# Temporal-stack audit — 2018 `I am the method` video

## Purpose

This is a finite low-light compositing control. It asks whether a fixed
visual overlay becomes clearer after temporal averaging or median stacking
of the preserved YouTube transcode. It is not an LSB recovery, a general
steganography search, or evidence about an unavailable original upload.

## Input

- Video: `solution/video/i-am-the-method-kQx2G3x7_WQ/i-am-the-method-kQx2G3x7_WQ.mp4`
- SHA-256: `0383e479b368c94983e0ce3e25d65e85ccf632f944be2f9258c38c4fc7fa52e4`
- Decoded dimensions: `1920 × 1080` at `29.970030` fps.
- Mean stack: `149` frames at one-second intervals.
- Median stack: `15` frames at ten-second intervals.

## Fixed derivatives

The published figure contains the temporal mean, temporal median, gamma
0.25 and CLAHE variants of both, plus a high-pass mean and temporal-luminance
standard deviation map. The same transforms are described in the JSON.

## Measurements

- Median per-frame grayscale mean absolute deviation from the temporal mean: `0.1365` / 255.
- 95th-percentile per-frame grayscale mean absolute deviation: `0.1404` / 255.
- Median pixel luminance standard deviation across the one-second stack: `0.0000` / 255.
- 95th-percentile pixel luminance standard deviation: `0.5769` / 255.

## Automatic QR control

| Derivative | Decoded text | Geometry response | Usable four-corner geometry |
|---|---|---|---|
| `temporal_mean` | `—` | `False` | `False` |
| `temporal_median` | `—` | `False` | `False` |
| `mean_gamma_025` | `—` | `False` | `False` |
| `median_gamma_025` | `—` | `True` | `False` |
| `mean_clahe` | `—` | `False` | `False` |
| `median_clahe` | `—` | `False` | `False` |
| `mean_highpass` | `—` | `False` | `False` |
| `luminance_temporal_std` | `—` | `False` | `False` |

No derivative produced decoded text or a usable four-corner QR geometry. A
geometry-only response with fewer than four unique corners is preserved in
the JSON as a detector false-positive rather than treated as a QR finding. The composites remain
available for inspection rather than being represented as a general negative
proof about every possible visual encoding.

## Conclusion

The predeclared temporal mean, median, gamma, local-contrast, high-pass, and variance derivatives produce no decoded text or usable four-corner QR geometry. One gamma variant causes a degenerate detector response with fewer than four unique corners; it is recorded as a false-positive geometry, not a QR result. The derivatives are published for inspection, but do not recover an address, command, or next-message target from the preserved transcode.