Fruit fly utopia
MaleCNS soma cloud plus a reduced PAM / PPL1 / PER rate model. Worlds include sugar, cocaine, doomscroll, and a printed “utopia” welfare mix that is not “max PAM”; cocaine loses. Wanting / reinforcement, not a trained policy.
The connectome is used as fixed wiring. Activity is mapped straight to controls. Nothing changes over time.
Dispute this rating →The evidence
No evidence, no level.The entire simulation is a hand-tuned scalar rate ODE over ~18 named populations; every coefficient is a literal in the source and no weight is ever written.
https://github.com/ML-Chen/fruit-fly-utopia/blob/30eb4bcaed9bd2c38dfc928cab72360d6266b8b4/src/lib/brain/simulate.ts#L255-L282The only connectome data loaded is soma position (3 floats) and a group id (1 byte) per cell — no connectivity is parsed.
https://github.com/ML-Chen/fruit-fly-utopia/blob/30eb4bcaed9bd2c38dfc928cab72360d6266b8b4/src/lib/brain/load-somata.ts#L25-L32Neurons are aggregated into 18 named groups, which are the units the rate model actually simulates.
https://github.com/ML-Chen/fruit-fly-utopia/blob/30eb4bcaed9bd2c38dfc928cab72360d6266b8b4/src/lib/brain/groups.ts#L1-L20
The MaleCNS data this project loads is soma XYZ positions plus a one-byte group label per cell — anatomy for a 3D point cloud. There is no connectivity, no adjacency matrix and no synapses in the shipped data, and nothing in the repo reads any. Not L1: there is no neuron model and no per-neuron state. The whole 'brain' is an 18-variable scalar population ODE with hand-written literature-derived coefficients, low-pass filtered by a `leaky()` helper; the resulting 18 rates only tint the point cloud. Membrane voltage, spikes and per-cell dynamics are absent. The README is honest about this ('The dynamics are a reduced rate model. Soma positions are real. The Hz numbers are model-equivalent, not patch-clamp'), and the welfare score is explicitly labelled a human-chosen objective rather than anything the fly computes.
Record
Nearby
All projects →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.
Research framework comparing a MaleCNS-constrained sparse recurrent controller with rewired graphs, conventional neural networks, and other controls on simple VizDoom tasks.
Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.