Fly Brain Sim

Welcome to
Fly Brain Sim

A fruit fly whose brain is the real scanned connectome. Feed her, scare her, teach her, and watch 138,639 neurons decide what she does.

Starting…

138,639neurons simulated

54.5 Msynapses

8,846neuron types

1,303descending neurons

How it works

🧠

The real wiring

Every neuron and every synapse count comes from the FlyWire scan of an adult female fruit fly. Senses go into real sensory neurons and her actions are read from real descending neurons.

🪰

Teach her

Place a lamp next to food or a flame. Dopamine neurons weaken the real mushroom-body synapses, and she learns to approach or avoid it.

🥚

A whole life

Mating, egg laying, larvae and new flies, with a reproductive tract driven by her real mating and egg-laying neurons.

Quickstart

01Pick a thing from the dock on the left and click the arena to place it.

02Put food down and watch the smell neurons pull her toward it.

03Add a lamp beside food a few times. Later, she goes to the lamp alone.

04Put the lamp beside a flame instead and she learns to keep away.

05Drop a male fly, then glass and walls, and see mating and egg laying.

06Open Saves to keep a world, or export it as a file.

What is real, and what isn't

Time scale. One fly-day lasts 2 minutes at life cycle ×1. Anything that takes minutes or longer in a real fly (starvation, egg development, ageing, memory fading) runs on that clock. Walking, reflexes and neurons run in real time. Courtship and copulation are shortened so you can watch them.

✅

Real: from the scan

• 138,639 neurons and 54.5 million synapses, and every dot at its scanned position.
• Smell, taste, touch, heat, hearing and looming go into real sensory neurons.
• Her actions (approach, feeding, alarm, escape) are read from real descending neurons.
• Mating uses the real pC1 hub and the vaginal-plate neuron DNp13. Egg laying uses the real egg-laying neurons.
• Learning weakens real Kenyon cell → MBON synapses, gated by real dopamine neurons.

🛠

Approximated or hand-made

• Steering and legs: a hand-made reflex, because the scan has no nerve cord.
• Wandering: a random wobble I wrote. The brain is silent with no input, while real brains are always spontaneously active. She walks at one constant speed, never pauses, grooms or follows walls, and hunger doesn't drive exploring.
• Training: real mushroom-body mechanism, but one cue class, so lamp colour is cosmetic and she can't learn a specific colour. The lamp enters through olfactory-style inputs, not vision. There is no short/long-term memory split, extinction or spaced-trial effect. Eating is a fixed reward switch and the approach is a turning rule.
• Egg laying: one input neuron (SMP550) is driven directly.
• Body: ovaries, sperm, development and ageing are my own models with real timings.
• Brain settings: synaptic gain is 0.34 of the published value, 1 ms steps, about 0.65× real time.

⚠️

Known unrealistic or missing

• A food blob is eaten in about 3 seconds. A real meal takes minutes.
• Odour is a smooth gradient, not turbulent plumes.
• A flat 2D arena: no flight, climbing, sleep, grooming, day and night, or temperature.
• Newly emerged females only wander. They never mate or lay.
• Colour vision, specific colour learning and real visual steering are not modelled.