SPIKING NETWORK · FRUIT FLY BRAIN

Fly DJ

A small synthetic fly brain DJs an endless generated set, from warehouse techno to lo-fi, mixing track into track on two decks. The tracks lay out the grid; the neurons decide which hits land, how hard, and which notes. You can steer the mood, rewrite patterns and ride the mixer while it plays.

now … 124 BPM energy 50% set #1 fly #4711

Decks

Every track is generated on the spot from ten genre templates, so the set never repeats. When a track reaches its outro the fly mixes in the next one, choosing the technique a DJ would: a long stem-by-stem blend for similar tracks, a build and drop when energy jumps, a breakdown bridge or an echo when tempo or style changes a lot.

Mood
Blend length

Played

    Live

    Click a step to add or remove a hit in the part the lead deck is playing; the brain still decides whether it lands. Click an instrument name to mute it, drag its slider for level. Double-click any slider to reset it.

    Left closes a low-pass, right sweeps a high-pass up.
    Pushes the set's tempo up or down.
    More reverb and echo on everything.
    spike raster, last 5 svertical axis: neurons, grouped by module

    Brain at the decks

    Each module plays its own instruments. Mute silences its neurons, Boost drives them with a steady stream of spikes, like an optogenetics experiment. Activity is measured against each module's recent average, so a boosted module gets wild for a few bars and then settles.

    Senses and volume

    How loud the room and the fly's own output reach the antennae.
    Rotation of the view. The bump on the central complex ring picks the lead notes.
    Spontaneous firing everywhere. More ghost notes and fills.
    Rewires the brain from a new seed. Same set, different player.

    The network is small and synthetic: 179 leaky integrate-and-fire neurons whose wiring follows fruit-fly motifs (sparse Kenyon cell input with APL inhibition in the mushroom bodies, a ring attractor in the central complex) but contains no connectome data. The next step is to drive the same decks from the full FlyWire model (about 138k neurons) by Shiu et al.: github.com/philshiu/Drosophila_brain_model.