How this works
Every adult in the dish runs a real nervous system live in this tab. Females run the FlyWire brain (139,248 neurons, connections with 3+ synapses, Shiu et al. 2024 model). Males run the Janelia male whole central nervous system, brain and ventral nerve cord together (165,114 neurons, 7.9 million connections with 4+ synapses; each cell type's excitatory and inhibitory input is calibrated to what the same type receives in FlyWire, so the circuits verified there behave the same way here). Senses drive real sensory neurons every frame: each fly's own motion becomes optic flow on T4/T5, anything approaching becomes expansion on LPLC2/LC4 of the facing eye, sugar smells through DM2 receptor neurons and tastes through the sugar GRNs, walls and flies touch the antennal bristles, gusts and airflow hit the JO-A wind neurons, song hits the JO-B neurons, and the male's front-leg taste neurons (ppk23) meet the female's cuticle, and dust settling on the antennae drives the antennal bristle neurons, which in both wirings recruit the grooming command neurons DNg15/DNg84 until the front legs have wiped it off (Hampel et al. 2015).
What the male's nerve cord does by itself: the giant fibre fires the jump muscle's motor neurons (TTMn), then the wing power motor neurons (DLMn, DVMn), and the wing's own sensors feed those back so flight sustains itself until the landing neurons fire; antennal touch recruits the front-leg motor pools (grooming); P1, the courtship command, drives pIP10 and the song motor neurons ps1/hg1 on one wing and the abdominal motor neurons that curl the abdomen. His leg motor pools shape the stance while walking.
What no connectome supplies, and how it is filled: inner state (hunger, startle, meals, activity bouts) drives the DNp09 walking command; a goal direction goes through the compass output PFL3 (which steers via DNa02 in both wirings), set by a stand-in that drifts, holds while a smell rises, casts when it fades, surges upwind in a breeze and follows walls; the legs' stepping rhythm is a pattern generator sized by the walking command, because the point-neuron model has no rhythm-making dynamics; P1 is switched on by the tap itself (its input from the leg taste neurons is too weak in the model) and stays on for tens of seconds as it does in the fly; on the female's side, song and cVA reach her sensory neurons but no path to pC1 survives in the FlyWire model at any pruning, so song drives pC1 directly, and pC1 in turn sets receptivity, the vaginal-plate neurons and, after mating, the oviDN egg-laying neurons; copulation lasts a set time (the abdominal-ganglion timer is neuromodulatory).
Why they fly: an air puff on the antennae is the classic flight trigger (JO-A reaches the giant fibre in the female wiring), so the breeze, with its gusts, launches her; looming launches both. In a breeze the sugar smell streams downwind as a plume; a hungry fly that meets it surges upwind and casts across the wind when it loses it, and a flying fly lands when the smell is strong. Flight is a point mass under wing thrust, gravity and drag against the moving air.
Mating: seeing her drives LC10a on one eye and the same-side DNa02, so he turns toward her and follows; his taps on her cuticle switch P1 on; song follows through the wiring; her receptivity builds; when he is behind her, she has paused with the vaginal plate open and his abdomen curls under, they couple, and the pair moves as one body, he going where she goes, until they separate. Afterwards her post-mating switch turns off receptivity and turns on egg laying on sugar. Runaway guard: under a long courtship the male's network can tip into a self-sustaining reverberation, a model artefact with no way back in a leaky integrate-and-fire net; his bristle, tap and pursuit drives are capped below the tipping point, and if it happens anyway the guard ends it as a seizure ends, every cell reset to rest, and the caption says so. Lifecycle: a dish day is 30 seconds of brain time. An egg hatches after a day (about a third never do), the larva crawls to the sugar, eats and grows for four days, pupates for four, and the fly that ecloses is a new adult with its own whole nervous system, female or male by coin toss, unable to fly for its first hours while the wings harden; adults age, die of old age after about 45 days or of starvation after a day and a half without sugar, and the body lies in the dish for a while. Larvae and pupae are simple crawlers with no connectome (the larval brain is a different animal); the room for adults is a slider (four by default, up to eight), because every adult is a running nervous system and they share one frame budget, so each extra one slows the dish; pupae wait for a place. A bead also dries out slowly, so the dregs of an eaten drop vanish rather than lingering. Auto sugar drops a fresh bead when the dish runs out. Body: NeuroMechFly v2 (flygym, Apache-2.0). Data: FlyWire v783 and the Janelia male CNS v0.9 (public release bucket). Compared with Eon Systems' embodied fly (the same FlyWire brain and NeuroMechFly body, with MuJoCo physics and imitation-learned walking controllers under seven descending neurons, synchronised every 15 ms): this page maps the same commands by hand, adds the male's real nerve cord, a full-window optic flow, looming and other-fly vision computed from the scene, taste, smell, wind, song and dust, and runs the brain-body loop every frame, about 16 ms.