← All topics (review home)

Skill: write-like-james @ 8a96578

Why the same faults kept appearing: root causes and the fixes now in force (7 October 2026)

Your review of the eleven first topics found faults that repeat. Each one traces to a rule that lived only as prose, with nothing mechanical checking the text, or to a rule written from one failure and applied too widely. The fixes below are now in the skills and, where a machine can see the fault, in a checker that runs on every lesson. The lessons you have not reviewed get a sweep against each one before their videos go; on the content you have approved the same checks run as warnings and the hits are kept as a backlog, not re-authored now.

The short version

Each pattern as a card

1. New jargon introduced without the coining frame

Your words: "you've just dropped the term renewable in there without introducing it — how did this happen? … It should be a rule that you introduce new jargon like this — why wasn't that followed here?" (Topic 29 L01); "why is new jargon not being introduced correctly?" (L02).

What went wrong. Topic 29's first lesson said "renewable" in its rule line and items before any sentence had said what it means; the second lesson's example sequence never reached its point ("coal takes millions of years to form, but people use it much faster"), so "fossil fuels" and "non-renewable" arrived as labels on nothing.

Root cause. The coining rule ("when X, we call it Y, because Z") lives in the write-lesson and sentence-atomise skills as prose for the author. Each topic declares its terms and the lesson that coins each (a list the author fills in), but nothing compared that list with the text: no check asked whether the coining lesson actually contains a "we call it" line, whether any earlier lesson or item uses the word, or whether later lessons say it again. The vault had a sweep for forward use, but it runs only where a course wires its term table in, and Grade 3–5 had not.

Fix forward. A term sweep now runs on every topic (the Grade 3 session wrote it today; it becomes a failing check in the course checker for Grade 4): a term used before its coining lesson, a coining lesson with no coining line, a coined term that later lessons or the summary never say, and an "is called X" line for a term the topic never declared. Required terms (the register's) are coined in the frame; a convenience term may be introduced in passing and is never tested or recapped (your Topic 12 ruling). A topic's batch does not go until it shows none of the first two.

Status. Topic 29 is being rewritten to your wording now. Grade 4 sweep: 7 lessons missing a coining line, 23 terms with no coining lesson declared, 26 coined terms not reused, 7 not in their summary, 57 undeclared coinings (some noise). These are worked through topic by topic before each batch; the eleven first topics first.

2. Misconceptions invented for the lesson

Your words: "you've invented a non-existent misconception. It's concerning cos I've seen this happen a few times — can you look for the root cause?" (Topic 25 L04); "I don't think we need to introduce this misconception at all" (Topic 21 L03).

What went wrong. A bar chart lesson told the child that the tallest bar is not the snowiest day, a confusion no child had; a rock-layers lesson stated and denied that a shell sank through solid rock.

Root cause (verified by the coordinator sweep, 7 Oct; corrected that evening — James: "this is never a rule I made and should be deleted"). The check-design skill's type (c) listed "the misconception: a named student states the register's misconception" as the third of four items in every concept lesson's check set — a slot, written by an earlier course's session, that James never ruled. Authors filled the slot whether or not a belief existed. The vault's own misconception rule (29 Sep, James's words) already said the opposite: no source, no claim item; a bank with no claim items is a fine bank. The item skill also says "the misconception trap from the atom's register entry"; but the assembled Grade 3–5 register has no misconception field on any atom; the course registers name misconceptions at SUB-DOMAIN level (subdomains[].misconceptions, e.g. LIF4.json), not per atom, so a planner reading the atom saw none (correction, 7 Oct afternoon: the Topic 14/17/35/37/38 pass found the sub-domain lists and now reads them; the fix below adds the sub-domain list to the sources an author may draw on). So every type (c) lesson plan demanded a misconception the register did not supply, and authors manufactured one; both of your cases were planned that way in their check designs. The Grade 3–5 register names a real, catalogued misconception for only some atoms (the physics catalogue covers forces, energy, light, sound); for the rest, the author filled the slot by inventing a plausible-sounding error and then often stated it in the lesson too. The rule "confront the age's misconceptions by name" was built for physics and says a lesson never states a belief the child does not hold, but the check-design step did not carry that boundary.

Fix forward. A misconception line or item is written only for a belief the register (its sub-domain misconception list or the atom's cross note), the misconception catalogue, a published study the item cites by name, or a reviewer has named; where none is named, the slot is a transfer item. The lesson text never states a misconception; a named one is confronted in the lesson's cases or tested as a distractor. The check-design skill is being amended today with that boundary.

Status. Topic 25 L04's line is deleted; Topic 21 L03's sentence is deleted and the belief moves to the quiz. The sweep of unreviewed lessons for stated misconceptions is part of the term sweep's pass (a "wrong-belief sentence" the lesson asserts and then denies).

