whether investigations draw on all three subjects; the anchor forms for the exit test; whether three Life-pointing atoms move to MS Life Science; the three sample topics; one missing atom (what a scientific law is).
Decisions I need from you
Each card: the default in force, why it is a question, my recommendation, one tap. A card is here only where your answer changes what a student learns or what you review.
1. Is this one course across grades 6–8, called Middle School Scientific Reasoning?
The course says: one course, "Middle School Scientific Reasoning", by default — the knowledge graph's standing call (five subject courses), which you approved for MS Physics on 13 Oct.
- The engineering-design half of the strand's name is gone from this course: all twelve Design and build atoms live in MS Physics (your 14 Oct tap), so "and Engineering Design" would promise content this course does not hold.
- A grade split would be arbitrary here: the TEKS process standards are worded identically at grades 6, 7 and 8, so no atom carries a grade stamp that could cut the course into grade shells.
- No state tests middle-school science before grade 8, so a per-grade course would have no official instrument to anchor its exit test to.
Cost of changing it: a rename is one line; a three-course cut has no standard to cut along.
My recommendation: one course, Middle School Scientific Reasoning.
A: One course, Middle School Scientific Reasoning / B: Another name or shape (say which)
2. Do the investigations in this course draw their cases from all three subjects?
The course says: yes, by default: every lesson's scene is a real investigation from physics, chemistry or life science, one named subject per lesson, so a reasoning atom is seen to travel; across a topic the three subjects each appear.
- A reasoning atom taught on one kind of scene is learned as a fact about that scene. A fair test judged on a seedling, a dissolving sugar cube and a rolling cart is learned as a rule about tests.
- The released grade-8 items ask the practice inside a content scene (a data table from a plant experiment, a trial log from a ramp); neutral everyday scenes (which detergent, which paper towel) would not rehearse that.
- The cost is coordination: a chemistry or life scene here must say nothing a sibling course has not yet taught or says differently; each scene is checked against that course's lexicon and kept to what a grade 6 student already knows.
Cost of changing it: none now; the sample topics are written to the default and would be re-scened.
My recommendation: all three subjects, one named subject per lesson.
A: All three subjects, one per lesson / B: Neutral everyday scenes only
3. Do the same released grade-8 forms anchor this course's exit test as anchor MS Physics's?
The course says: yes, by default: STAAR grade 8 science 2022 (PDF) and the 2023–2025 online released forms; Florida's grade 8 Statewide Science Assessment item specifications and sample test; the AzSCI grade 8 sample test and scoring guide — this time the PRACTICE items only: Florida's Nature of Science reporting category outright, and the STAAR and AzSCI items whose work is the practice (reading a trial table, naming the variable, judging a claim against data).
- Florida tests the nature of science as its own reporting category at grade 8; Texas and Arizona embed the practices in content items. So Florida's items are the richest anchor and the other two forms are read for item shapes.
- The forms are fetched when the test is authored (the state sites refuse automated download; they are fetched by hand); an examiner's record names which released item shape each test item mirrors, and every test item is written fresh.
- One consequence to know: the same fetch settles which words the released items use for the variables (Florida's benchmark writes "test variables (independent variables)"), and the coining in Topic 2 follows the items, per your 9 Oct ruling.
Cost of changing it: none now; the anchor set is read at the test.
My recommendation: the same anchor set, practice items only.
A: Same anchor set as MS Physics, practice items / B: A different set (say which)
4. Do all thirty-six atoms stay here, or do the three that point at life science move into MS Life Science?
The course says: all thirty-six stay here, by default; MS Chemistry and MS Life Science host no reasoning atom today.
- Three register cross notes name the Life Science strand: SR-026 design a fair test, SR-051 together is not the same as caused, SR-063 bigger sample, safer conclusion. No cross note names the Matter strand on an unhosted atom.
- Under card 2 those three lessons take a life-science scene here (a seedling fair test; a claim about screen time and sleep; a survey of one class against a thousand students), which is what the cross notes ask for, without a second home.
- Moving them would take Topic 3's capstone (the designed fair test) and Topic 6's centre out of the only course whose job is to teach them; the MS Life Science chat has been asked whether any investigation of theirs needs one, and has not claimed any.
Cost of changing it: three lessons move and two topics here are re-cut; MS Life Science then owes the lessons and their recall items.
My recommendation: keep all thirty-six here.
