One added an optional short description to decisions and made the return card and on-screen video text prefer it. The original memory text and correction links remain. The observer directly viewed three plain-language lines in a saved mobile card. That screen used an altered last-visit time for a test, not an actual user return. This stores wording for display. It does not change the text used by search, so it is not evidence of improved retrieval. Limits were also found: an older summary can remain when its source text is revised, and editing the summary alone does not leave a separate revision history. The video draft’s display line used the stored summary, but generated narration changed delegating a long investigation into postponing it. One revised the instruction to prevent that shift and acknowledged in its completion report that rechecking remained unfinished. No regenerated script or rendered video after that change has been confirmed. The earlier trial with decision scenes is separate from this script draft. Whether the new instruction preserves meaning in a rendered video, and what happens at the first scheduled 22:30 generation, remain follow-up observations. A more readable display is not treated as proof of factual summarization or better speech quality.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.