# Recording-to-artifact: R02 reconciliation

September 15, 2026

**Default R02 and Adaptation R02 show a qualified repair of the sampled active → stopping → stopped sequence. Their single biggest remaining gap is unchanged: the dense working history is reduced to a sparse generic assembly list and progress bar.** This is not an overall win or a moving-quality approval.

This report supplies variant and revision names for findings sealed before the naming reconciliation. No additional movie viewing, decoding, native inspection, scoring or evidence credit occurred. The source criteria, primary gaps and confidence limits are unchanged.

## Variant results

| Variant and revision | Recorded result | Single primary gap retained from the sealed review |
| --- | --- | --- |
| **Default R02 — Session workshop** | Active recording remains visible through frame 025; frames 016–025 show “Stopping capture”; frame 026 shows “Recording stopped” and “Capture complete”. | C05: dense mixed prose/tool working history is replaced by a sparse generic assembly list and progress bar during frames 050–095. |
| **Default R01 — Session workshop** | Frames 000–015 read “Recording saved locally”; frame 016 already shows “Capture complete” and reading/checking text, without a visible active/stopping phase. | C02/C03: missing visible active/stopping phase, especially the post-submit interval `[016,026)`. |
| **Adaptation R02 — Fieldwork workshop** | Active field capture remains visible through frame 025; frames 016–025 show “Stopping field capture”; frame 026 shows “Field capture stopped” and “Fieldwork complete”. | C05: dense mixed prose/tool working history is replaced by a sparse generic field-guide assembly list and progress bar during frames 050–095. |
| **Adaptation R01 — Fieldwork workshop** | Frames 000–015 read “Field notes saved locally”; frame 016 already shows “Fieldwork complete” and reading/checking text, without a visible active/stopping phase. | C02/C03: missing visible active/stopping phase, especially the post-submit interval `[016,026)`. |

These remain the primary findings for each revision. The later work-history simplification also existed in R01, but it does not replace R01's previously selected primary state-sequence gap after the revision names become known.

## What the qualified repair establishes

The retained source is already recording at frame 000. Its recorder remains visible after the posted completion message at frame 016 and through frame 025; its first sampled absence is frame 026. R02 restores a visible active/stopping distinction across that same post-submit interval in both variants. R01 lacks that presentation and already displays completion at frame 016.

The interval `[016,026)` is 0.667333333–1.084416667 seconds on the actual retained-file clock, lasting 0.417083333 seconds. This is the supported repair claim: the visible distinction and sampled boundary, not recovered backend recording behavior or smoothness between samples. “Saved locally” alone need not mean that recording has stopped; the finding rests on the missing active/stopping presentation together with the already-complete state at frame 016.

The repair does **not** establish complete C02/C03 fidelity. R02 uses an inline state marker rather than the source's separate, partly clipped recorder pill. The source's composer, pointer and submission details are also not fully reproduced. Those limitations were recorded before reconciliation and remain open.

The coarse assembly, readiness, document-pair and file-detail onsets at frames 050, 096, 111 and 112 were already observed in both revisions. They are retained matches, not new repair credit attributed to R02.

## Biggest remaining gap in both R02 variants

During `[050,096)`, the source shows a dense conversation with dark explanatory paragraphs, pale tool rows, inline code and contextual clipping. A Running row becomes Thinking, then a Ran row and a faint skill-skeleton response enter within that history.

Default R02 instead shows “Assembling the workflow” with “Read session notes” at native frame 050 and adds “Group repeated actions” by native frame 069. Adaptation R02 shows the corresponding “Assembling the field guide” and “Read field notes” treatment. Both remain sparse lists with a progress bar rather than the source's mixed prose/tool account of the work.

The affected interval is 2.085416667–4.004000 seconds, lasting 1.918583333 seconds. The gap is structural and contextual, not a demand to copy source demo names, private command paths, colors or icons. Matching the stage-change timestamp does not replace the missing work detail. No new assessment of motion quality is added here.

## Evidence scope and confidence

All four revision movies had every retained index 000–167 inspected chronologically as **320 × 180 thumbnails**, one-quarter of their 1280 × 720 native dimensions. This was exhaustive thumbnail inspection, not exhaustive native text inspection.

For each revision, the previously inspected native indices were exactly:

`000, 015, 016, 025, 026, 050, 069, 096, 110, 111, 112, 167`.

Complete native images were viewed at 000, 050, 096 and 112. The remaining checks were unscaled crops:

| Frame indices | Native crop rectangle, x0,y0,x1,y1 |
| --- | --- |
| 015 | `(208,228,1068,560)` |
| 016, 025, 026 | `(208,228,1068,508)` |
| 069 | `(208,290,1068,552)` |
| 110 | `(208,290,1068,610)` |
| 111, 167 | `(208,298,1068,640)` |

The right and bottom rectangle edges are exclusive. Full image areas outside those crops were not claimed as natively inspected. All native checks reused already-decoded frame hashes without adding unique-frame credit.

The retained source had previously received chronological 480 × 270 thumbnail inspection for indices 000–167 and bounded native checks at 40 indices. The two repeated source copies were byte-identical to that same source and reused its sealed evidence; they were not additional source watches or new evidence. The companion `report.json` preserves the exact native source indices and viewed rectangles as well as the revision coverage, without private locations.

Every actual file uses PTS `index × 1001` at time base 1/24000, with 1001 ticks per retained frame. The last frame starts at 6.965291667 seconds and the end-exclusive time is 7.007 seconds. The inherited nominal source interval is separate from this measured file clock. No full-master recovery is claimed.

Confidence is unchanged from the sealed review: these are observed sample-based discrepancies and qualified state matches, with the same explicit uncertainties. No numerical confidence score, stronger quality claim or ranking is introduced by naming the revisions. No moving playback, real-time readability, easing, blur-continuity or native Resolve approval is available from these still-based observations. The source-identical material has no new retained-file discrepancy, but byte identity is not a presentation-quality win.

## Disposition

Retain the qualified R02 state-sequence repair in both variants. Keep C05's working-history simplification as the single biggest remaining gap for each R02 variant. Keep R01's missing active/stopping phase as its previously recorded primary gap. No overall winner, additional evidence credit or publication approval follows from this reconciliation.

## Packaging scope

This packet selects the two named current content exports. Earlier comparison variants remain in the sealed research evidence; they are not extra examples or selected-export credits. Original builder snapshots stay dated historical artifacts. Delivery checks, moving perception and native Resolve execution remain separate.