Digital fly / Experiment 01

First taste.

A real neural signal. A simple visible consequence.

Recorded experiment

We ran the fruit fly’s reconstructed brain network on a Chromebook. Press Play to see its recorded mouth-motor activity move the fly’s mouthpart. Try Wires off to see why the connections matter.

1 brain-second → 8 seconds of replay
SugarReady to replay
An illustrated fly with a motor-driven mouthpartThe mouth angle follows the recorded spikes of two feeding motor neurons. This is a symbolic body, not a muscle simulation. Sugar and bitter colors identify the chosen input; they do not control mouth motion. Sugar input Mouthpart

Illustrated body · mouth movement follows measured model spikes

Each mark is an actual mouth-motor spikeTwo recorded motor cells

This page replays completed simulations. The buttons select real recorded trials; they do not start a new brain simulation.

The connection makes the difference.

The same model and input strength, tested three times per condition. Bars show recorded mouth-motor spikes; dots show individual trials.

What is real here—and what did we draw?

Real model: the full original FlyWire v630 network, simulated locally using the published neural equations. Each mouth movement comes from its recorded motor spikes. There is no “if sugar, move mouth” rule.

Our drawing: a simplified fly. The same fixed conversion turns motor spikes into a visible angle for every condition. The mouth is a signal display, not a calibrated muscle. The colored drop only labels the selected input; this experiment does not simulate physical contact, swallowing or hunger.

The control: Wires off stimulates the sugar cells but sets all connection weights to zero. The sensory spike times were identical to the paired sugar run, but the mouth-motor signal disappeared. This first control does not tell us whether some other wiring could produce a similar result.

The limit: this reproduces a small response in a simplified, calibrated model. It does not recreate a living fly or demonstrate consciousness. Three trials are a first proof of concept, not a full replication of the publication.

Results, sources & how to reproduce it

The parameters and original neuron IDs were fixed before these trials. Sugar and bitter cells were each driven at 100 Hz when selected. No graph trimming, response training or parameter search was used.

Shiu et al., Nature (2024) · Original model and MIT-licensed code · FlyWire data attribution and terms

Download recorded results · JSONDownload experiment guide · MarkdownDownload experiment runner · PythonOpen source and reproduction instructions