Does the fly
actually learn?
Everything people have wired the fruit fly connectome into — Doom, Mario, robot bodies, lab rigs — rated by how much of the biology is actually simulated. A moving fly is not a learning fly.
The fidelity meter
Every project gets one level. Evidence attached.The connectome is used as fixed wiring. Activity is mapped straight to controls. Nothing changes over time.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Synapses change under a modeled rule grounded in fly biology — for example dopamine-gated change on Kenyon-cell → MBON synapses in the mushroom body. Learning begins here.
Behavior is compared against real fly experiments, and the comparison is published.
Rated so far
All 141 →MaleCNS simulation connected to ViZDoom through modeled visual inputs and a fixed button readout. Includes plasticity experiments and negative validation results; learned survival has not been demonstrated.
FlyWire mushroom-body LIF sat at Texas Hold’em. Innate wiring can discriminate hands; outcome-driven dopamine did not teach cards. Honest negative result.
MaleCNS mushroom body (KC→MBON, PAM / PPL1) learns Tetris as odor-gradient navigation with the fly’s own dopamine-gated plasticity rule.
FlyWire sub-circuits as spiking networks that play games using the fly’s own dopamine update rule.
8 ratings proposed · 133 projects indexed and waiting · every rating below reviewed status is marked [draft]
Know one we missed?
Disagree with a rating?
Submissions and disputes are public. Every level ships with the evidence that earned it.