A fly's brain, switched on.
In 2024 a research team published a computer model of a fruit fly's whole brain, built from the map of its wiring. We rebuilt that model so it runs in a browser, checked it against the authors' own code, and put it online.
Open the labFly Brain Lab
The model running live. Switch on taste, antenna or hearing neurons and watch what the fly does.
The explainer
Not published yet.
What the Fly's Wiring Map Predicts Without a Simulation
Our own analysis: how much of the model's feeding prediction is already in the wiring map, and what public re-runs of the model report. Preprint, version 1.6, 11 October 2026; not peer-reviewed.
Experiment and paper code
Plans, code and result files for the paper: github.com/terramind-org/science. The lab's own code is not public yet. The model we ported is the authors' open code: github.com/philshiu/Drosophila_brain_model.
- We ported the authors' model to JavaScript so it could run in a browser.
- The first version was wrong. On the paper's sugar experiment it gave 81.8 spikes a second where the authors' code, in our run on the map version the paper used, gives 67.5. A spike-by-spike test on a small network found the cause: the original drops input that arrives while a neuron is recovering from a spike, and ours had kept it.
- After the fix the two match. They are identical spike for spike on that test network (300 artificial neurons), 5,053 of 5,053, and the sugar experiment gives 68.2 against 67.5.
- We found we were not first. People had already run fly brain simulations in a browser, including fly-explorer by Brandon Cho and webgpu-fly by Ahmet Barış Günaydın.
- An independent check found our mistakes before launch. A separate AI reviewer that had not written the page went through every sentence on the lab page and found 3 errors and 16 places that needed a caveat. We corrected all 3 errors and reworded most of the 16. A few minor ones are still open. A second pass found that we had misread the paper's tables, taking the spread of the numbers for the numbers themselves. On the three tables we checked, the paper's firing rates are within 3 spikes a second of ours.
The model. Shiu, P.K. et al. (2024). A Drosophila computational brain model reveals sensorimotor processing. Nature 634, 210–219. https://doi.org/10.1038/s41586-024-07763-9
The wiring map. Dorkenwald, S. et al. (2024). Neuronal wiring diagram of an adult brain. Nature 634, 124–138. https://doi.org/10.1038/s41586-024-07558-y. FlyWire releases its data under CC BY-NC 4.0 (https://flywire.ai/guidelines).
Cell types and neuron positions. Schlegel, P. et al. (2024). Whole-brain annotation and multi-connectome cell typing of Drosophila. Nature 634, 139–152. https://doi.org/10.1038/s41586-024-07686-5
Courtship circuit. Wang, K. et al. (2021). Neural circuit mechanisms of sexual receptivity in Drosophila females. Nature 589, 577–581. https://doi.org/10.1038/s41586-020-2972-7
The lab's own "Sources and credits" panel has the full list.
No corrections since the lab went live on 10 October 2026. Later that day the lab moved from /fly-brain/ to /fly-brain/lab/; its text and the model were not changed. Any correction we make will be listed here and stay here.
Our port of the model is not peer-reviewed.