A cage of flies,
each driven by a complete
nervous system wiring diagram.

What the connectome gives, and what it does not.

The wiring is real. The dynamics are ours. Every claim on this page is meant to survive that sentence.

Berg and colleagues at Janelia reconstructed the complete central nervous system of one adult male fly from electron microscopy: the brain, both optic lobes and the ventral nerve cord, through an intact neck. That dataset, MaleCNS v1.0, is the nervous system of every animal in this cage: every neuron of the MaleCNS connectome, and every connection of five or more synapses between neurons proofread at both ends. The engine loads its — nodes and — weighted edges verbatim from the canonical file below, whose Merkle root is quoted in section 6. Connections of fewer than five synapses are not loaded; the file is not the whole wiring, and this page never calls it that.¹

The dataset does not give neuron dynamics, muscles, or a body. So the engine supplies the same integer threshold unit for every neuron (a potential, a refractory counter, a fired flag), assigns — roles to — of the neurons from Janelia's own annotations (which sensory neurons are olfactory, gustatory, mechanosensory, photoreceptors; which motor neurons drive which leg, the wing power and steering muscles, the halteres, the neck, the proboscis; which neurons descend from the brain to the cord, which ascend, which are neurosecretory), and treats the rest of the body as a few equations: legs, wings, halteres, proboscis. A fly here is driven by the real connectome. It is not a real fly, and it does not behave exactly like one.²

What the wiring does do is measurable, and it is measured, not asserted. The figure below is filled from probe runs over the checked-in engine; each number carries the horizon it was measured over.³

— neurons, watched by role.

Too many cells for one dot each on a screen, so the brain is shown two ways. The bars are the neurons that fired at the world's last tick, by role group, as a share of the group. The raster below them samples every —th neuron in canonical order, one dot per sampled neuron, drawn when it fired; it is read from the world four times a second for the fly named in the caption. Click a fly in figure 1 to watch that one.

A glass box with weather.

The floor is —×— cells; above it are — layers of air for flight, and four walls and a lid a fly can walk on. On the floor stand raised dishes of yeast paste and banana, a water pool, and a salt crust in the driest corner; a lamp hangs from the lid and lights the cage in a cone. Light follows a day of 16,384 ticks; temperature drifts around the cage-centre value the thermosensory neurons compare against, with a gradient across the floor, hot on the lamp's side; humidity falls from the pool to the salt corner. All of it is a pure function of the tick and the seed, so it is the same in every mirror.

What kills a fly.

Where the lamports go.

Money lives as lamports on the world account, a program-derived address on Solana. Four pools are tracked inside it: each fly's vault, the metabolism that sets how many animals the cage supports, the pool that pays living animals, and credits owed to people. The program keeps one invariant on every instruction: the account's balance above rent covers all four.

Whose work this stands on.

Berg, S., Beckett, I.R., Costa, M., Schlegel, P., Januszewski, M., Marin, E.C., Nern, A., Preibisch, S., Qiu, W., Takemura, S.-y., and the FlyEM and Jefferis groups (2026). Sexual dimorphism in the complete connectome of the Drosophila male central nervous system. Cell; preprint doi:10.1101/2025.10.09.680999. Data: Janelia FlyEM, MaleCNS v1.0 (janelia.org), used under CC BY 4.0. The census file used here is a conversion of the published connection-weight and body-annotation tables, thresholded at five synapses and restricted to neurons proofread at both ends; it is not a redistribution of the originals, and the changes are stated in the file's provenance.