Y
Connectome · Interactive

How a thought moves
through a fly's brain

In 2024 and again in 2026, scientists finished mapping every neuron in a fruit fly's brain — the first complete wiring diagrams of a whole brain. This is a stylized walk through one of its most beautiful circuits: how a smell becomes a memory. Hover a neuron to learn what it does, then click it to fire a spike and watch the signal cascade downstream.

Speed
0 spikes fired
smell → relay → sparse code → decision
ORNALPNKCMBONLHDAN

Stylized depiction built on the real Drosophila connectome's olfactory-learning circuit (FlyWire, 2024 · Google Research/Janelia, 2026) — nota live simulation of the fly's brain. Node counts are small, representative stand-ins (the real mushroom body has ~2,000 Kenyon cells); the wiring pattern — divergence into a sparse Kenyon-cell code, then convergence onto ~34 output neurons — is faithful to the biology.

A smell arrives

Receptor neurons on the antenna catch odor molecules and converge, type-by-type, onto ~50 glomeruli in the antennal lobe. Projection neurons then carry each channel deeper into the brain.

It becomes a sparse code

In the mushroom body, ~2,000 Kenyon cells each sample ~6 random projection neurons and fire only on coincidence — turning every smell into a unique, sparse signature the fly can learn from.

A decision comes out

Just 34 output neurons pool that whole layer into approach or avoid. Dopamine neurons rewrite those synapses when there's a reward or a shock — that is a memory, physically stored.

What this is — and isn't

This is a stylized, honest prototype — not a live simulation of a fly's brain and not a raw dump of connectome data. The real olfactory circuit has thousands of Kenyon cells and tens of thousands of synapses; a browser can't render all of that at 60fps, so the node counts here are small, representative stand-ins. What is real is the architecture: convergence at the antennal lobe, massive random divergence into a sparse Kenyon-cell code, and convergence again onto a tiny set of output neurons. That shape is exactly what the connectomes revealed.

Sources & open data: FlyWire whole-brain connectome (Dorkenwald et al., Nature, 2024 — ~139,255 neurons) · nature.com · the complete male Drosophila CNS (Google Research + Janelia + Cambridge, Cell, 2026 — ~166,000 neurons) · research.google · explore the data yourself in FlyWire Codex and Neuroglancer.

I wrote about why I built this — and what it means that we can now map and begin to emulate whole brains — in this post.