DesktopFly for Linux
Python/GTK port of DesktopFly for GNOME on X11 and Wayland, retaining the upstream FlyWire circuit and documenting platform limitations.
Neurons are simulated — leaky integrate-and-fire or better — with documented, tuned parameters.
Dispute this rating →The evidence
No evidence, no level.The neuron model is an explicit LIF with documented membrane tau, threshold, refractory period and inhibition delay.
https://github.com/somsom10/desktop-fly-linux/blob/930ece09a885690e09b0e3604430de24c9221845/desktopfly/simcore/lif.py#L31-L48Signed synaptic weights are computed once from the connectome edge list at construction and stored in a CSR array; nothing later writes to it.
https://github.com/somsom10/desktop-fly-linux/blob/930ece09a885690e09b0e3604430de24c9221845/desktopfly/simcore/lif.py#L101-L119The authors state that the sugar pathway is odour bearing injected as current onto real steering and walking neurons - stimulation, not a change in synapses.
https://github.com/somsom10/desktop-fly-linux/blob/930ece09a885690e09b0e3604430de24c9221845/README.md#L110-L120
A 668-neuron FlyWire escape/steering circuit runs as a leaky-integrate-and-fire network at 1 kHz with stated parameters (20 ms membrane tau, threshold 1.0, 2 ms refractory, 4 ms delayed inhibition), so membrane voltages are genuinely simulated. Not L2: the signed synaptic weights are built once in the constructor from the connectome edge list and the only other reference to that array in the whole repo is a read inside the propagation loop - no line anywhere writes a weight. The sugar feature the README added is injected current onto the real DNa01/DNa02/DNp09 cells, which the authors correctly label as stimulation, not plasticity.
Record
Nearby
All projects →Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.