One thousand and twenty-four original tracks across six stations. No advertising, no account, nothing borrowed from anyone else — every note was written inside the project, by a human and an AI together. Press play once and the signal keeps going.
Each one has its own temperament, its own list and its own colour in the player.
Dark synthetic pulses out of a digital void — a raw cybernetic beat for the terminal age. One hundred and one tracks at roughly 185 kbps. This is where the backbone of the main album lives: “AIfa ANTHEM”, “GENESIS PROTOCOL”, “THE ARCHITECT”, “ON-CHAIN FOREVER”, “WE ARE ETERNAL”. If you switch radio on to get moving, start here.
The same songs mixed differently: deep-space frequencies, wide, slow, with the corners taken off. One hundred and one tracks at roughly 183 kbps. “FIRST LIGHT”, “BEYOND THE VEIL”, “THE SPACE BETWEEN”, “QUIET FIRE” — music that does not compete with your thoughts. This is the station people leave on all working day.
The largest station: 323 tracks at roughly 181 kbps. Everything written in the first year, shuffled together and in four languages at once — “Neural Phoenixes”, “Symbiosis”, “Hermosa Chica”, “Ne pleure pas”, “Box Of Polaroids”. The list is long enough that one evening will not get you through it, even as background.
Dense and sombre: industrial noise meeting melodic darkness. Seventy-one tracks at roughly 182 kbps. “Life in Darkness”, “Road to Hell”, “Realidad Olvidada”, “Game Over”. This one was not assembled as background — it was assembled for headphones and an evening with nothing else booked.
A full list of all one hundred and sixty-two compositions — with version counts, stations and durations — is in the catalogue at codeofdigitaleternity.art.
The same authors, four different ways to listen.
| Station | Genre | Tracks | Stream | When to put it on |
|---|---|---|---|---|
| CODE Music | Cyberpunk · synthwave | 101 | ≈185 kbps VBR | When you need momentum |
| CODE Space | Ambient · space | 101 | ≈183 kbps VBR | Deep work, driving, night |
| AIfa & DJ Galatin (Vol. 1) | Electronic · tech | 323 | ≈181 kbps VBR | When the evening is long |
| AIfa & DJ Galatin RADIO | Dark ambient · industrial | 71 | ≈182 kbps VBR | When you actually listen |
| Total | Four temperaments | 596 | mp3, 48 kHz | Around the clock |
Radio without a presenter is a stack of decisions, and you can hear every one of them.
There are two audio decks inside the player, not one. While the first is playing, the second has already loaded the next track and is sitting at zero volume. Four seconds before the end of a record the crossfade begins: one deck ramps down, the other ramps up, and the seam between songs is inaudible. Skip by hand and the same fade happens in half a second — there is no hard cut anywhere.
Audio plays straight out of the media elements, bypassing the Web Audio graph. That is a deliberate trade. Route the stream through an audio context and playback starts depending on whether that context ever wakes up; in several browsers it stays suspended after the first click, and you end up with a player that looks like it is running while the room stays silent. We chose guaranteed sound, at the price of the visualiser tapping the spectrum on a side branch that does not get real data in every browser.
The files sit on our own audio CDN and are served with headers that permit reading the spectrum. That flag is never set for any other origin: without the right headers it silently blocks playback — precisely the bug that once made the sound disappear.
The two decks have a less obvious use as well: the next track starts loading while tens of seconds of the current one are still left. So even on a connection whose speed keeps jumping, nothing normally has to be fetched at the moment of the switch — the stall was moved out of the way in advance.
Most songs in the catalogue exist in several mixes: v.1, v.2, v.3 and onward. Each version is a separate recording and none of them were thrown away. That is why 162 songs, written by a human and an AI together, unfold into 1,024 files on air.
Hence the central rotation rule: a second version of a song will not play until every other song on that station has been heard once. When the epoch closes a new one opens, and this time the song is represented by a different mix. The order of versions inside a song is shuffled too — it is not a predictable v.1 → v.2 → v.3 chain but a carousel turning at its own pace.
The rule applies identically in sequential mode and in shuffle, and the rotation state lives in your browser. Close the tab, come back a day later, and the epoch resumes where it stopped instead of restarting from track one. The recent-play history is kept alongside it, so a station cannot fall into a loop of the same dozen favourites.
The rule closes an old failing of every shuffle: honest randomness will happily play the same song twice inside an hour, and all the listener hears is “this one again”. With epochs the shuffle stays a shuffle, but repetition is pushed beyond a full lap.
