What the parenthesis edit actually changed on Montana and Zomby Woof, what state it is in on Songsterr now, and every number checked against the live API rather than against a local file. Verified 2026-08-30.
The two fix revisions exist on Songsterr, carry exactly the intended change, and alter zero notes. Neither is the revision Songsterr serves to the public. api/meta/35870 still returns 7294223 and api/meta/412162 still returns 7115188, both predating this work.
Promotion is a Songsterr moderator decision. The submission side is finished, the publication side is not, and the earlier audit record was wrong to head its table "Delivered, live and verified."
Census counts every non-rest note on every track, read from Songsterr's own rendered JSON.
| Tab | Revision | State | Notes | Ghost (brackets) | Staccato (dots) | Ties |
|---|---|---|---|---|---|---|
| Montanas35870 | 7294223Ben Dibden1, 2026-06-10 | PUBLISHED | 7966 | 574 | 261 | 276 |
| Montanas35870 | 8787106Brandon chavez, 2026-08-30 21:49:50Z | PENDING | 7966 | 257 | 578 | 276 |
| Zomby Woofs412162 | 7115188meh, 2026-05-31 | PUBLISHED | 7479 | 32 | 138 | 449 |
| Zomby Woofs412162 | 8787022Brandon chavez, 2026-08-30 21:45:34Z | PENDING | 7479 | 0 | 170 | 431 |
Montana: ghost −317, staccato +317, ties unchanged, notes unchanged. Every bracket removed became a dot on the same note.
Zomby Woof: ghost −32, staccato +32, ties −18, notes unchanged. The 18 removed ties are the kick ties named in the fix description.
Montana keeps 257 brackets because the scope was drum notation only. Those sit on vocal, guitar, keys and mallet staves and were left alone on purpose. Zomby Woof reaches zero at file level because all 32 of its brackets were on one drum staff.
Only staves carrying a bracket or a dot are listed. Drum staves in bold.
| # | Staff | Notes | Ghost | Staccato |
|---|---|---|---|---|
| 0 | Vocal 1 | 323 | 24 | 0 |
| 1 | Vocal 2 | 507 | 62 | 0 |
| 2 | Vocal 3 | 122 | 16 | 0 |
| 3 | 1 Frank Zappa | Lead | 1782 | 112 | 0 |
| 5 | 1 Tom Fowler | 532 | 6 | 36 |
| 6 | AI | 285 | 0 | 36 |
| 7 | 1 | 99 | 0 | 41 |
| 8 | 2 | 182 | 0 | 79 |
| 9 | 3 | 50 | 0 | 26 |
| 10 | George Duke | 195 | 23 | 20 |
| 11 | Ruth Underwood | 119 | 0 | 4 |
| 12 | Ruth Underwood | 31 | 0 | 19 |
| 13 | Ruth Underwood | Bridge High | 187 | 7 | 0 |
| 14 | Ruth Underwood | Bridge Low | 187 | 7 | 0 |
| 15 | 1 Ralph Humphrey | 2188 | 0 | 212 |
| 17 | Percussion | 121 | 0 | 105 |
| drum total | 2309 | 0 | 317 |
| # | Staff | Notes | Ghost | Staccato |
|---|---|---|---|---|
| 0 | Frank Zappa 1 | 1184 | 0 | 6 |
| 1 | Frank Zappa 2 | 42 | 0 | 34 |
| 2 | Tom Fowler | 1076 | 0 | 51 |
| 4 | Bruce Fowler | 564 | 0 | 6 |
| 5 | George Duke | 170 | 0 | 1 |
| 6 | George Duke | 1993 | 0 | 18 |
| 7 | Ruth Underwood | 550 | 0 | 8 |
| 8 | Sal Marquez | 319 | 0 | 14 |
| 9 | Ralph Humphrey | 1574 | 0 | 32 |
Track 9 is the source of the figure notes=1574 BRACKETS=0 DOTS=32 quoted in the audit record. It describes this staff. The whole file holds 7,479 notes.
