# Batch09: structural fidelity findings

September 15, 2026

## Scope

This report records one primary remaining fidelity gap per treatment, shared by the reviewed candidate variants. The findings come from complete retained export chronology examined as thumbnails and selected native-frame details. Frame order and timing were checked against the actual files' presentation clocks. The reviewer had studied the retained visual references beforehand; this was a source-familiar review.

These are observations about visible export structure. They are not diagnoses of implementation code, rendering systems, or toolkit quality.

## Five primary gaps

### Recording to artifact: the capture-stop event is missing

The candidate sequences begin with an already-saved-session story. Work-state changes and final artifact rows are visible, but an independently active recording control does not visibly transition to a stopped state. The primary repair target is to make stopping capture a distinct event before the work and artifact stages.

### Phone–desktop pair: a persistent foreground anchor is missing

The candidates retain phone-first states and a staged desktop entrance. The missing relationship is a foreground layer that persists while the devices redistribute. Without it, the composition becomes a flat device pair rather than retaining the reference's foreground, device, and background hierarchy. Preserving that relationship does not require copying a presenter or brand.

### Embedded map detail: the map loses meaningful result relationships

The candidates retain the hard change to map detail, typed request, later marker reduction, cleared field, and unfinished second request. Their generic dots and unlabeled street grid do not carry the reference's relationships among landmarks, differentiated markers, and listing information. The primary repair target is meaningful map and result detail, using appropriate original or fictional content.

### Targeted revision: the selected preview does not reflect the edit

The main chart, title, and notes update, but the selected rail thumbnail remains a generic triangle. It never identifies the revised chart. Updating the selected preview to match the edited slide is the primary structural repair target; the local instruction, waiting interval, and discrete main-slide update are already present.

### Physical-screen bridge: hand–keyboard contact is schematic

The candidates include the laptop, changes in framing, keyboard and hands, return to the display, localized screen washout, and attachment-transfer states. The close keyboard view still uses flat hand silhouettes with weak contact and foreshortening over the keys. The primary gap is physically convincing hand–keyboard contact, not an absent bridge sequence. This is a static visual judgment, not a claim that the hand animation was watched and found stiff or that illustration is inherently inferior.

## Structural and static limits

**No genuine moving perception was demonstrated.** Ordered stills and verified presentation clocks do not establish natural motion, easing, camera comfort, typing rhythm, flash comfort, or readability during playback. Those dimensions remain unassessed.

Complete thumbnail chronology is not complete native-detail inspection. The evidence is limited to the retained excerpts; original source-master provenance was not authenticated. The review does not verify real product behavior or issue a production-acceptance result.

No overall winner or moving-quality win is claimed. The findings support the five specific structural repair targets above, without an implementation or toolkit verdict.

## Packaging boundary

The selected packet contains five original code/SVG kits and ten verified content renders. The source excerpts total 1,367 retained frame instances; the two outputs per treatment total 2,734. The critic inspected their complete thumbnail chronology and selected native details, not moving footage or full source masters. The guides preserve their dated builder-stage notes; the later render verification and this report establish the current export/review scope. No earlier pre-render statement is a claim that the later exports are absent. Actual downloaded-kit execution and delivery remain separate checks. Native Resolve execution and quality approvals remain zero.