Teacher desk run: Day 53 fits 25 minutes (3 safety/source, 4 prior prediction, 5 adult action, 7 exact child record, 4 limit check, 2 exit). With a genuinely approved local 9 cm marked tabletop passage and locally measured large smooth-edged object, adult can make one slow flat try without force and record the observed category. If dimensions/setup/approval are missing, teacher writes REAL OBSERVATION NOT AVAILABLE, uses fictional Model P and teaches source reasoning only. Day 54 refuses to compare different surfaces as if orientation alone changed. Day 57 refuses a “shape alone” claim when width also differs. A child is never physically tested in a passage. Day 59 offers a private teacher listener and records real feedback only if it occurs.
Child desk run: From P I can say a drawn 6 cm span lies between 9 cm lines and the turned 12 cm span overlaps. I cannot say a classroom card passed or stopped. I may ask a question, predict for a real safe attempt before it happens, then help record an actual result if one occurs. I can point, speak, sign, use AAC, direct an adult, follow an attributed ADULT REPORT or use secured broad tactile cards. The Science claim stays the same; source label stays visible.
Assessor desk run: Held Day 55 uses a new oval/7 cm paper gap; Day 60 uses new 9/13 cm rectangles/11 cm gap. Neither appears in P/Q/T or worked swaps. First independent wording/prompt/access is captured before feedback. Model-only answers can show source reasoning but cannot satisfy actual movement/safe-investigation evidence. There is no fixed actual outcome in the key. A prior prediction is credited only if made before actual movement. One check cannot establish a term/year standard or growth.
Funder/public desk run: Original model and vector rights, seven exact pinned official rows, Queensland entry, optional free family participation, safe boundaries and reproducible hashes are visible. No classroom pilot, child outcome, universal access certification, translation or ACARA/QCAA endorsement is claimed. Local teacher, child/accessibility and safety review remain needed before such claims.
Mechanical/visual receipt: Verifier checks ten six-step 25-minute scripts, 30 routes, 20 worked swaps, ten optional family prompts, two unseen checks/keys, seven official rows and source hash, local links, six A4 SVG/PDF pairs and file hashes. PDFs are rendered to a contact sheet and visually inspected after generation; repeat after any artwork change. This is a desk review, not a classroom field trial.