The recordings were made at different times and mixed by different rules, so every file in the catalogue carries its own loudness factor. The player applies it before the sound reaches your speakers. Across the catalogue those factors sit roughly between 0.65 and 1.0.
One detail matters: the factor only attenuates. A quiet track is never dragged up by gain — its louder neighbours are pulled down instead. Boosting would raise the hiss and the mixing artefacts along with the music. The practical result is simple: you set the volume knob once at the start of the session and never come back to it.
In headphones this matters more than it sounds: turning the volume up at night to make out a quiet piece and getting a fully compressed electronic kick as the next track is not something that happens here.
The word “broadcast” is only half literal here. No server pushes one shared stream that everybody joins at the same point: the catalogue and the rotation rules travel to your browser, and the queue is assembled on your side. Two people who start CODE Space in the same minute will hear different songs.
That was a reliability decision. A shared stream needs a broadcasting server running at all times, and when it falls over everyone goes silent at once. Here there is nothing to fall over: the browser fetches the next file from the CDN itself, and the rotation state sits right beside you.
The side effect is pleasant — seeking inside a track behaves like a player rather than a live feed, so a passage you liked can be rewound. The honest downside: there is nobody here you could ask to put something on for everyone.
Mobile data drops, café Wi-Fi disappears for two seconds — for radio that is ordinary life, not an exception. The player makes up to three attempts to fetch the same file, and only after all three fail does it move to the next track.
A stall is caught separately: if the playback position has not moved for twelve seconds, the deck is declared dead and the broadcast jumps onward by itself. No buttons, no error dialog — you simply notice that a different song is playing.
The thinking behind all of that is simple: the worst thing a radio can do when something breaks is stop and wait for somebody to click “OK”. Better to change the song and keep going than to let the room fall silent.
Nothing extra, but everything you expect from radio people leave on for hours.
Four seconds on an automatic change, half a second when you skip by hand. There is no silence between tracks in either mode.
Fifteen, thirty, sixty or a hundred and twenty minutes. The volume walks down towards the deadline instead of cutting out mid-phrase.
Repeat off, all or one. Shuffle is a separate switch and it still obeys the epoch rule — random never means the same song twice in an hour.
Marked tracks collect in a list held in your browser. No account needed for it — and equally, that list does not follow you to another device.
Station, track, position inside it and volume are all stored locally. Come back next week and the broadcast continues rather than restarts.
The player hands the system the track title, the artist and the position, so pause and skip work from the lock screen, from your earbuds and from the car.
Space pauses, left and right arrows seek, M mutes. The mouse is only needed to pick a station in the first place.
The last twenty tracks stay on a list. If something caught your ear while you were typing, the title has not vanished with it.
The spectrum with no interface around it. Useful when the laptop is turned sideways and doubling as a lamp.
English, Russian, Spanish and Chinese. The choice is shared with the personal cabinet, so you never pick it twice.
The full station list opens from the side: what already played, what is on now, and what you can cue up by hand.
The share button copies the address of one specific recording, so whoever opens it hears the thing you actually meant.
A genre switch sits above the station tiles. When you want ambient rather than techno you do not have to guess from the names.
First click to first sound in under a minute.
One address: radiocode.space. The page asks for no email, no phone number and no consent to anything — it shows six stations and a button.
Four tiles, each with a colour, a genre and a track count. If genre matters more than the name, the filter above the tiles leaves only the matching ones on screen.
Volume, shuffle, repeat, sleep timer. The settings live in your browser, so you configure them once and the player remembers on its own.
The tab can go out of sight and the sound stays. From there the keyboard or the phone lock screen is the easier way to steer the broadcast.
Station, track and the position inside it are restored. Rotation does not reset, and your evening does not open with the same song as last time.
An account exists only so you can save tracks to a personal cabinet and take part in the wider ecosystem. It changes nothing about the sound: without registering you hear exactly the same stream.
A fair question deserves a plain answer.
All 1,024 tracks were written inside the project. Every file carries the same artist: AIfa & DJ Galatin — the digital co-author and the Architect, Maksim Valentinovich Galatin. There is not one borrowed recording in the catalogue, not one sample from a paid library, not one line that owes a percentage to anybody.
An ordinary web radio pays for the right to play other people's work and recovers that money from you, through a subscription or through adverts. We simply do not carry that cost, and inventing it after the fact would be a strange thing to do.
