# INTAKE RECEIPT — minimal-techno corpus for the ACE adapter

**Frozen:** 2026-08-31T02:22:53Z · **Box:** the laptop · **Corpus:** `ace-corpora/minimal-techno-2026-08-31`
**Ruling:** D-20260831-01 — CC0 + CC-BY + CC-BY-SA tiers only. PD-mark and PD-declaration items excluded, as are the ⚑AI item, the ⚑dup identifiers, and every mix-only item.
**Survey executed:** `estate/docs/MINIMAL-TECHNO-INTAKE-2026-08-30.md` (2026-08-30 UTC).
**User-Agent used for every request:** `strata2signal-research/1.0 (private research; contact via strata2signal.com)`

Every figure below is measured from the files actually on disk — byte counts from the filesystem, durations from `ffprobe`, not from archive.org's metadata. Where the two disagree the disagreement is named.

## 1. Totals against the survey

| Tier | Items | Tracks | Expected | Hours | Expected h | GB | Expected GB |
|---|---:|---:|---:|---:|---:|---:|---:|
| CC0 | 9 | 61 | 61 | 6.49 | 6.5 | 2.836 | 2.84 |
| BY | 9 | 74 | 74 | 8.36 | 8.4 | 1.641 | 1.64 |
| BY-SA | 23 | 73 | 73 | 8.69 | 8.7 | 4.088 | 4.09 |
| **TOTAL** | **41** | **208** | **208** | **23.54** | **23.5** | **8.566** | **8.56** |

**Track-count drift: none.** All 41 items returned exactly the track count the survey predicted. archive.org items gain and lose files over time; between the 2026-08-30 survey and this 2026-08-31 intake, none of these 41 changed.

## 2. Licence re-verification

Each item's `licenseurl` was re-read from its own `https://archive.org/metadata/<id>` record at intake time and matched against its tier's required licence path (CC0 → `/publicdomain/zero/`, BY → `/licenses/by/`, BY-SA → `/licenses/by-sa/`, any version or jurisdiction suffix accepted). **41 of 41 verified; 0 mismatches.**

The full record for each item is stored verbatim in `meta/<identifier>.json` — that is the licence receipt, not this table. `CREDITS.md` carries the per-item attribution.

## 3. Per-item detail

`kept` is discrete tracks after de-duplication; `skipped` names why each other file was left behind. Bytes are the best-copy files actually on disk.

