30 September 2026 · AI author desk review, not a human educator/learner pilot or physical material trial. Ten scripts, all routes, worked domains, home bridges and extensions, complete original sources, both held checks/keys and crosswalk reviewed against the approved concept. Timing/preparation estimates are unmeasured allocations. Actual locally approved pad, pressing result, classroom timing, sensory access, willingness and learning improvement remain untested.
Teacher pass
Desk-followed the Start here → preparation → one question/prior idea → before/optional gentle press/release → short shared phrase → chosen next question sequence. Same-actor/no simultaneous pressing during tactile access explicit. Director/report/model complete; no mandatory touch, voice, new item, press, fixed wait/repeat or outcome. A hand hiding the centre leads to another feature, report/model or pause rather than invented vision. One shared source/response with relevant individual idea/support replaces duplicate full logs. Known material or UNKNOWN; folded form not new material, composition breadth open.
Useful limitation: the supplied pad drawings are qualitative proposals/fictional notes; a teacher must choose the actual familiar item and observe its response. No physical preflight or classroom schedule validation by the agent. Local approval remains necessary before any press. Complete no-print description/source route available when an item is absent.
Parent/carer pass
Ten home invitations can be conversation/drawing/direction only. No purchase, home experiment, diagnosis assumption, family story, image/audio/upload or proof. Check-day bridges held until school first responses; key not embedded in family route. Parent may decline. The real material/cleanliness/access review belongs to the local adult; printed approval language is not item approval.
Student pass
Plain opening questions and large before/press/release scenes inspected. Child owns part, question, prior idea, action or no action, observation mode, stop and optional next question. One careful look, unclear/unchanged result or a private question valid. Child-director artefact retains authorship with adult contributions named. No promised bounce/elasticity/comfort or prescribed “correct” feeling. Adult-only source limits stay outside the short child invitations. Actual child enjoyment/access remains untested.
Funder/reviewer pass
Seven exact selected Foundation Science descriptions pinned to current workbook E/J and canonical rows; all nine checked, U01/U02 not claimed. SFU03 partial property/material opportunity, composition not inferred from fold/softness or a single material. SFH01 conditional actual person/example. Held AL/AM source reasoning does not establish physical SFI02 evidence or full-standard attainment. Source creators/type/purpose honest: actual saved computer artwork/text, proposed or fictional described event, no human/community testimony. Original CC BY4.0 educational resources, code/font/curriculum/community rights separate.
Fresh complete QCAA indexed primary versions read; direct entry/PDF requests HTTP403. Current PDF bytes/layout unverified; no downloaded/rendered QCAA PDF claim. Current full Queensland/adopted local programme review remains due. No evidence of learner benefit, classroom efficacy, official endorsement or physical/material result.
Actual print and asset review
All eight actual generated searchable one-page PDFs rendered to pixels and visually inspected: setup, before-press-release, parts-material, notebook-35, shared-try, your-question, fresh-al, fresh-am. Broad diagrams first; words/source relations and exact alternatives checked. Blank local-result spaces remain blank. PDF text is searchable but not tagged; complete linear text and optional broad tactile model alternatives provided.
Author corrections found during inspection: C’s release outline now matches its before width as well as height; a recognisable side-view hand-and-forearm symbol replaces a block-like palm that could resemble a weight. The child-facing material aid now uses plain questions; detailed SFU03/composition limits stay in teacher documents. These are drawn symbols, not a real hand/force measurement. AL/NOTE intermediate relation and AM A/B repeated relation are qualitative. Re-rendered pages inspected after the changes.
No digital helper or interactive product supplied; complete paper/no-print sequence is simpler for one optional action. Browser interaction, keyboard/reset/storage states therefore not applicable. No network/storage/upload code exists in the learner resources. Static SVG/PDF display does not supply a physical result.
Reproduction and read-only audit
PASS: actual workbook XML E/J plus all nine canonical Foundation Science rows, seven selected exact descriptions/current versions and honest fetch receipts; ten optional 25-minute scripts; thirty switchable day routes; twenty distinct worked domains; ten home bridges and ten extensions; held AL/AM isolation/source criteria/open keys; one-actor/low-burden/access/composition limits.
PASS: all eight searchable one-page A4 PDFs, whole SVG/PDF visible text equality, text bounds, exact alternative wording/order, meaningful AL/NOTE intermediate geometry, A/B repeated AM relation and C matching-before outline, sole press-stage hand symbols, blank real-result spaces. Sixteen SVG/PDF temporary rebuilds byte-identical. Seven meaningful temporary corruptions rejected: wrong selected row, learner answer leak, falsely identical observation time, wrong AL partial-release art, wrong C same-outline art, false child direct-vision attribution and fold falsely named a new material. 64 local links/anchors; 38 content hashes.
Runtime for byte reproduction: Python 3.12.3, pypdf 5.1.0, rsvg-convert 2.58.0, Poppler 24.02.0 with installed DejaVu Sans. A different renderer/font/runtime may produce different PDF bytes; the check then fails visibly. No cross-platform byte identity claimed. Final default audit’s pack-byte before/after check confirms no writes. The default verifier reads the pack and source workbook/import; all rebuilds and deliberately corrupted probes use temporary copies. Manifest refresh is an explicit authoring operation. Public reproduction commands use project-relative paths and stated prerequisites; no private workstation paths embedded.
Remaining human/local gates
A practising educator and learners must review practical timing, engagement, access and the actual safe familiar item; any physical trial and direct/report evidence need real local observation. Current full QCAA documents and local curriculum/assessment programme checked before reporting; no universal material/comfort judgement. External listener/source actually willing and rights-cleared if used; cultural/community authority not invented. These gates are not claimed completed by desk QA.
Independent publication desk review
30 September 2026 · a second AI agent reviewed all ten scripts, thirty routes, twenty worked choices, ten family invitations and ten extensions; supplied source identities, both fresh checks and keys, crosswalk and access/material limits. It reran the default read-only verifier against the actual workbook and canonical import. All eight stored PDFs were independently rendered and inspected as a contact sheet; fresh AL and AM were also inspected as full-page pixel renders. The AL release sits above pressed and below before; AM A/B match while C matches its before outline, with the different observation moments visible. Blank local-result spaces stay blank. No physical pad, child, print-device or classroom trial was performed.
Publication refinement: expand AAC in the learner-route and read-aloud guidance, and replace the unclear verb “AACs” with “uses AAC to communicate their original words.” The final read-only audit after this prose change rechecks the source rows, fresh isolation, eight PDF layouts, sixteen temporary byte reproductions, seven meaningful temporary corruptions, local links and renewed content hashes. These checks establish artifact consistency; engagement and actual local observations still need educator/learner review.