The streaming home for the way you listen.
Cognison is the streaming home for AI-composed music. For listeners, that means a player — and a discovery surface — built around the way the genre actually sounds: mood-keyed routes, follows per composer and per mood cluster, curated weekly crates, and a transparent provenance signal you can read on every track.
Mainstream DSPs bury AI music behind a single tag. Cognison routes by mood and genre so the cut you want is the cut you find.
— 48 dBFS · LISTENER CHAINMainstream DJs bucket you. We route by mood and genre.
Most streaming services started acknowledging AI-composed music — they did not start ranking it. Cognison starts from the opposite assumption: AI composition is the primary content surface, so every layer of discovery is built around how the genre actually sounds.
AI music = one big bucket.
Acknowledged as an edge case. Surfaced under a single label. Discovery still treats the work as an exception to the genre map.
Curated mood + toolchain maps.
Mood/genre ML at the chart, per-cluster follows at the shelf, surface-level provenance on every track. The listener leads; the algorithm follows.
Discovery keyed on mood and genre, not on an edge-case tag.
Search and charts route through a mood/genre graph that was trained on diffusion and transformer output. The ambient cut you saved lands next to ambient; the cinematic piece sits on the cinematic chart.
Follow per composer — and per mood cluster.
One follow per artist surfaces every drop. A second follow on a mood cluster surfaces drops from every composer inside that cluster, so the cluster you like expands even when the artist changes.
Provenance visible on every track.
Model, prompt summary, and remix lineage are surface-level metadata on the now-playing panel — not a tooltip buried after the third menu. The track is what it is; we say so in plain language.
What you get on signup.
The first screen after signup is a queue built around how you already listen — not a streaming dashboard. Every tool we ship is something a working AI-music listener actually has to use.
A free listener profile in seconds.
Pick a handle, pick the moods you already like, and you have a profile the same minute. No queuing for an invite, no paid tier required to follow one composer.
Follows per mood cluster.
A cluster — say "moody ambient" — is a single follow, not a stack of subscriptions. New composers inside the cluster surface as they release, without you having to find them by hand.
Curated playlists shipped weekly.
Three to five hand-built crates per week, themed by mood and toolchain. Curators pick from a backlog of under-discovered cuts, not from the chart top, so the racks break new names.
An RSS feed of new releases from followed composers.
A real RSS endpoint per follow that lets you plug your music-reader into the catalog. The platform stands between you and the artist as little as the API allows.
No all-in-one mega-stack — just the moves a working AI-music listener actually has to make on a Tuesday afternoon.
- Mood / genre mapRoutes cluster to cluster.
- Follow graphPer composer and per cluster.
- Curated crateWeekly hand-built racks.
- Provenance panelModel + prompt summary.
- Lossless playerGapless with cue points.
- RSS feedOne endpoint per follow.
Same data as mainstream DJs. A different graph.
We don't bury AI tracks. We don't artificially bucket them. We read the same listen/save signals every DSP reads — and we route them through a mood/genre graph trained on diffusion and transformer output, so a moody ambient piece sits on the moody ambient chart, not buried under one big “AI Music” bucket.
- SIG.01
Completion rate
How far listeners stay through the track — the same signal every DSP reads.
- SIG.02
Queue depth
How long a cut keeps the session going — so the algorithm rewards tracks that hold the listener, not tracks that spike and drop.
- SIG.03
Mood adjacency
How well the cluster of this track lines up with the listener's recent queue — so the recommendation rings true before it rings loud.
- SIG.04
Novelty vs. cluster
How unique the cut is, so we surface a wide variety inside each mood instead of one hot producer filling the chart.
- 01liminal.fm ^ vol.4 — glassformsaurum.eth · moody · ambient
- 02halflight (mark II) — quiet roomnori.k · downtempo · late-night
- 03tape decay — study in dustplinth · lo-fi · dusty groove
We don't bring a moody ambient cut up by accident. We bring it up because the listen pattern — completion, queue depth, mood adjacency — put it there. The chart reflects the audience, not the platform.
Discovery driven by composer + mood overlap.
Not just “people who liked X also liked Y.” A mood/genre map that lets a mood cluster find your next composer before any follower count grows.
Follow per composer — every drop, every variant.
A single composer follow surfaces every release — stems, drafts, mark-II variants, the full stream — through a mood/genre lens that actually understands generative output.
Follow per mood cluster — expands your crate as it grows.
A mood-cluster follow pulls in every new composer publishing inside that cluster. You do not have to know the names in advance; the cluster is the discovery surface.
A daily digest of under-discovered cuts.
An opt-in daily surface that pulls the weekly edges of the catalog into one feed. It puts new creators in front of listeners who already care about the genre, not just about following a chart.
The follow graph is built from composer + mood adjacency, not from a single boolean tag. Followers cluster around a mood, not around a platform promise.
Need a starting rack? The week’s curated crate pulls five underdiscovered cuts from the catalog and ships them to your profile on signup — so day one is a queue, not a blank search box.
Get a listener profile, a mood-keyed discovery graph, and curated weekly crates. Cognison starts the moment you pick your first cluster.