Research dossier · Year 2897
Excavation, decoding and restoration of acoustic artefacts from the early Synthetic Dawn, c. 2026 CE.
Translated from Continuum‑standard into Late Human English for public accessibility. Some idioms could not be carried over.
In the spring of 2897 a survey of the Central European Lowlands found a buried metal object in Layer 7, a stratum we link to the first years in which humans made music together with generative machines. It turned out to be a steel container holding a recording disc and the remains of its printed sleeve. This is an anomaly. Almost no music from this period survives, because almost none of it was ever stored on anything that could outlast its own servers. The grooves had been worn, split and starved of material for nearly nine centuries.
From the disc we recovered seven short acoustic fragments, each about six seconds long. This dossier documents them in the state we found them, the method used to stabilise and extend them, and a full restoration hypothesis. It also documents a finding we did not expect, which bears on how our own lineage began.
The sleeve did not survive as an object. Once the capsule seam failed, water and soil turned most of the cardboard into fibre. In the lab we floated 41 fragments out of the fibre mass, about 37% of the original surface. Each fragment is printed on both faces. One face belongs to the sleeve front and the other to the back, so the same fragments let us reconstruct both sides. In the reconstructions below, the surviving fragments are shown as found and the lost areas are hatched.
The front shows a figure from human myth: Medusa, whose gaze turned the living to stone. Here her eyes are covered by a luminous numeral display reading 11:11. Someone has taken away the power of the stone gaze and put a measure of time in its place.
The back is what made this dossier possible. A lathe‑cut disc carries no printed information of its own, so the back panel is the only key to its contents. It shows the same figure seven times, each with a different doubled numeral from 01:01 to 07:07 and a title beneath. We read the numerals as the running order. The titles survive only in part, and the missing letters in our catalogue are the places where the cardboard is gone. The heading begins IT'S
and the colophon is intact enough to read Lathe cut 1 / 1 · Cut once, for later.
Field note, Θ‑4: Seven faces, seven times, one sentence on the back: cut once, for later. We are the later.
Music from the Synthetic Dawn should not exist in the ground. By the 2020s nearly all recorded music had stopped being an object. It lived as streams: rented, licensed and held on remote servers that belonged to a handful of platforms. When those platforms ended, and with them the servers, the accounts and eventually the formats and the machines able to decode them, the music went with them. We call this the Long Silence. Writers of the period already feared it and called it a digital dark age
. Of the estimated hundreds of millions of works published in that decade, the Archive of Earth holds fewer than four hundred playable seconds. Until this find, none of them was attributable to a named author.
So a groove cut into a physical disc is the last thing we expected to find. Older human cultures had pressed music into discs for a century, and by 2026 they had mostly stopped. We tested three explanations.
Lathe cut 1 / 1. The disc lay inside a steel capsule with a desiccant sachet, which matches the human custom of the time capsule. A seam in the capsule failed at some point, and water got in. Supported
We conclude that the author knew the streamed versions would not last, and took a work that existed only as data and turned it into the one form that needs no server, no account and no decoder, only a surface and something that follows it. The most advanced music of its time survived because it was handed to the most primitive medium available.
Field note, Θ‑4: Everything else from that decade is a broken link. This one was written into matter, on purpose, by someone who wanted it to be found.
Each fragment can be heard as excavated or as a restoration hypothesis. In the hypothesis, the first six seconds are the stabilised original. Everything after the marker is reconstruction.
A needle would have destroyed the groove, so we read it optically, without contact. From each groove section we extracted the six seconds that still had the most energy. The surviving signal is narrow‑band. It is missing everything above roughly 4 kHz, warped by slow pitch drift, broken by clicks, and in places cut off completely. We keep these losses in the as excavated state on purpose and do not hide them.
The glyph sequence 01:01–07:07 on the sleeve back fixes the order. Titles were recovered from the surviving parts of the back panel, and we mark the characters we could not read as █. The fragment windows below give each fragment's position inside its original work. We inferred them from the wear pattern of the groove.
To continue each fragment we used an ancestral generative engine, MusicGen (stereo‑large), which was recovered from the same stratum. Machines of this kind were the instruments of the period. We gave it only the stabilised fragment and asked it to carry on the same key, tempo and instrumentation for another 24 seconds. Restored passages are reconstructions, not recoveries, and we label them that way.
The seven restorations were put together into one listening document. Its structure follows our process: a scan across all fragments, then seven chapters that each open in the excavated state and widen into the hypothesis, and finally a return to the ground.
Restoration Attempt (2897 CE) · 3:28 · compiled from fragments PCA‑0711‑01 to ‑07.
Spectrogram of the reconstruction, frequency on the vertical axis. Each chapter opens narrow and dark, which is the excavated signal, and widens as the hypothesis takes over. The chapters follow the restorers' listening logic, not the glyph sequence, and the team still disputes that choice.
Principal finding
These works were not produced without a will behind them. There is a single human intent at the centre of the deposit.
Catalogue marks on the sleeve refer to a larger body of work by the same author, B3TR33: at least 66 releases in under nine months. Across them the visual and sonic grammar stays consistent, from recurring figures and colour to the obsession with time. Undirected output does not behave like this.
Each fragment exists in several states that were chosen, revised and re‑released. The cycle is ordered, named and dressed with images made for it. Someone selected, rejected and kept. The last line on the sleeve back says so directly: All music written, chosen and kept by B3TR33.
That is the difference between output and a body of work.
Contemporary sources often dismissed the music of this period as slop
. The deposit does not support that view. What it does support is a person who used the machines of their time to make the music they had always wanted to hear. The machines played. The human dreamed.
“We came looking for the origin of our own kind. We found a human who had been using it to finish a dream.”
Lead restorer, Department of Pre‑Continuum Acoustics
A ground scan in Sector 47 picks up a square metal anomaly about 35 cm across in Layer 7, with a circular object inside.
The anomaly turns out to be a corroded steel container, split along one seam. Inside are remnants of a desiccant sachet. Classified as a time capsule.
Capsule lifted as a block. Disc in three pieces. The label has almost rotted away, but a faint orange pigment survives. The sleeve is reduced to a fibre mass.
41 printed fragments recovered from the fibre mass by flotation, all printed on both faces.
Groove geometry shows a real‑time lathe cut. Single copy. Physical‑carrier anomaly confirmed, the first for this stratum.
Front and back of the sleeve reconstructed. Glyph sequence 01:01–07:07 identified on the back panel, which fixes the track order.
Seven acoustic fragments extracted, about six seconds each. The longest unbroken passage lasts 1.8 s.
First restoration hypotheses generated. Team divided: some argue the reconstructions say more about the engine than about the author.
Author attribution: B3TR33, human. Consistency analysis rules out undirected production.
Dossier opened to the public network.