A: Keep all thirty-six here / B: Move the three Life-pointing atoms to MS Life Science
5. Are the variables, one-result-is-not-enough, and together-is-not-caused topics the right three samples?
The course says: Topics 2, 4 and 6 are being authored first, by default, as the three "quite different" samples you asked for today.
- Topic 2 (the three kinds of variable) is a coining topic: three happenings named in the student's words, each then given its standard's name — the process-and-its-name split you ruled on 30 Sep, three times over, and the words the released items use.
- Topic 4 (one result is not enough) is a data topic: results tables and graphs drawn by code, the odd result, a pattern that is not a guarantee, repeats; it also shows how a lesson here recalls a hosted MS Physics atom (describe the pattern) in one sentence.
- Topic 6 (together is not the same as caused) is a judgement topic with the band's best-known reasoning misconception, confronted on documented cases (ice cream and drowning; books at home and school results).
- Alternatives that are also "different": Topic 3 (fair test or not, your own example-sequence shape), Topic 1 (observation or interpretation), Topic 7 (theory is not a guess).
Cost of changing it: a sample already authored is kept as a finished topic; a swap costs one more topic's authoring before your read.
My recommendation: Topics 2, 4 and 6.
A: Topics 2, 4 and 6 / B: Swap one (say which)
6. Does the course gain one atom for what a scientific law is?
The course says: no new atom yet, by default: the register has none, and a register addition is yours to rule.
- TEKS 6.1H reads "Distinguish between scientific hypotheses, theories, and laws", and Florida SC.6.N.3.2 reads "a scientific law is a description of a specific relationship under given conditions". The strand has an atom for the hypothesis (SR-012, hosted in MS Physics) and one for the theory (SR-080, Topic 7 here) and none for the law.
- Without it a student meets "law" only inside MS Physics's Newton's-laws topics, where it is never explained; a released item asking which is a law and which a theory has no lesson behind it.
- Proposed: SR-080b "A law describes, a theory explains" (Fact; state that a scientific law describes what always happens under given conditions and a theory explains why), placed in Topic 7 right after SR-080, six bank items, TEKS 6.1H and Florida SC.6.N.3.2.
Cost of changing it: one lesson in Topic 7 and one register edit; none if you say no.
My recommendation: add the atom.
A: Add SR-080b, a law describes / B: Leave the register as it is
Applied without asking
Your rulings on investigation lessons from Grades 4 and 5 (each is in the vault; listed here as applied, not asked):
- A fair test is a rule that judges cases (7 Oct): Topic 3 opens with an example-sequence lesson on "Is this a fair test?" in your 9 Oct shape — a fair test (only the ramp height changes), a minimally different unfair one (the height and the car), another unfair one, a very different fair one, a minimally different fair one — quizzed "is this a fair test? yes or no", with "and why" as the hard items of the PP100. The design-a-fair-test routine is a separate lesson after it.
- The student's words before the term (Grade 4 register, your 9 Oct coining ruling): Topic 2 names the thing you change, the thing you measure and the things you keep the same first, and coins "independent variable", "dependent variable" and "control variables" on them; each happening and its name are two atoms and two kinds of item (30 Sep). The coining follows the words the released items use (card 3's fetch confirms them); Florida's own words "test variable" and "outcome variable" are said once as the test's words, never coined.
- The control group keeps Grade 5's words (9 Oct): "In an experiment, the group that gets none of the thing you're changing is called the control group. It is kept for comparison." Topic 3 recalls it; SR-025's lesson adds why the comparison is needed in a test with no group at all.
- "Fair test", "observation", "criterion", "constraint" and "control group" are Grade 4 and 5 coinings and are recalled in their lexicon lines, never re-coined; "hypothesis" is MS Physics Topic 11's coining and is recalled.
- Evidence supports that an idea is true (your Grade 5 wording), and the evidence is a measurement or an observation, never the thing itself; an investigation measures with an instrument a student can read (13 Oct).
- Video-led, the quiz cap, one summary video, the PP100 bank, no mixed practice, no strand opener (13–15 Oct): every lesson one video of at most 180 spoken words ending "Okay, now your turn." with one attached question; a quiz capped at ten, five or fewer usual, an atom with several cases tests every case; one summary video under two minutes with its three questions on the page; a PP100 bank of six items per atom; no separate practice stage; no course opener video.
