stonkfly-lab
Smaller mushroom-body trading lab with odor inputs, delayed dopamine, two-fly agreement, paper BTC, and a dashboard. Inspired by Stonkfly; not the full MaleCNS graph.
Read, and deliberately left unrated. The reasoning is below.
Propose a rating →The evidence
No evidence, no level.The author states the project does not use the MaleCNS graph and is a compact mushroom body only.
https://github.com/paappraiser/stonkfly-lab/blob/09e4529e2e1a135083838abd00ccacfd95c63a29/README.md#L18-L19All connectivity is generated from a seeded RNG rather than loaded from connectome data.
https://github.com/paappraiser/stonkfly-lab/blob/09e4529e2e1a135083838abd00ccacfd95c63a29/flylab/mb.py#L24-L42The three-factor learning rule updates these synthetic KC->MBON matrices.
https://github.com/paappraiser/stonkfly-lab/blob/09e4529e2e1a135083838abd00ccacfd95c63a29/flylab/mb.py#L65-L75
Off the scale: this project uses no connectome data at all, and says so. 'This is not the 166,700-neuron MaleCNS graph. Compact mushroom body only.' The mushroom body is generated from a seeded RNG — `rng.choice` picks each of 800 Kenyon cells' PN inputs and `rng.uniform` fills the PN->KC and KC->MBON matrices — and nothing in the repo loads a connectome, a FlyWire or MaleCNS file, or any neuron ids. It is a synthetic model of the mushroom-body motif, not a simulation of a measured circuit. Worth noting for the index rather than the scale: the plasticity itself is real work — eligibility-traced, delayed three-factor reinforcement on KC->MBON weights, with frozen and shuffled control arms — and the README is unusually candid ('Profitable trading has not been demonstrated'). If it were ever run on a connectome graph it would be an L2 shape; on a random expansion it is not on the scale.
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.