Goal: support a narrow claim with valid entries, keep a slipped grip visible, and compare a genuine prior prediction only when it exists. Proposed codes: AC9SFU03, AC9SFI03, AC9SFI04 (conditional). Prepare: actual dated record if any or L, property claim, Card 107.
- Source · 3 min. Mark ACTUAL/date or L MODEL on every displayed copy; never substitute tidy L for a real mixed record.
- Read raw · 4 min. Read P/C/P/C valid entries and any interrupted entry from chosen source in order.
- Model · 5 min. Of L only: “P tips are BELOW twice; C tips are AT and ABOVE in this pretend set; slipped C is not a valid tip.”
- Child try · 7 min. Card 107 routes cite exact two or more raw entries, state a source-bounded material observation and one unknown; compare with an actual earlier prediction only if documented, otherwise NO PRIOR PREDICTION.
- Check · 4 min. Ask whether L proves card backing always works in every room or for everyone. Child limits to these model entries and separates route use.
- Exit · 2 min. Save exact child claim, source, support and prompt.