- Numbers and stems (9–15 Oct): a space for thousands, one unit through a comparison, one set of numbers per lesson; a stem never says "X is drawn" without the drawing; a two-way verdict is an MCQ.
Decided here and recorded, not asked (your answer either way would not change what a student learns): topic titles name the thing; Topics 4 and 5 each mix two of the register's sub-domains, because "one result is not enough" (SR-060) is built on "the odd result" (SR-033) and "bigger sample, safer conclusion" (SR-063) on both, so the prerequisite order beats the sub-domain labels; Topic 2 lists six atoms and its author decides whether each happening and its name share one video (the vault allows one lesson, two tags, six bank items per atom); Topic 1's "what you observed" and "observation or interpretation?" are the pattern you struck in MS Physics Topic 1 (a sort obvious on its own, its contrast the next atom), so the author may fold them into one lesson and records it; one static test of three forms at the end, because thirty-six lessons under one test is the size of MS Physics's Motion strand; the hosted atoms this course recalls are recalled in the host's lexicon words (Card C), and where the host's lesson is not yet written the recall line is agreed with that chat before the lesson ships; the arrow palette is MS Physics's (a graph here is drawn with its dt_graph helper so graphs look the same across courses).
The no-standard sweep (your 13 Oct ruling: "If no state requires it, then we shouldn't be including it")
The rule applied: an atom stays only if TEKS, Florida or Arizona names its content at grades 6–8, or NGSS names it inside the endpoint the kick-off set (the eight practices SEP-1 to SEP-8 and MS-ETS1). Every atom's standards field was read; the full table is in records/ARCHITECTURE_CHECK_2026-10-15.txt, and build_tools_check_architecture.py fails if a placed atom has no qualifying code.
Cut: none. Every one of the thirty-six unhosted atoms carries a TEKS or Florida code at grades 6–8.
Two atoms carried an NGSS practice code only, and now carry a verified state code (edit log, 15 October): SR-001 "What you actually observed" and SR-002 "Observation or interpretation?" cited SEP-3 alone. Florida SC.6.N.1.6 reads "Explain how scientific inferences are drawn from observations and why it is important to use multiple trials"; telling the observation from the inference is its first half, so both atoms now carry it (and SR-003, MS Earth and Space's, and SR-060, whose content is the benchmark's second half); SR-001 also carries TEKS 6.1E, "collect … qualitative data as evidence". They were kept either way: SEP-3 is inside this course's endpoint.
Register findings from the sweep (stated once, not filled):
- Arizona names no practice on its own at grades 6–8; every Arizona 6–8 code embeds the practice inside a content standard. Zero of the thirty-six atoms carries an Arizona code, by construction, not by omission.
- Florida SC.6.N.1.3 ("the difference between an experiment and other types of scientific investigation") is cited on five evidence-strength atoms (SR-005, 006, 032, 042, 063) whose content does not match it, and no 6–8 atom carries its actual content (Grade 5's SR5-006 does). Each of the five keeps a TEKS code, so nothing is cut; the citations are flagged for the three courses that host them, not edited here.
- What a scientific law is has no atom (card 6).
- SR-006 and SR-060 are near-duplicate behaviours (ten trials beat one; one result could be a fluke). SR-006 is MS Physics Topic 11's; Topic 4 here recalls it and teaches the fluke, not the comparison.
Card A — The course and its order
The call: one strand, seven topics in the order of an investigation, one static test of three forms at the end, also the end-of-course test.
| Part | Topics | Lessons | Why it sits here |
|---|---|---|---|
| Asking and testing | 1–3 | 16 | What you observed and the question you can test; the three kinds of variable; fair test or not, and how to design one. Everything after needs "a fair test" and the three variable names. |
| Reasoning from results | 4–7 | 20 | One result is not enough; from many cases to a rule; together is not the same as caused; how science checks itself. Each leans on the data-reading atoms MS Physics teaches (read a value, describe the pattern, does the data support the claim), recalled in one sentence. |
Total 36 lessons; 24 further atoms of the strand are hosted elsewhere (Card C).