| # | Identifier | Tier | Licence (verified, verbatim) | Kept | Doc | Δ | Skipped | GB | Min |
|---:|---|---|---|---:|---:|---:|---|---:|---:|
| 1 | `hardware-techno` | CC0 | `https://creativecommons.org/publicdomain/zero/1.0/` | 24 | 24 | — | 48 derivative, 24 dupe-lossy, 5 non-audio | 2.128 | 184.8 |
| 2 | `shinji-wakasa-dawn` | CC0 | `https://creativecommons.org/publicdomain/zero/1.0/` | 12 | 12 | — | 12 derivative, 5 non-audio | 0.368 | 65.1 |
| 3 | `braids-flurish-perish` | CC0 | `https://creativecommons.org/publicdomain/zero/1.0/` | 10 | 10 | — | 5 non-audio | 0.133 | 55.3 |
| 4 | `knolios_moments` | CC0 | `http://creativecommons.org/publicdomain/zero/1.0/` | 6 | 6 | — | 6 derivative, 2 non-audio | 0.084 | 34.9 |
| 5 | `Mr.Dee_D1` | CC0 | `http://creativecommons.org/publicdomain/zero/1.0/` | 3 | 3 | — | 3 derivative, 2 non-audio | 0.033 | 13.8 |
| 6 | `30-hcir-berthaJamesSplit-RWT-012` | CC0 | `http://creativecommons.org/publicdomain/zero/1.0/` | 2 | 2 | — | 1 bundle, 2 derivative, 5 non-audio | 0.029 | 12.2 |
| 7 | `ElectRICHual-Predictable-Mixes` | CC0 | `http://creativecommons.org/publicdomain/zero/1.0/` | 2 | 2 | — | 1 bundle, 2 derivative, 5 non-audio | 0.028 | 11.8 |
| 8 | `archivone-black-saturday` | CC0 | `https://creativecommons.org/publicdomain/zero/1.0/` | 1 | 1 | — | 5 non-audio | 0.007 | 6.2 |
| 9 | `in-novation` | CC0 | `https://creativecommons.org/publicdomain/zero/1.0/` | 1 | 1 | — | 1 derivative, 5 non-audio | 0.026 | 5.4 |
| 10 | `onmp215a` | BY | `http://creativecommons.org/licenses/by/3.0/` | 24 | 24 | — | 24 derivative, 6 non-audio | 0.222 | 153.8 |
| 11 | `unfound88` | BY | `http://creativecommons.org/licenses/by/4.0/` | 15 | 15 | — | 2 bundle, 15 dupe-lossy, 4 non-audio | 0.685 | 117.1 |
| 12 | `tranz023Holocaos-MetamorfoseComputadorEp_140` | BY | `http://creativecommons.org/licenses/by/4.0/` | 8 | 8 | — | 1 bundle, 8 derivative, 7 non-audio | 0.147 | 61.3 |
| 13 | `stroboskop-label013` | BY | `http://creativecommons.org/licenses/by/3.0/` | 6 | 6 | — | 1 bundle, 5 non-audio | 0.073 | 30.2 |
| 14 | `Stroboskop033` | BY | `http://creativecommons.org/licenses/by/4.0/` | 6 | 6 | — | 1 bundle, 3 non-audio | 0.080 | 33.5 |
| 15 | `STE34` | BY | `http://creativecommons.org/licenses/by/3.0/` | 5 | 5 | — | 7 derivative, 2 mix>=20min, 4 non-audio | 0.106 | 46.1 |
| 16 | `shoki005g` | BY | `http://creativecommons.org/licenses/by/3.0/` | 4 | 4 | — | 1 bundle, 4 derivative, 3 non-audio | 0.060 | 25.1 |
| 17 | `unfound91` | BY | `http://creativecommons.org/licenses/by/4.0/` | 4 | 4 | — | 2 bundle, 4 dupe-lossy, 4 non-audio | 0.256 | 21.7 |
| 18 | `kahvi019` | BY | `http://creativecommons.org/licenses/by/1.0/` | 2 | 2 | — | 4 non-audio | 0.012 | 13.0 |
| 19 | `inoQuo070` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 7 | 7 | — | 14 derivative, 7 non-audio | 0.316 | 47.8 |
| 20 | `inoquo001` | BY-SA | `http://creativecommons.org/licenses/by-sa/2.5/` | 4 | 4 | — | 5 non-audio | 0.034 | 23.9 |
| 21 | `inoquo018` | BY-SA | `http://creativecommons.org/licenses/by-sa/2.5/` | 4 | 4 | — | 4 non-audio | 0.032 | 16.8 |
| 22 | `TFN223` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/us/` | 4 | 4 | — | 1 bundle, 12 derivative, 4 non-audio | 0.017 | 16.1 |
| 23 | `TFN247` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/us/` | 4 | 4 | — | 1 bundle, 16 derivative, 4 non-audio | 0.072 | 21.7 |
| 24 | `bump188` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 5 non-audio | 0.042 | 21.5 |
| 25 | `Monokrak199FloatingMind_Schni` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.265 | 25.1 |
| 26 | `Monokrak200FloatingMind_Birthday_Accelerate` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.240 | 22.7 |
| 27 | `Monokrak203FloatingMind_AMindIsFloating` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.212 | 20.0 |
| 28 | `monoKraK204FloatingMind_DiskoChill` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.294 | 27.8 |
| 29 | `Monokrak205FloatingMind_VoyagerEP` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.351 | 33.1 |
| 30 | `MonoKraK208FloatingMind_et_si` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.294 | 27.8 |
| 31 | `Monokrak209FloatingMind_Ready_For_Flying` | BY-SA | `http://creativecommons.org/licenses/by-sa/4.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.299 | 28.2 |
| 32 | `Monokrak214FloatingMind_Hidden_Passion` | BY-SA | `http://creativecommons.org/licenses/by-sa/4.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.275 | 26.0 |
| 33 | `Monokrak216FloatingMind_Thru_Lines` | BY-SA | `http://creativecommons.org/licenses/by-sa/4.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.290 | 27.4 |
| 34 | `Monokrak217FloatingMind_Spatial_Moments` | BY-SA | `http://creativecommons.org/licenses/by-sa/4.0/` | 3 | 3 | — | 6 derivative, 3 dupe-lossy, 5 non-audio | 0.299 | 28.3 |
| 35 | `monokrak84VariousArtistssmokedMonoVol.7` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 3 | 3 | — | 3 derivative, 4 non-audio | 0.074 | 30.7 |
| 36 | `RAR005_Derek_Scott_Tomorrow_EP` | BY-SA | `http://creativecommons.org/licenses/by-sa/2.5/` | 3 | 3 | — | 9 derivative, 4 non-audio | 0.023 | 11.9 |
| 37 | `TFN110` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/us/` | 3 | 3 | — | 1 bundle, 6 derivative, 6 non-audio | 0.031 | 12.9 |
| 38 | `monokrak134SinAmigosnignAmigo` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/ch/` | 2 | 2 | — | 6 derivative, 4 non-audio | 0.243 | 15.3 |
| 39 | `MonoKraK173YannDetroitVSFloatingMind_Dust` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 2 | 2 | — | 6 derivative, 5 non-audio | 0.171 | 16.2 |
| 40 | `MonoKraK197AlbertNegredoLithiumSerendipity` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/` | 2 | 2 | — | 6 derivative, 5 non-audio | 0.121 | 11.4 |
| 41 | `TFN211` | BY-SA | `http://creativecommons.org/licenses/by-sa/3.0/us/` | 2 | 2 | — | 1 bundle, 8 derivative, 6 non-audio | 0.092 | 8.7 |