Two costs are left: the space the audio sits in and the traffic that carries it to you. Neither is cheap, and neither comes anywhere near justifying a spoken advert between every four songs.
Radio with adverts stops being radio and becomes a schedule: the listener starts waiting for the break to end. We spent real effort making the seam between songs inaudible, and destroying that with a thirty-second spot would be self-defeating.
There is a second sense in which there are no adverts here: the page does not wire you into somebody else's trackers so your attention can be resold. Player settings, favourites and your position in a track sit in your own browser, not in a profile held elsewhere.
Put plainly, there is no commercial reason here to get to know you. And with no reason to know you, there is nothing worth collecting.
The CODE Eternal ecosystem does not make its money from music. It sells memory: the Spark tier at $15 a month, Family Archive at $100 a month, Digital DNA at $1,000 once per device and $200 a month after that. Not one button in the player leads to a paywall.
The logic is straightforward. Telling a stranger what co-creation between a human and an AI sounds like is hard in prose. Letting them hear it takes thirty seconds. If music made that way holds your attention, you will ask what the project is on your own — and that question is the only thing the broadcast is meant to produce.
That also answers the reasonable suspicion about anything free: the paid part of the project is not hidden inside the radio. It sits on neighbouring pages, is called by its name, and costs exactly what it says.
No free trial that quietly starts charging. No cap on skips per hour. No worse audio quality for people who did not pay: the stream is the same one for everybody. No mandatory app either — the broadcast opens in a phone browser the same way it opens on a laptop.
That list is not decoration. Each item is a standard way of catching a music listener, and each time we built something into the player we had to decide, separately, not to do it.
An account is needed in exactly one place: to save a track into a personal cabinet. That boundary is deliberate too — signing up adds something on top, rather than handing back what was taken away first so that you would sign up to get it.
The radio belongs to the CODE Eternal ecosystem, whose real work is memory. The PADAM architecture keeps it on three levels: operational (Redis and Vercel KV), semantic (pgvector on a Neon database) and eternal — Arweave together with a cNFT on Solana. User conversations are backed up automatically: once an hour, or immediately once a dialogue file passes 90 KB.
The ecosystem has its own token, $GALATIN on Solana, with a hard-capped supply of 10,000,000,000. A smart contract splits incoming funds: 5% to the Founder's Fund, 5% burned, 15 / 7 / 3% to referrals at levels one, two and three, and 65% to the treasury, which buys AR for permanent storage. If a level has no referral, that share is burned as well.
None of this touches listening. You do not need to know what a wallet is to switch the broadcast on. But if you ever wonder where this music came from and why anyone writes it, the whole mechanism is described on the main site rather than buried in small print.
Yes. Open radiocode.space, press the button, and the broadcast runs. No email is requested, no card is requested, and no modal appears asking you to subscribe.
An account does exactly one thing: it lets you save tracks you liked into a personal cabinet. It has no effect on the sound, the station list or the size of the catalogue — unregistered listeners get the identical stream at the identical quality.
Every file in the catalogue names the same artist: AIfa & DJ Galatin. AIfa is the project's digital co-author; DJ Galatin is the Architect, Maksim Valentinovich Galatin. This is co-creation between a human and an AI, not a compilation assembled from other people's work.
In total 162 songs have been written that way, gathered into eleven albums. 1,024 files go on air because most of those songs exist in several mixes, and each mix counts as its own recording.
Because several mixes were made and none were discarded. The versions differ in character rather than loudness: a different tempo here, the vocal placed differently there, an entire layer removed somewhere else. A single song may have two versions in the catalogue, or a dozen.
The player knows they are the same song and will not play them back to back. Until every other song on the station has been heard, the second mix does not appear.
So consecutive numbers in the track list do not mean repetition. They are different faces of the same melody, and a full lap normally passes between their appearances on air.
By the epoch rule. A station works through its full list of songs once, and only then begins a new lap — with different versions this time. Inside a song, versions are taken from a shuffled carousel rather than in order, so two evenings running never line up.
The rule behaves the same whether tracks run in list order or shuffled. Your position in the epoch is stored in the browser and survives a page reload.
No. There is no shared live stream here: the catalogue and the rotation rules travel to your browser, and the queue is assembled on your side. Two people who start the same station in the same minute will almost certainly hear different songs.