Card B — The topics
The call: seven topics of four to six lessons, each with an intro video, a quiz after every lesson, one summary video with its three questions, and a PP100 bank of six items per atom; no mixed practice. Every atom is this course's own; the atoms it recalls from MS Physics and MS Earth and Space are in Card C.
| # | Topic (student-facing title) | Atoms (in teaching order) | Lessons | TEKS stamp | The big idea it serves |
|---|---|---|---|---|---|
| 1 | What you observed, and the question you can test | SR-001, 002, 010, 011, 013, 014 | 6 | 6.1A, 6.1E, 6.3C | Say only what you saw, heard or measured; sort a statement into what was observed and what was added to it; a question evidence can answer against one opinion or definition settles; sharpen a vague question into one you can investigate; a claim that could be proven wrong; the result that would count against it |
| 2 | The thing you change, the thing you measure, the things you keep the same | SR-020, 020b, 021, 021b, 022, 022b | 6 | 6.1B | In the student's words first, then the standard's name: the one thing changed on purpose is the independent variable; the thing measured to see what the change did is the dependent variable; the things kept the same so they cannot affect the result are the control variables |
| 3 | Fair test or unfair test, and how to design one | SR-023, 024, 025, 026 | 4 | 6.1B, 6.2D, 7.1B, 8.1B | Is this a fair test? (an example sequence with near misses); why changing two things at once tells you nothing; the group you compare against (the control group, Grade 5's words); design a fair test by naming the independent, dependent and control variables |
| 4 | One result is not enough | SR-033, 034, 060, 061 | 4 | 6.2B, 6.5A, 6.1E | The odd result that does not fit; a pattern in past results makes an outcome likely, not certain; one result could be a fluke; repeat it and see whether it holds |
| 5 | From many cases to a rule, and the three kinds of reasoning | SR-041, 042, 063, 040, 044 | 5 | 6.5A, 6.2B, 6.4B, 8.3A, 8.3C | Form a rule from cases that all behave the same way; why more, and more varied, cases make the rule safer; a bigger sample gives a safer conclusion; a conclusion that must follow from its statements; sort a piece of reasoning into rule-to-case, cases-to-rule, or clues-to-best-explanation |
| 6 | Together is not the same as caused | SR-050, 051, 052, 053, 054, 055 | 6 | 6.5B, 6.2B, 6.1B, 6.3C, 6.4B | Two things that change together; together is not the same as caused; a third thing behind both; which way the cause runs; what a fair test of a cause would need; spot the conclusion that claims more than its evidence supports |
| 7 | How science checks itself | SR-064, 080, 083, 084, 085 | 5 | 6.3C, 6.4B, 6.1H | Spot the cherry-pick; "theory" in science means a well-supported explanation, not a guess; other people check a result before it is accepted; what science can and cannot answer; what makes an answer scientific rather than authority, tradition or preference |
| End-of-course test (three forms) — the whole course, anchored to the released grade-8 practice items (card 3) |
Totals: 36 lessons in 7 topics; 7 PP100 banks; 1 static test of three forms after Topic 7. If card 6 is tapped, Topic 7 becomes 6 lessons and the course 37.
Hosted elsewhere, never taught here (James, 13 and 14 Oct 2026; MS Earth and Space's five proposed 16 Oct): SR-003, SR-004, SR-005, SR-006, SR-012, SR-030, SR-031, SR-032, SR-043, SR-062, SR-081, SR-082, SR-090, SR-091, SR-092, SR-093, SR-094, SR-095, SR-096, SR-097, SR-098, SR-099, SR-100, SR-101.
Cut on James's ruling (13 October, the no-standard sweep): none.
Why this grouping
- Every prerequisite edge points backwards in this order, into grades 3 to 5, or at a hosted atom (Card C). Checked atom by atom today (
records/ARCHITECTURE_CHECK_2026-10-15.txt); zero forward references. - Topic 1 joins the two observation atoms to the four question atoms because a testable question starts from what was observed (SR-010's prerequisite is SR-001), and "could this be proven wrong" is the test of the question before any variable is named.
- Topic 2 is the course's coining topic and stands alone so that Topic 3's fair-test judgement can use the three names from its first lesson.
- Topic 3 holds the fair-test rule, the why, the control group and the design routine, in that order, on your 7 and 9 Oct rulings: the example-sequence lesson first, the routine that runs one last.
- Topics 4 and 5 follow the register's prerequisite chain rather than its sub-domain labels: the odd result leads to "one result is not enough", which with "why more cases make the rule safer" leads to "bigger sample, safer conclusion". The three kinds of reasoning close Topic 5 because the sort needs the rule-from-cases lesson before it.
