Author checks completed · frozen for independent review. Source work began 30 September 2026; final artifact review is on 1 October 2026. No actual learner/experiment/classroom clearance supplied. Source request dates and original model dates are retained, rather than rewritten as new fetches.
What this pack contains
Ten flexible 25-minute scripts, 30 filled access choices, 20 original worked domains, ten free standalone home offers and ten extras, original named Shadow Window source, two new open held C/D briefs with separate descriptive key/next moves, exact seven Year 5 rows plus full level/standard, source/rights ledger and eight original A4 SVG/searchable PDF pairs with exact linear/optional tactile/no-print/no-sound alternatives. No helper/audio is needed. Counts inventory authored work; they are not curriculum coverage or observed learner outcomes.
The ordinary ideal diagrams use source0/bar4/screen8 then bar2/screen8; the approved fresh C source2/6 and D3/9→2/6 are reserved. All units are arbitrary, ideal straight rays. Actual extended lamps/ambient light/edges can differ; no numeric projection quota or actual-data replacement.
Portable local checks
From this pack's directory, with Python3/Pillow/pypdf, DejaVu Sans, librsvg and Poppler available:
PYTHONDONTWRITEBYTECODE=1 python3 verify_pack.py
PYTHONDONTWRITEBYTECODE=1 python3 verify_negative_probes.py
PYTHONDONTWRITEBYTECODE=1 python3 print/generate_print.py --check-reproduction
Default verification is read-only, including temporary genuine rebuilds. The explicit author commands below mutate only this leaf's outputs/manifest and are not needed to teach:
PYTHONDONTWRITEBYTECODE=1 python3 print/generate_print.py
PYTHONDONTWRITEBYTECODE=1 python3 verify_pack.py --write-manifest
Use --output with a temporary directory for a new print build. Deterministic PDF metadata retains the original model design date as a fixed reproduction value; it is not a claim about actual generated-file creation time or a remote source's date. The temporary byte comparison proves reproduction in the recorded runtime; cross-platform font/render identity is not certified.
Completed checks
- Default read-only verifier passed: exact seven raw/canonical Year 5 descriptions plus full level/achievement standard, truthful current-source flags, original source/type/purpose and ideal arithmetic, ten contiguous flexible clocks/30 concrete routes/20 distinct contexts/home10+extra10, first/later/open choices and 56 local links.
- Independent diagram checks passed: actual blocker/screen endpoints and straight edge-ray projections; ordinary span4→8; fresh C/D have no ray/projected-shadow answer and retain fixed unfilled receiving lines; D before/proposed share the same arbitrary scale. SVG/PDF full-page A4 and searchable visible-text fidelity/bounds passed. These mathematical checks are model consistency, not actual lamp trials.
- 17 genuine temporary rebuilds (eight SVG, eight PDF, exact-alternatives document) were byte-identical in Python 3.12.3, Pillow 10.2.0, pypdf 5.1.0, librsvg 2.58.0 and Poppler 26.05.0 with installed DejaVu Sans. Stored files were not overwritten by default checks.
- 48 meaningful temporary faults rejected at their intended source/model/teaching/geometry/rights/inventory gates. Non-cache mutations had manifest and fresh-file hashes renewed before semantic checks; the incidental-cache fault deliberately cannot enter the inventory. Probes included fabricated events/HTTP/layout, wrong raw row/column, model physical units/actual-data substitution, C/D changed facts/key, ordinary held-source leak, broken clocks/missing routes/contexts/home, speed/later-sitting failures, ray/blocker/receiver/scale defects, filled fresh projection, non-A4 page and trimmed upstream font terms.
- All eight stored full PDF pages were rendered by Poppler at 100 dpi and actually viewed individually, including both fresh briefs, blank actual-record columns and source card. No clipping/overlap found; receiving-line scale and credit/purpose lines were inspected. Full-page raster inspection is an author screen/layout check, not a physical print, tagged-PDF validation, learner trial or human accessibility clearance.
- Full author desk read-through of all scripts, routes, contexts, home offers, learner cards, sources, key and alternatives completed. Corrected a receiving-screen versus emitting-display ambiguity in Day22/route22A, removed an unrelated movement phrase from Day30, and simplified a stray cross-subject home evidence phrase. The child invitation remains a useful action; detailed evidence scopes stay in educator notes.
- Owned Python formatting/lint checks passed without caches. The final manifest binds all deliverable files; default verification and negative probes preserved stored SHA/size/mtime. An external final receipt records the exact inventory/hash and frozen Year 4 preservation comparison. Independent/root review and release remain separate.
Four author desk perspectives
| Role | Practical critique and remaining limits |
|---|---|
| Teacher | Start here gives a reusable ordinary setup and one shadow action before theory. Locally decide apparatus/roles/edge definition, choose one script/route and keep a short useful note; no 30-route log or compulsory photos. Main risk: treating ideal heights as real readings. Source card, blank record and lesson prompts separate them. Actual class pacing/access/apparatus remains untried. |
| Learner | Own purpose, inquiry alone, early optional actual shadow, varied questions/making and keep/change/stop prevent ten paper audits. No numeric projection, grip, posture, speech, audience or complete-check speed target. An actual source/communication need can change the route; no learner usability trial supplied. |
| Parent | Home entries begin fresh and need no kit, earlier drawing, printer, lamp trial or record. Choose one short imaginative/conversational offer and skip extras. Their reasoning is source/model participation, not automatically actual inquiry. No family feedback supplied. |
| Funder/reviewer | Exact provenance, narrow U03/inquiry conditions and actual reproduction/fault/render receipts make the pack inspectable. Full Science year, human-endeavour breadth, all-state adoption, classroom pilot and benefit are not claimed. Independent root review/publication remains separate. |
Remaining human/local scope
Review actual child access/communication, apparatus/light/space safety, local adopted authority/timetable and actual evidence interpretation before use. Actual trials/operators/observations/measurements, other-investigator sources, maker outcomes and willing reader responses are supplied only when they really occur. Complete planned/director/source routes remain genuine with actual physical components open. First versus later responses and unfinished/not-yet-observed/later sitting/no-speed distinctions apply to both fresh starts. Scripted/computer-tested is not classroom piloted.