# Contrapunctus I final-cadence audit — `I am the method`

## Scope

The timing audit selects the audio/score alignment from the musical body and its two internal rests. This follow-up does not refit that alignment or search for a message. It describes the sound that remains after the mapped final MIDI note.

## Result

The documented MIDI ends at score 156.000 s, which maps to video 136.842 s under the independently stored timing alignment. Literal PCM silence begins at 143.089 s, leaving a 6.247-second audible tail.

Immediately before score end, the reference holds the MIDI notes D, A, F#, D. Their pitch classes form `D major (D–F#–A)`. In the final stationary portion beginning at 139.089 s, all of the eight strongest narrowband spectral peaks are in those D/F#/A pitch classes; its median frame-to-profile cosine is `0.994` (5th percentile `0.989`).

This is consistent with a long synthesized/rendered final cadence (including its D-major/Picardy-third sonority), not an additional sequence of changing notes. Rich timbre creates overtone frequencies such as C# from A; the peak table therefore uses narrowband local maxima instead of treating every broad chroma bin as a separate written note. It does not rule out an unspecified keyed transform, but the tail is no longer evidence for a standalone audio message.

## Eight strongest narrowband peaks in the final stationary tail

| Frequency (Hz) | Nearest equal-temperament note | Pitch class | Relative PSD power |
|---:|---|---|---:|
| 292.969 | D4 | D | 1.000 |
| 183.105 | F#3 | F# | 0.972 |
| 73.242 | D2 | D | 0.948 |
| 109.863 | A2 | A | 0.899 |
| 366.211 | F#4 | F# | 0.279 |
| 146.484 | D3 | D | 0.271 |
| 585.938 | D5 | D | 0.180 |
| 439.453 | A4 | A | 0.136 |

## Reproducibility

- Audio SHA-256: `bbb2d00633a0c9278b2f0b4de12c9eeffaf8d1aadd4ccc5425d3bedcb0f572fc`
- MIDI SHA-256: `9d051a6938b09571846983281c2c7e14a447659e8bf0c74724ce6e07c4070528`
- Timing audit input SHA-256: `1f213bd4d3435a9df765e79a29bbdbaba978bc46bb5e8fa5e90b8b45482eac38`
- Alignment inherited without refitting: score = (video - 0.000000) × 1.140000`.
- Plot: `final-cadence-audit.png`

## Limit

Spectral peaks are an acoustic summary, not source separation and not a proof that no hidden transform can exist. The conclusion is deliberately limited to the obvious hypothesis that the post-score sound is a separate extra melody or unparsed note stream.