- Topic 6 is the register's correlation sub-domain whole; its fair-test-of-a-cause lesson (SR-054) needs Topic 3's design routine, and its overclaim lesson (SR-055) recalls MS Physics's "does the data support the claim?".
- Topic 7 gathers how science checks itself: the cherry-pick (which needs the overclaim and the repeat), the theory, other people's checks, and the boundary of what science can answer.
What changes: the folder gets topicNN_build/ per topic as each is built; Topics 2, 4 and 6 are briefed first (card 5), one agent per topic, no blueprint, at most two agents at once.
Card C — The strand's atoms taught elsewhere, and where this course recalls them
| SR atom | Home | Recalled here by |
|---|---|---|
| SR-012 A hypothesis is a testable prediction | MS Physics Topic 11 | Topic 1 (could this be proven wrong; what result would count against it), Topic 2 (the thing you change) |
| SR-006 Why one piece of evidence is stronger | MS Physics Topic 11 | Topic 4 (one result is not enough) |
| SR-030 Read a value off a chart; SR-031 Describe the pattern | MS Physics Topics 4 and 5 | Topics 4, 5 and 6 (every results table and graph) |
| SR-032 Does the data support the claim? | MS Physics Topic 15 | Topic 6 (spot the overclaim) |
| SR-062 Someone else gets the same answer; SR-081 Evidence can change the answer | MS Physics Topic 16 | Topic 7 (checked by other people) |
| SR-003 An inference goes beyond what you saw; SR-004 More than one explanation can fit; SR-005 Stronger and weaker evidence | MS Earth and Space Topics 4, 5 and 9 (proposed) | Topic 5 (a conclusion that must follow) |
| SR-043 The best explanation for the clues | MS Earth and Space Topic 13 (proposed) | Topic 5 (match the reasoning to its kind), Topic 7 ("theory" doesn't mean "guess") |
| SR-082 Provisional is not the same as worthless | MS Earth and Space Topic 6 (proposed) | not recalled here |
| SR-090 to SR-101 Design and build | MS Physics Topics 13 and 18 | not recalled here |
A recall is one sentence in the host's lexicon words and one recall item, never a re-coining. Where the host's lesson is not yet written (MS Physics Topics 11, 15 and 16; the Earth and Space five), the recall line is agreed with that chat before the lesson here ships.
Card D — What is taught in Grades 3–5 and retrieved here
The call: the prerequisite atoms that live in the grade 3–5 courses are not taught here. Each lesson that leans on one states the idea in a sentence and asks one recall item before using it (the stand-alone entry rule: no student has taken those courses yet).
| Topic | Leans on (grade 3–5 atoms) |
|---|---|
| 1 | SR3-002 observation or idea about it; SR4-001 a question you can test |
| 2 | SR4-002 the one thing you change; SR4-003 keep everything else the same; SR4-004 what you measure to see the result |
| 3 | SR4-005 fair test or unfair test; SR5-002 plan the whole fair test; SR5-004 the group you leave alone (the control group) |
| 4 | SR4-012 the result that doesn't fit; SR5-003 run it more than once |
| 7 | SR5-007 someone else gets the same answer (through SR-062) |
The full lean list, atom by atom, is in the check record. Grade 4 and 5's coined words (fair test, observation, control group, prediction) are reused word for word from their lexicons.
Card E — Tests and benchmarks
The call: a PP100 bank per topic (six items per atom, drawn from); one static test of three forms after Topic 7, the end-of-course test, anchored to the released grade-8 practice items (card 3). Every test item is written fresh for the test, a non-author examiner re-performs the freshness claim on scenes, and every form is read cold with keys hidden.
- Topics carrying the heaviest tested content, from the state codes on their atoms: Topic 3 the fair test (TEKS 6.1B at all three grades, 6.2D, Florida SC.6.N.1.1 and SC.7.N.1.4); Topic 2 the variables (6.1B, SC.7.N.1.4); Topic 6 cause and correlation (6.5B on every atom); Topic 5 (6.5A, 6.2B, 6.4B, 8.3A); Topic 7 (6.3C, 6.4B, SC.6.N.2.1, SC.6.N.3.1). The per-topic counts are in the check record.
- Anchor set (card 3): STAAR grade 8 science 2022 (PDF) plus the 2023–2025 online released forms, Florida's grade 8 Statewide Science Assessment item specifications and sample test (the Nature of Science category), the AzSCI grade 8 sample test and scoring guide. Fetched by hand when the test is authored.