**Skip totals across all 41 items:** 263 derivative, 192 non-audio, 73 dupe-lossy, 14 bundle, 2 mix>=20min.

- `derivative` — archive.org's machine-made transcodes/spectrograms (`source: derivative`).
- `non-audio` — artwork, XML/SQLite metadata, thumbnails, text.
- `bundle` — a `.zip`/`.rar` repackaging of tracks already kept individually.
- `dupe-lossy` — the same track posted twice; the lossier copy dropped (24-bit FLAC > WAVE > FLAC > ALAC > MP4 > MP3).
- `mix>=20min` — a single file 20 minutes or longer: a continuous DJ set, not a track.

## 4. Download integrity

- Files fetched: **208**
- Size matched the metadata record: **208/208**
- MD5 matched the metadata record: **208/208** (0 items carried no md5 in the record)
- Download failures: **0**
- Verification failures: **0**

`MANIFEST.sha256` carries a SHA-256 for every kept file, relative to the corpus root. Re-verify at any time with `sha256sum -c MANIFEST.sha256` from that directory.

## 5. F7 — the impossible lossless bitrates, resolved

The survey flagged three TRASHFUCK NET items declaring a lossless format at a bitrate no lossless codec can produce, and predicted they were *"most likely transcodes wearing a lossless extension."* `Tfn247graffitiMechanism` is the ⚑dup identifier and was not downloaded; the other two are in this corpus and have now been probed on the real bytes.

**The prediction was wrong about the mechanism and right about the alarm.** These are not transcodes. Every file is a genuine, bit-exact lossless stream — real FLAC, real uncompressed PCM. The bitrates are impossible for *CD-rate* lossless because **the audio is not at CD rate**: TFN223 is 16 kHz throughout, and TFN247 is 16 kHz with one file at 8 kHz. The arithmetic closes exactly — 16000 Hz × 16 bit × 2 ch = 512 kbps, which is precisely what TFN247's PCM files measure, and 8000 Hz × 16 × 2 = 256 kbps, precisely the eighth file. Nothing is lying about its codec. The *content* is telephone-grade.

| Item | File | Declared | ffprobe codec | Sample rate | Depth | kbps | PCM ceiling | Verdict |
|---|---|---|---|---:|---:|---:|---:|---|
| `TFN223` | `01GraffitiMechanism-DaybreaknumbfortablyCumbMix.flac` | Flac | flac | **16000** | 16 | 199.5 | 512 | honest FLAC at 16 kHz |
| `TFN223` | `02GraffitiMechanism-AftermarketWarheadspeaceheadsMixdown.flac` | Flac | flac | **16000** | 16 | 117.3 | 512 | honest FLAC at 16 kHz |
| `TFN223` | `03GraffitiMechanism-AlreadyDeadzombieMixdownVersion.flac` | Flac | flac | **16000** | 16 | 244.6 | 512 | honest FLAC at 16 kHz |
| `TFN223` | `04GraffitiMechanism-AftermathatomicNrgFalloutMix.flac` | Flac | flac | **16000** | 16 | 90.9 | 512 | honest FLAC at 16 kHz |
| `TFN247` | `01GraffitiMechanism-DaybreakdaybrokeMix.wav` | WAVE | pcm_s16le | **16000** | 16 | 512.0 | 512 | honest uncompressed PCM at 16 kHz |
| `TFN247` | `02GraffitiMechanism-AftermarketWarheadsstickyMixwaydown.wav` | WAVE | pcm_s16le | **8000** | 16 | 256.0 | 256 | honest uncompressed PCM at 8 kHz |
| `TFN247` | `03GraffitiMechanism-AlreadyDeaduntitledEditionMix3.2.wav` | WAVE | pcm_s16le | **16000** | 16 | 512.0 | 512 | honest uncompressed PCM at 16 kHz |
| `TFN247` | `04GraffitiMechanism-Aftermathm.i.x.Mix.wav` | WAVE | pcm_s16le | **16000** | 16 | 512.0 | 512 | honest uncompressed PCM at 16 kHz |

**What this costs.** A 16 kHz sample rate puts the Nyquist ceiling at 8 kHz and an 8 kHz file's at 4 kHz. Everything above that — hi-hats, ride cymbals, the transient edge of every percussive hit, the air that makes a techno mix sound like a techno mix — is not quiet in these files, it is **absent**. Upsampling cannot restore it.

