139,255
brain neurons
The 30-second version
A whole animal nervous system, end to end, running on your GPU.
A real fly's brain was reconstructed neuron-by-neuron from electron microscopy — the FlyWire project. We loaded that exact wiring diagram into the browser and turn it into a living simulation: each neuron charges up, fires when it hits threshold, and passes the spike on to whatever it's connected to. The same rule, 139,255 times, every millisecond.
The brain talks to the fly's spinal cord (a second real connectome), which drives the legs of a physically simulated body. Press a key and you inject current into specific command neurons, then watch the consequences ripple all the way down to the feet. If the wiring says "this makes the fly walk forward," it walks forward. If it doesn't, it doesn't.
What's actually inside
Four real datasets, wired together.
01
Brain
FlyWire connectome — the full reconstructed brain of an adult fly: 139,255 neurons and ~15 million connections (Dorkenwald et al. 2024, Nature). Runs as a leaky integrate-and-fire network in a fused WebGPU kernel.
02
Spine
Janelia MANC connectome — the fly's ventral nerve cord, its "spinal cord": 23,188 neurons, 5.2 million connections. A second live network that turns brain commands into leg movement.
03
Body
TuragaLab flybody — a physically accurate fruit-fly model with 67 body parts and 111 muscles, simulated with real physics (MuJoCo) compiled to run in the browser.
04
Eyes
A tiny 64×16 retina rendered from the fly's own head position each frame, fed back into the brain's optic neurons — so the fly can actually see the target it's walking toward.
The brain and spine are joined the way real biology does it: a command
neuron named DNa01 in the brain is the same cell as
DNa01 in the spinal cord. No learned shortcuts — wire for
wire from the real animal.
Straight about it
What this is — and what it deliberately isn't.
What it is
- A whole-animal nervous system — brain, spine, and body — running end to end in a browser tab, no install.
- Built from real, published data: two connectomes, a physical body model, and a trained policy, each peer-reviewed.
- Reproducible by design — every shared run re-executes the same neurons deterministically.
What it isn't
- Not a replacement for scientific simulators (NEST, Brian2, NEURON) — those are faster, more detailed, validated.
- Not biophysically detailed: neurons are simple integrate-and-fire units — no ion channels or neuromodulation.
- Not quantitatively validated whole-brain, and it runs slower than real time. Full list in LIMITATIONS.md.
How to play
Race the fly to the target, on as few spikes as you can.
Each key fires a famous descending neuron — a command cell that carries instructions from brain to body. Tap them to steer:
Q W forward
E backward
R turn
A escape jump
S looming dodge
F backward (MDN)
SPACE new round
M science view
Reach the red target in the least time × fewest spikes. When you win, you get a replay link — whoever opens it watches the exact same simulation re-run: same target, your same keystrokes, against the same brain. You're sharing a brain trace, not a video.
How fast — honestly
Slower than real time. The only one behind a single URL.
The brain kernel is limited by memory speed, not math. We benchmarked it against the scientific standards on the same M2 Pro laptop — biological-time rate, higher is faster:
The original 1 kHz real-time target was unreachable for all three on that hardware — so we report what we measured, not what we hoped. The point was never "faster than NEST." It's a real fly brain on a phone in 30 seconds — the part the others can't do. Full disclosure in LIMITATIONS.md.
Why this one's different
Other groups simulate fly brains too. None of them open in a tab.
- It's just a link. Comparable projects need Python, a GPU, and a setup hour. This needs one click.
- All three layers, wired together — brain and spine and body — not just brain plus body.
- Every shared run is reproducible. A replay URL re-executes the real neurons deterministically — anyone can verify, study, or remix it.
Part of a research line
One front of a broader effort: Geant4-class simulators, ported to WebGPU, in a browser tab.
webgpu-fly is built by Ahmet Barış Günaydın, an independent researcher. It shares its thesis — and its kernel-fusion and viewer machinery — with these sibling projects:
Q
webgpu-q ↗
The closest sibling. Browser-native quantum chemistry — Hartree–Fock, CCSD(T), DFT, EOM-CCSD — ported from PySCF and GPU-accelerated. Same "real science in a tab" thesis, different physics.
🧬
webgpu-dna ↗
Sister project: Geant4-DNA radiobiology — Monte-Carlo electron tracks, radiolysis chemistry, DNA-damage scoring — in the browser, within 0.985× of the reference. The 4D-viewer pattern came from here.
⊗
kernelfusion.dev ↗
The research umbrella. Single-kernel fusion for GPU workloads — collapsing per-dispatch overhead — which is exactly what makes a 140k-neuron LIF step feasible per frame here.
▤
gpubench.dev ↗
"How fast is your GPU in the browser?" Real WebGPU compute tests across 592 devices and 7 vendors — the honest hardware baseline behind any number on this page.
◗
neuropulse.live ↗
Watch a real 3.8B-parameter transformer think, tensor by tensor — every glow is a live activation read back from WebGPU. The same read-back-and-render idea as this fly's retina.
A·B·G
barisgunaydin.com ↗
The author's site — an index of every project, paper, and note behind this research line, plus contact. Ahmet Barış Günaydın, independent researcher.
The honest part
This isn't the fastest fly-brain simulator — it's the most reachable one. It trades scientific completeness for a real connectome on a phone in 30 seconds. Treat it as a teaching, demonstration, and intuition-building tool — not a validated platform for publishing fly-brain dynamics.
The full, honest list of what's simplified, untested, or approximate — the performance ceiling, the LIF simplifications, the brain→spine→body approximations, the browser matrix — lives in LIMITATIONS.md.