We Can Map a Whole Brain Now. I Built a Way to Watch One Think.

I have spent twenty years building things that move information around — telecom networks, AI platforms, little companies scattered across a dozen countries. So when I read, in the same year, that we now have a complete map of every neuron in a brain and that AI models can generate whole playable worlds from a text prompt, I couldn't let it go. Two different fields, quietly crossing the same line at the same time: we can now map intelligence, and we can manufacture something that behaves like it.

I wanted to understand the first one with my hands, not just my feed. So I built the interactive thing above. Fire a neuron and watch a signal cascade through a fruit fly's brain.

Let me tell you what's real about it, what isn't, and why it kept me up.

Someone mapped an entire brain. Twice.

For most of history, a brain was a black box. You could poke it, scan it, watch it light up — but you could not hold the actual wiring diagram, the who-connects-to-whom of every single cell. That is what a connectome is, and until very recently we didn't have one for any complete brain.

Then, in October 2024, the FlyWire consortium at Princeton published the whole-brain connectome of an adult female fruit fly in Nature: 139,255 neurons and around 50 million connections, reconstructed from electron-microscope images and proofread by a global community of scientists. Every neuron. The complete graph.

Then they did it again. In September 2026, Google Research, HHMI Janelia, and the University of Cambridge published the complete male fruit-fly central nervous system in Cell — about 166,000 neurons and 125 million connections, brain and nerve cord together. It's the largest brain map by neuron count to date, and because it's the male, scientists can now line it up against the female map and ask what's different, cell by cell.

A fruit fly is not a human. But this is the first time we've had the complete circuit diagram of any brain — and the tools they built to make sense of it are open to anyone.

The part that surprised me: you can just look

I assumed this data lived behind a lab door. It doesn't. Google's Neuroglancer — the WebGL viewer they built to fly through terabytes of brain imagery — is open source. And FlyWire's Codex lets you search neurons, trace circuits, and download the connection tables yourself. The raw data is too big to load in a browser (the full connection table is hundreds of megabytes; the image volumes are terabytes), but you can pull a single circuit as a small edge list and do something with it.

So I did.

What you're actually looking at

The visualization above traces one of the most studied circuits in the fly: how a smell becomes a memory. It's the olfactory pathway into the mushroom body, the fly's learning center. The flow runs left to right:

Receptor neurons on the antenna catch odor molecules (~50 receptor types).
They converge onto glomeruli in the antennal lobe — one knot per smell channel.
Projection neurons (~150 of them) relay each channel deeper in.
In the mushroom body, ~2,000 Kenyon cells each sample about six random projection neurons and only fire when several fire at once. This is the clever bit: it turns every smell into a sparse, almost unique code.
Just 34 output neurons read that entire layer and boil it down to approach or avoid.
And dopamine neurons rewrite those last synapses when there's a reward or a shock. That rewrite is the memory.

I want to be honest about what this demo is, because the whole point of the exercise was to not fool myself. It is a stylized prototype, not a simulation of a living brain, and not a raw dump of the connectome. The real mushroom body has thousands of Kenyon cells and tens of thousands of synapses — a browser can't animate all of that smoothly, so the neuron counts here are small, representative stand-ins. What's faithful is the shape of the wiring: convergence, then a massive random fan-out into a sparse code, then convergence again onto a tiny decision layer. That architecture is exactly what the connectomes revealed, and it's genuinely beautiful once you can watch a signal move through it.

Map, then emulate

Here's where it stops being a fun toy.

A wiring diagram is static. It tells you who connects to whom, not what the brain does. The obvious next question is: if you have the complete wiring, can you run it? In 2024, researchers did exactly that — they built a connectome-constrained simulation of the entire fly brain, a simple model where each neuron is a leaky little integrator wired up according to the real connectome. When they switched on the neurons that sense sugar, the model correctly predicted which downstream neurons would respond and which motor neurons would drive feeding — and then they confirmed it in a real fly.

Read that again. A wiring diagram alone, with crude assumptions about the rest, predicted real behavior in a living animal.

It is not a mind in a computer. It doesn't capture the full chemistry, the neuromodulators, the plasticity, a hundred things we're still guessing at. Nobody has "uploaded" a fly. But the direction is unmistakable: map the circuit, then emulate the circuit, then check it against reality.

The other line we crossed

Now hold that next to what's happening in AI. In the same window of time, generative models stopped producing just text and images and started producing worlds — coherent, explorable, physically-plausible environments you can walk through, generated on the fly from a prompt. "World models," people are calling them. We went from AI that describes a scene to AI that hosts one.

So here we are, standing between two mirrors. On one side, we're learning to read biological intelligence down to the last wire and press play on it. On the other, we're learning to conjure synthetic worlds and the agents inside them out of nothing but data. Reading brains and writing worlds, both maturing at once.

I'm an optimist about tools — I build them for a living. But I've also learned that the exciting version and the dangerous version of a technology are usually the same version, just pointed differently. The same map that helps cure a disease is a blueprint for control. The same model that generates a training world can generate a convincing fake one. The honest response isn't hype and it isn't panic. It's paying attention early, while a fruit fly is still the headline and not the afterthought.

That's really why I built the demo. Not to claim I understand a brain — I don't. But there's a difference between reading "166,000 neurons" and actually clicking one and watching the signal ripple outward. The number is a fact. The ripple is a feeling. And I think we're going to need the feeling to make good decisions about the facts.

Open the full interactive version → — hover any neuron to learn what it does, then fire it and watch the cascade.

Real data and open tools used: the FlyWire whole-brain connectome (Dorkenwald et al., Nature, 2024), the complete male Drosophila CNS (Google Research / Janelia / Cambridge, Cell, 2026), FlyWire Codex, Neuroglancer, and the connectome-constrained fly-brain simulation (Nature, 2024). The visualization is my own stylized rendering of the olfactory–mushroom-body circuit — faithful in structure, not a literal simulation.