3. A verdict with no question, far from the claim

Your words: "woah — 'No, Lincoln is not right.' comes wayyy after what he said. Also, there is no question — I think you're implementing some rules incorrectly here… if you say 'no' it has to be in response to a question." (Topic 25 L07).

Root cause (verified by the coordinator sweep, 7 Oct). Your Topic 7 note on Colby ("No, Colby is not right — answer at once") was landed in the vault this morning as "the verdict is the next sentence, zero delay" and applied within hours to a case with no question asked; in the Colby case the question was in the sentence before. A rule landed and applied the same morning should carry its originating example as the boundary case before anyone acts on it.

Fix forward. A verdict either answers a question asked up front ("So, is Lincoln right?") straight after the evidence, or, better, recaps the claim and contradicts it directly and firmly: "So, when Lincoln said Austin has a cold winter climate, he was not right. Austin does not have a cold winter climate; it was one cold day." The vault rule is being corrected by the Grade 3 session today; the summary and verdict beats of unreviewed topics are swept for a bare "No" with no question before it.

4. "So" opening a told fact

Your words: "you've said 'so' when it's not justified — you can't deduce that amplitude is the name of this feature based on this lesson — that's just a fact you're telling them." (Topic 8 L02).

Root cause. The rule-slide move ("So, here's what we've seen…") made "so" the habitual opener for any closing line, including a naming. Nothing distinguished a deduction from a told fact.

Fix forward. Vault rule today: "So" opens only a deduction; a name or a definition is a told fact ("This height is called the amplitude. So the rock's wave has a bigger amplitude than the pebble's."). Scope: a first naming; the recap opener before a rule line stands until you say otherwise.

5. Lettered cases for young children

Your words: "try and avoid naming things A, B, C, D for the young kids — it's a bit abstract — just stick with the pebble's wave and the rock's wave." (Topic 8 L02, L03).

Root cause. No rule said so; letters are the habit of HS physics figures and item drawings.

Fix forward. Vault rule today: in grades 3–5 the cases are named people and things, never wave A and wave B; item drawings name waves by position (top, middle, bottom) and arrows by colour. Applied to Topics 7 and 8; the sweep of other topics' item drawings for lettered labels is on the list.

6. Vague phrasing

Your words: "'the mark that using each one leaves' — too vague → how each energy resource impacts our environment" (Topic 29 intro); "what do you mean leave no mark?" (L05); "what does 'go with the ripple' mean? … be precise"; "it sounds confusing to say the water 'stays put' because it does move up and down" (Topic 8 L01); "'played a part too' — in what?" (Topic 36 L04).

Root cause. The say-the-specific-thing rule exists, and the sentence pass that catches a vague phrase was run on articles; scripts were written separately and not put through the same pass, and the video-script skill's own checklist did not include it.

Fix forward. Every script beat goes through the sentence pass as the article does (the write-video-script skill now says so), and the brief's checker reads the beats for the banned vague frames ("leaves a mark", "stays put", "plays a part" without an object). A phrase a reader could ask "what does that mean?" of is rewritten as the fact.

7. Heat as an object that travels

Your words: "I prefer 'the hotter thing heats the colder thing' rather than treating heat like an object that moves from one object to another … if the temp is the same, neither object heats the other." (Topic 6 summary).

Root cause. The energy strand's carrier rule ("light, sound and heat carry energy from one thing to another") is a true sentence of the energy atom; the heat topic borrowed its grammar and so said "heat moves from X to Y" everywhere.

Fix forward. The heat topic says "X heats Y" in every lesson, item and figure (done, cold-read 112 items). The carrier sentence stays where the energy atom says it, until you rule on the tension (your card A45 on the Grade 3 pages).

8. Contrasts without parallel structure

Your words: "Notice the parallel sentence structure when I'm contrasting cold in vs heat out — you should be using structural parallelism in cases like this — that should be in my writing skill pack already." (Topic 6 L01).

Root cause. It was in the writing pack (sentence-atomise, "Line It Up") but write-lesson had no mechanical check for a contrast pair, and the author wrote the two halves in different frames.

Fix forward. A contrast takes one sentence frame; the sentence-atomise skill gained its mechanical check today (check_parallel) and write-lesson calls it on every contrasting pair.

9. Convenience terms recapped and tested

Your words: "energy changer is not a term that they need to know or be assessed on … it's okay for you to introduce the term … but just don't test the kids on it … The summary should focus on things that are testable, and the students need to know." (Topic 12).

Root cause. The authoring brief treated every coined term as a lexicon entry to recall and recap; the summary skill recapped the lexicon.

