24 October 2026 · for James, Dan and David. James ruled on all eight proposal questions on 23 October. The build is go, pending the four questions below.
Takeaways
The ruled course designs at 41.1 hours (about 44 with everyday slack) against v1's 112.1 hours. Same scope; fewer, bigger lessons and lighter checking.
The hours rose from the proposal's 37.3 because of three rulings: topic intros kept (+1.35 h), End of Topic Tests priced at the platform's 11-question adaptive minimum (+0.90 h), and the NY-exam content added (+1.50 h).
670 v1 lessons become 339, each a 90-second video or a short article, then a quiz of 2 to 5 questions. Dan's 124 lessons and 108 of David's 112 videos carry over.
What this page needs: four votes — one from Dan, one from Dan and James together, two from David. Everything else is settled and listed below for the record.
Where the hours go
v1: 112.1 hv2 as ruled: 41.1 h designed
Every row is computed and sum-checked by data/duration_v2.py: the six bars total 41.07 h, asserted in the script.
Quizzes are priced at the default of 3 questions. The 2-to-5 flex is assumed to average out; if the atom audit moves the average, this row is re-priced.
Lessons per unit — the merge
now (670 lessons)after merges (339)
Full atom table: MERGE_REPORT.md — every v2 lesson with its source atoms and a one-line reason. The ~13 NY lessons and the density re-cut (two lessons, ruling 8) are added on top of the 339 when the v2 register is written.
Vote
Who is voting?
needs Dan
1. Dan — happy to proceed on the ruled base, with your 124 lessons re-cut to the 90-second shape?
James ruled the base: the v1 course's skeleton (670 lessons merged to 339) plus your additions — your ~13 NY-depth lessons, your sequencing changes, your register corrections.
Your 124 lessons keep their teaching. The re-cut to "a 90-second video or an article, then a short quiz" is mechanical: your per-atom text blocks already sit at the right grain.
This base keeps 108 of David's 112 delivered videos and all 9,475 reusable items in their current places.
Recommendation: yes — the re-cut changes your lessons' shape, never their substance.
See it: MERGE_REPORT.md · dan_base/COMPARISON.md · dan_base/LESSON_SAMPLE.md
needs Dan + James
2. Where does the build live — Dan's hs-chemistry-v2 repo, or James's course factory?
Dan's repo as the build home: the skill pack and these rulings wired into it; James's side ships the merge plan, the 9,930-item verified bank and the validators to it.
Centralised: the build runs in James's course factory and Dan reviews from outside.
Recommendation: Dan's repo as the build home — he owns the course, and James's side supplies the merge plan, the item bank and the validators.
needs David
3. David — will you share your image library, so v2 reuses images instead of regenerating them?
Your pipeline has already produced many of the images v2 needs. James ruled: contact you and reuse what exists, never regenerate it.
The ask: push the image library itself — or the CDN address scheme plus a manifest keyed by lesson and figure — into the shared tracker repo, so the build can look an image up before drawing one.
Recommendation: yes — reuse costs a manifest; regeneration costs a render per figure and a second look to keep consistent.
needs David
4. David — can you confirm the PP100's ~11-question minimum and the numeric grading contract?
James says the End of Topic Tests (PP100s) are adaptive, with a minimum of about 11 questions asked. The duration model now prices every topic test at that minimum.
The numeric grading contract — what tolerance applies, how typed answers are normalised — is the same confirmation already gating v1's test registration.
Recommendation: one platform probe answers both, for v1 and v2 at once.
James has ruled these eight — tap Disagree if you would rule otherwise
Each ruling is settled and the build proceeds on it. A disagreement posts to the course inbox with your name and your reason.
Whose base carries v2?
James ruledThe v1 skeleton plus Dan's additions. Dan's NY-depth atoms join; his 16-topic map does not replace the structure.
What shape is a lesson?
James ruledA video targeting 90 seconds (hard cap 2 minutes) or an article, then the quiz. Images: reuse David's existing library, never regenerate what exists.
How many questions after a lesson?
James ruledDefault 3, flexing 2 to 5 by the atom: a simple fact atom may take 2; a late-topic capstone atom up to 5.
Which standards does the course cover?
James ruledThe union: TX §112.43 + NY Physical Science, and the built Florida content stays.
Is the matter on-ramp taught or retrieved from middle school?
James ruledA slim taught version (~10 lessons, ~1 h), with a test-out so strong students skip it.
Test-out, and what happens to the items the trim frees?
James ruledTest-out yes, from the PP100 bank. The freed items are a large bank, every item tagged with its atom and a difficulty: the atom is the main difficulty factor (capstone and routine atoms harder, early recall atoms easier); free numeric entry outranks MCQ.
Do David's topic intro videos stay in the student path?
James ruledYes — kept in the counted path (+1.35 h). The summaries stay too.
Any trimming, and how fine do simple-equation lessons cut?
James ruledNo trimming — the design should land inside the budget as it stands. PP100 length is not ours to set: the test is adaptive, minimum about 11 questions. Simple equations like density take one lesson for the equation and one lesson for both rearrangements, never one lesson per rearrangement.
What happens next
The v2 register is written from the merge map plus Dan's additions.
The 88 video briefs and the article re-cuts are specced unit by unit; the PP100 banks and unit tests are assembled from the 9,475 verified items, each tagged with its atom and a difficulty.
Dan's harvesting map (his atoms → v2 lessons) is drawn the way the merge map was.
Nothing above starts on a blocked question: Q1 gates the re-cut, Q2 gates where the work lands, Q3 gates figures, Q4 gates test assembly.
The evidence files, for anyone who wants the detail
MERGE_REPORT.md — the full atom table, 670 → 339.
dan_base/COMPARISON.md and dan_base/LESSON_SAMPLE.md — Dan's build, assessed.
david_videos/REUSE_TABLE.md — per-video reuse verdicts, 108 of 112 fit.
data/ITEM_INVENTORY.md and data/duration_v2.py — the item stock and the hours model.