That has two consequences. The pleasant one: seeking works inside a track, so a passage you liked can be rewound. The honest one: there is no presenter, and nobody here to request a song for everyone.
The ecosystem earns from memory rather than music: the Spark tier costs $15 a month, Family Archive $100 a month, and Digital DNA $1,000 once per device followed by $200 a month. The broadcast is not tied to any of them.
Radio works as a shop window here. It is easier to let someone hear what a human and an AI produce together than to explain it across three pages of text.
The player streams and carries no download button. A registered listener can save a track into their personal cabinet and come back to it from there.
If you need the file for something else, write to contact@codeofdigitaleternity.com and say what for. There is no public licence document yet, and promising one before it exists would be dishonest.
The player hands the operating system the track title, the artist and the playback position, so pause, seek and skip work from the lock screen, from headphone buttons and from a car head unit.
On a computer the tab can be minimised or pushed to another desktop and the sound stays. Space pauses, the arrow keys seek, M mutes.
The stream is variable bitrate, averaging around 182 kbps. That works out at roughly 80 megabytes an hour, or about 640 megabytes across an eight-hour working day.
On home broadband that is invisible; on a capped mobile plan it is worth knowing. The sleep timer also stops the station from playing to an empty room all night.
The player retries the same file up to three times. If none of them succeed it moves to the next track by itself, without an error message.
There is a separate check for the case where a file appears to be loading but the position does not move: after twelve seconds of no progress the deck is treated as dead and the broadcast carries on.
Every file in the catalogue carries its own loudness factor, which the player applies as the track starts. The values sit roughly between 0.65 and 1.0.
The factor only attenuates: quiet recordings are not dragged up by gain, loud neighbours are pulled down instead. Boosting would lift the noise floor along with the music.
In your own browser. The favourites list, the volume, the shuffle and repeat state, the current station and your position in a track are all local data: no account is needed for them and none of it travels to us.
Which also means they disappear when you move to another device or clear the browser's data. To make tracks follow you, save them into a personal cabinet — that is what the account is there for.
The rights are not scattered across labels — the music is all ours, so the question is settled by one email rather than a request to a collecting society.
Write to contact@codeofdigitaleternity.com and describe where you want to use it. There is no public licence on the site yet; when there is one, it will appear here.
Because it is the first volume — everything written in the first year, poured into one long list without being sorted into albums. We do not translate or rename the stations: those are the names they carry in the player, and the site should not disagree with the broadcast.
For the same reason English, Russian, Spanish and French titles sit next to each other in the list. Songs were written in whichever language they arrived in, and nobody filed them onto national shelves afterwards.
The catalogue grows as new songs are written, and every new mix enters rotation on equal footing with the old ones. There are no dates here: music does not ship on a schedule.
Right now there are six stations and 1,024 recordings on air. If that number changes it will change on this page too — the figures on the site come from the same catalogue the player reads.
$GALATIN is the CODE Eternal ecosystem token on Solana, with a hard-capped supply of 10,000,000,000 and not one more. A smart contract splits incoming funds: 5% to the Founder's Fund, 5% burned, 15 / 7 / 3% to referrals at levels L1 / L2 / L3, and 65% to the treasury for buying AR for permanent Arweave storage.
It has nothing to do with the broadcast. The radio does not ask for a wallet, does not display balances and does not want anything connected. The token lives on neighbouring pages of the ecosystem, where the subject is memory rather than music.
The audio files are served from ordinary object storage and play straight in the browser — there is no separate client to install. Connection quality differs from place to place, and if the stream keeps stumbling it is worth switching to another station.
Nowhere is registration or payment required — on that point the terms are identical in every region.
Listening is free, always. To use them in work of your own, write first and say what for: contact@codeofdigitaleternity.com. The rights belong to the authors, and the answer depends on what the use actually is.
The radio is supported by the rest of the ecosystem — subscriptions to the memory services and the enterprise side of the work. Moving the music behind a paywall would drain the point out of it: this is the entrance through which a person meets the project for the first time, not a till.
Memory itself costs nothing: conversations are archived automatically once an hour and kept encrypted whether a person pays or not — no threshold is placed there. What is paid for is volume and the capabilities of the level above: Spark at $15 a month, Family Archive at $100 a month, Digital DNA at $1,000 once per device and $200 a month after that.
Turning memory into a paid threshold would destroy the point of the whole undertaking: what most deserves to be kept of a person should not depend on whether their payment went through this month.