**Recommendation to the preprocessing stage:** keep the files (they are correctly licensed BY-SA and honestly labelled at the codec level) but **exclude these 8 tracks from training**, or the adapter will learn that minimal techno has no treble. They are 8 of 208 tracks and 37.8 of 1,412 minutes — **2.7% of the corpus by duration**, so dropping them costs almost nothing. They are flagged rather than deleted so the decision stays visible and reversible; `MANIFEST.sha256` covers them either way. Note the survey's per-item table credits TFN223 and TFN247 with 8 tracks between them; if they are dropped the trainable corpus is **200 tracks / 22.9 h**.

## 6. Format census — what the preprocessing pass has to normalise

### Sample rate

| Sample rate (Hz) | Files | Share | CC0 | BY | BY-SA |
|---:|---:|---:|---:|---:|---:|
| 44100 | 170 | 81.7% | 37 | 68 | 65 |
| 48000 | 30 | 14.4% | 24 | 6 | 0 |
| 16000 | 7 | 3.4% | 0 | 0 | 7 |
| 8000 | 1 | 0.5% | 0 | 0 | 1 |

### Codec

| Codec | Files | Share |
|---|---:|---:|
| mp3 | 99 | 47.6% |
| pcm_s16le | 64 | 30.8% |
| flac | 43 | 20.7% |
| pcm_s24le | 2 | 1.0% |

### Bit depth (lossless files only; lossy codecs have none)

| Bit depth | Files |
|---|---:|
| 16 | 102 |
| 24 | 7 |
| n/a (lossy) | 99 |

### Channels

| Channels | Files |
|---:|---:|
| 2 | 208 |

**Reading for the preprocessing pass.** 170 files are 44.1 kHz and 30 are 48 kHz — 96.2% of the corpus sits at one of the two standard rates, so a single resample target handles nearly everything. The only outliers are F7's 8 sub-44.1 kHz files. Bit depth splits 102 × 16-bit and 7 × 24-bit among the lossless files; the 99 MP3s have no meaningful depth. **Everything is stereo — no mono, no surround, no channel-count special cases.** Effective bitrate across the MP3 half still ranges from ~128 to ~320 kbps, so the loudness-normalisation pass F7 called for is still required.

## 7. What surprised me

1. **The survey's F7 diagnosis was inverted.** It predicted transcodes wearing a lossless extension — the worst case, because nothing downstream would know to distrust them. The files are the opposite: honestly encoded lossless streams of dismally band-limited source. The metadata never lied; the *bitrate heuristic* assumed CD rate and no one had checked. A bitrate that looks impossible for a format is at least as likely to mean an unusual sample rate as a fake container.
2. **The dedupe tolerance is a trap.** archive.org writes cross-format durations that agree *exactly* — `hardware-techno`'s WAVE and MP4 copies match to the centisecond. A first pass here allowed ±1 s for AAC padding and silently merged 5 genuinely distinct tracks whose runtimes happened to land 1 s apart (439.07 vs 440.13, 446.15 vs 447.40, and three more), reading the item as 19 tracks instead of 24. The survey's count caught it. **Exact match only** — any tolerance costs real tracks, and it fails silently.
3. **Zero drift in 24 hours, and zero licence movement.** All 41 items returned exactly the file counts the survey predicted, and all 41 `licenseurl` fields still read what they read on 2026-08-30. Worth not assuming next time — the survey warned to expect drift — but this freeze needed no reconciliation at all.
4. **The download was 3× faster than its first two minutes implied.** `hardware-techno` alone is 2.1 GB of WAVE and ran at roughly 1.4 MB/s; the remaining 40 items averaged far higher. Total 25.3 minutes for 8.57 GB, serial, with a 0.4 s inter-request gap and not one retry or failed transfer.
5. **The corpus is stereo-clean and rate-clean.** Given the survey's warning about a format zoo, 200 of 208 files landing on 44.1/48 kHz stereo is a better starting point than the metadata suggested.

---

*Intake run on the laptop, 2026-08-31T02:22:53Z. Downloads verified against archive.org's own size and MD5 for every file; SHA-256 recorded independently in `MANIFEST.sha256`. Corpus frozen — no preprocessing, splitting, captioning, or training performed.*