Fix forward. Two kinds of term: required (the register's, the standards') — coined in the frame, tested, recapped; convenience — introduced in passing, never tested, never recapped. The write-summary skill now recaps only testable, required knowledge. Topic 12 reworked; the term sweep's undeclared-coining list is where other convenience terms will show.

10. A "first person" ban that you never asked for (my own)

Where it came from. You have two rules here: a rule sentence speaks to the student as "you" (your 5 Oct note on "friction is helpful when you want the grip"), and a coining is framed "when X, we call it Y, because Z". On 5 Oct a render voiced the friction rule as "when I want the grip"; I fixed that by edit and then wrote into every video brief "never I / we for you". The pipeline read that as a general first-person rule for grades 3–5 and began holding ordinary classroom lines: "Let's check with two thermometers", "We build circuits with batteries only", and this afternoon your own coining frame, "we call it a conductor". Nothing is wrong with first person; the brief now says only what you said: a rule sentence speaks to the student as "you", never "I". The pipeline has been told to stop holding lines for "we" or "let's".

The pattern. A rule written from one failure case, without its scope statement, applied everywhere. The standing rule is that every new rule carries its origin and what it does not cover before adoption; I skipped that in the moment.

Scope of every fix on this page

The rules apply everywhere; approved content is not re-authored now. Your words this afternoon: "The rule still applies. I'm just saying that we don't need to go back and re-author stuff right now if it's already been approved." So each fix lands in the skills and checkers, binding on every course from now (the Grade 5 build inherits them), and is applied to the Grade 4 topics you have not yet reviewed before their videos go. On content you have approved — Grade 3, and Grade 4 Topics 1, 2, 3 and 11 — the checkers report the same hits as warnings and the list is kept as a backlog for a pass you may call later; nothing there is re-authored now.

Example sequences: the before and after you asked for

Your rule: the near miss differs from its neighbour in one feature, the deciding one; nothing irrelevant changes. The 6 October audit found 114 sequences in Grade 4; 54 passed, all 114 do now. The root cause was timing and tooling: the four juxtaposition rules landed in the skill on 6 October, after Grade 4 was written on 2–3 October, and there was no mechanical check. Now every sequence lesson carries a block of its cases and the vault's sweep fails any step that changes more than one feature or changes the object at a boundary.

Three of the sixteen on the sample page:

The full sixteen, old run in red beside new in green with Apply or Keep under each: the sample page. Topics 1, 2, 3 and 11 are left as you signed them off.

What happens next

  1. The Grade 4 sweeps run now: terms (coining, reuse, undeclared), stated misconceptions, bare verdicts, lettered labels, vague frames. Each hit is fixed in the lesson before that topic's batch goes; the eleven first topics first.
  2. Your rulings on Topics 21, 25, 29 and 36 are being applied as this page goes up, including the two lesson deletions.
  3. Each fix forward is a skill step and, where a machine can see it, a checker row in the vault, so it runs on Grade 3 and every later course.

11. A quoted student or teacher at the top of most lessons

Your words (7 Oct, evening): "why do you keep creating these contrived questions asked by a student or teacher at the start of every lesson?" … "some of the student statements are completely pointless" … "it is also fine to use the student question / misconception setup some of the time — but the problem is at the moment you're overusing it, and doing it in scenarios that don't warrant it."

What it looks like. "Here's Miss Hayes's class at the window. Rain falls from a gray cloud onto the playground. Landon says snow fell from a cloud like that one last winter. What do rain and snow have in common?" — Landon adds nothing. "Levi says all of this rain came out of our lake. Is he right? No, he's not right." — a claim no child made, answered in the next breath. Most Grade 4 lessons open this way; Topic 20 does it seven times in a row.

Root cause. Not a new fault: three existing things colliding, none of them a rule you made in this shape. (1) The check-design skill's per-lesson "misconception" slot (pattern 2) trained authors to put a named child's claim at the front of every lesson. (2) The sentence-atomise step "name the actor" was read as "name a person" — your rule is that a sentence has a subject, a noun, doing a verb; "Rain falls from the cloud" already has one. (3) The headache guideline (open on a scene, end on a question about the happening, do not answer it straight away) is good and stays; the cheapest way to satisfy all three at once was a child who asks the question.

Fix forward — scoping, not a new rule. In the shared rulebook (vault e62dce7): the misconception item exists only where the register or catalogue names a belief and the lesson addresses it, never as a slot; "the subject is a noun, not a person" written into the sentence step with your words; the headache note gains "who asks the question": the narrator, over the scene, with "let's look at this example" as the natural next line, and a student speaks only when the claim is the lesson's point and the belief is real. And vault-sync gains a step 0: a rule is standing only on your words or your confirmed ruling; anything else is filed as proposed and not applied. Sample for your verdict: G4_OPENERS_SAMPLE.html (five improved, three kept). The topic-wide sweep of the unreviewed topics follows your taps.