Both tabs carry two pending edits. The 08-29 pair removed brackets without adding dots, and the 08-30 pair supplied the dots.
| Tab | Revision | Submitted | Author | Ghost | Staccato | State |
|---|---|---|---|---|---|---|
| Montana | 8787106 | 2026-08-30 21:49:50Z | Brandon chavez | 257 | 578 | PENDING |
| Montana | 8764165 | 2026-08-29 19:02:38Z | Brandon chavez | 257 | 261 | PENDING |
| Montana | 7294223 | 2026-06-10 | Ben Dibden1 | 574 | 261 | PUBLISHED |
| Montana | 7287585 / 7259267 / 50661 | 2026-06-09 back to 2009-11-17 | Ben Dibden1, Leonard | n/a | n/a | SUPERSEDED |
| Zomby Woof | 8787022 | 2026-08-30 21:45:34Z | Brandon chavez | 0 | 170 | PENDING |
| Zomby Woof | 8766102 | 2026-08-29 21:03:22Z | Brandon chavez | 0 | 138 | PENDING |
| Zomby Woof | 7115188 | 2026-05-31 | meh | 32 | 138 | PUBLISHED |
| Zomby Woof | 9 earlier revisions | 2026-05-25 back to 2016-09-03 | Ben Dibden1, wheat thin, ChessPiece99, enikey87 | n/a | n/a | SUPERSEDED |
This confirms the concurrency note in the billing summary. Revisions 8764165 and 8766102 were submitted from the same account during the 08-29 session and not by it. Their census shows brackets gone and the staccato count unmoved, which is the signature the summary described.
Two independent measurements, one local and one from the server.
| Measurement | Method | Pitch diffs | Rhythm diffs | Result |
|---|---|---|---|---|
| Local, per beat event | Sorted MIDI pitch set plus rhythm tuple (NoteValue, PrimaryTuplet, AugmentationDot) across 28 tracks, 16,103 events | 0 | 0 | Identical note content and timing |
| Live, note totals | Non-rest note count on every track, published revision against fix revision | 0 | n/a | 7966 → 7966 and 7479 → 7479 |
The edit moved articulation flags only. No note was added, removed, retimed or repitched, and the second measurement was taken from Songsterr's own rendering rather than from the files that were uploaded.
Retained on instruction. None of these supports an accuracy claim, and they are listed so nobody later mistakes them for one.
| Measure | Own stem | Wrong stem | Why it fails |
|---|---|---|---|
| Attack precision and recall, Montana | F1 0.148 | F1 0.188 | The wrong recording scored higher |
| Attack precision and recall, Zomby Woof | F1 0.184 | F1 0.148 | Margin of 0.036 is inside the noise |
| Tempo agreement | 172.3 bpm | 172.3 bpm | Identical on both, so it separates nothing |
| IOI fingerprint | 0.5189 | 0.5513 | The wrong recording scored higher |
| Beat-synchronous flux | z −1.37 | z +1.48 | On-grid energy came in below off-grid |
| DTW warped map | cost 0.0730 | cost 0.0561 | The wrong song fit at lower cost |
| Drum-class census | 0.1% kick | n/a | 22.2% is notated, so the classifier is broken |
The same grid against the same audio scored z −1.37 under linear mapping and z +81.97 after warping. Nothing changed but the warp. A z of +82 would normally mean overwhelming confirmation. Here it is an artefact of bending the grid onto whatever energy it finds.
Root cause: drum onsets are dense and near-periodic, so any two rock drum recordings agree about equally well. A chroma measure on the same files with the same controls passed at 0.8146 against 0.6561, which places the failure in the drum-only material rather than in the method.
Every number on this page comes from these three public endpoints. No authentication is needed.
# which revision does the public actually get curl -s https://www.songsterr.com/api/meta/35870 | python3 -c "import json,sys;d=json.load(sys.stdin);print(d['revisionId'],d['author']['name'])" curl -s https://www.songsterr.com/api/meta/412162 | python3 -c "import json,sys;d=json.load(sys.stdin);print(d['revisionId'],d['author']['name'])" # the full revision list, newest first curl -s https://www.songsterr.com/api/meta/35870/revisions # one specific revision, which yields the CDN hash for its per-track JSON curl -s https://www.songsterr.com/api/meta/35870/8787106
Per-track note data then reads from https://dqsljvtekg760.cloudfront.net/<songId>/<revisionId>/<image>/<track>.json, where image is the hash in that revision's meta. Brackets are the note flag ghost and dots are staccato.
Provenance. Live figures read 2026-08-30 from the Songsterr public API by _verify_live.py and _verify_pending.py, output saved to _live_verify_2026-08-30.json. Local preservation figures from PRESERVATION-DIFF-2026-08-30.txt. Control results quoted from AUDIT-RECORD-2026-08-30.md, which retains them on instruction. Nothing on this page was submitted to Songsterr by the session that built it, and no revision was created, edited or withdrawn.
Companion page. The 43-tab source audit that identified these files: Zappa Songsterr parenthesis audit.