- The PP100 banks carry the "and why" items: a fair-test verdict is a yes-or-no quiz item after the lesson and a verdict-with-its-reason item in the bank (your 9 Oct ruling).
Card F — Register observations, one edit made and logged
- SR-001, SR-002, SR-003 and SR-060 now carry Florida SC.6.N.1.6, and SR-001 TEKS 6.1E (edit log, 15 October; the sweep section above).
- Florida SC.6.N.1.3 is mis-cited on five evidence atoms and its own content (experiment against observation study) has no 6–8 atom; flagged for the hosts, not edited.
- No atom for a scientific law (card 6).
- SR-006 and SR-060 are near-duplicates across two courses; recorded, taught on different moves.
- SR-013 "Could this be proven wrong?" carries no TEKS code; 6.1H (hypotheses, theories, laws) and 6.3C (argumentation with empirical evidence) are near it; left blank as a finding.
- SR-044 "Match the reasoning to its kind" carries grade-8 TEKS codes only (8.3A, 8.3C); its three kinds are taught in Topic 5's own lessons and in MS Earth and Space's SR-043, which it recalls.
- SR-020b, 021b, 022b cite Florida SC.7.N.1.4, whose own words are "test variables (independent variables)" and "outcome variables (dependent variables)"; the coining follows the released items (card 3), the other words said once.
- SR-085 "Science vs other ways of knowing" is a Fact atom whose behaviour is a sort; its lesson states the one feature (an answer that evidence could overturn) and sorts cases on it.
Checks run (15 October 2026, build_tools_check_architecture.py, output in records/ARCHITECTURE_CHECK_2026-10-15.txt)
- Atoms placed: 36; hosted elsewhere: 24; cut: 0. Every register atom appears in exactly one topic or on the hosted line; none is left over; none is doubled.
- Prerequisites: every prerequisite of every placed atom sits in an earlier topic, earlier in the same topic, in grades 3 to 5, or is a hosted atom (listed as a lean). Zero forward references.
- The no-standard sweep: zero placed atoms without a qualifying code; zero kept on a pre- code; zero kept on NGSS alone after today's edit.
- Topic sizes: largest 6 (Topics 1, 2, 6), smallest 4 (Topics 3, 4).
- Standards coverage: every TEKS and Florida code the thirty-six atoms carry is carried by a placed atom.
About this build (kickoff record)
- Register:
Science Knowledge Graph/data/atoms/06-08/scientific-reasoning.json, pulled 15 October 2026. Per-course choices are the register's: K–8 letter-suffixed ids with gaps left for insertion; Fact / Categorical / Transformation / Routine; bare behaviours; verdict and reason split by default; frameworks TEKS, NGSS, Florida, Arizona. - Endpoint: the grade-8 STAAR, Florida SSA and AzSCI practice items (Florida's Nature of Science category outright); NGSS SEP-1 to SEP-8 and MS-ETS1 (the kick-off's endpoint; the MS-ETS1 atoms are MS Physics's).
- Sources read: the kick-off notes; the vault's CLAUDE.md, 00 Protocol, the ExplAIn Yourself principles, the PP100, video, summary and article decisions and the five rules the kick-off names; the knowledge graph's HANDOFF, CLAUDE and decisions files and the full register; MS Physics's CLAUDE.md, architecture, authoring brief and its Topic 5 chain end to end; MS Earth and Space's architecture; the Grade 5 session start and systemic-issues records; your Grade 4 and 5 rulings of 2, 5, 7 and 9 October on fair tests, criteria, constraints and the control group; the Grade 3–5 reasoning registers and the Grade 4 and 5 investigation and design lexicons. Nothing from any sibling course's lesson text.
What happens next
The reasoning-misconceptions table is written first (records/REASONING_MISCONCEPTIONS.md: "an inference is what you saw", "one trial proves it", "together means caused", "a theory is a guess", "the control group is the one you change", with the wording that plants each and the case that breaks it), then the folder gets topicNN_build/ for Topics 2, 4 and 6, one agent per topic from records/AUTHORING_BRIEF_MSSR.md, at most two at once: scripts, quizzes, the summary, the PP100 bank with its type sheet, the full gate chain to 0 FAIL, a non-author review, a blind read, then the three topics on the flat scripts page for your read. The rest of the course waits for that read. The first video batch waits for your word before it leaves the machine.