Mosca
FlyWire v783 subgraph (4,000 neurons on fluids, 500 on the platformer) as the mask of a trainable rate RNN (Dale signs from the data, weights learned inside the mask) that forecasts 2D Navier–Stokes and plays a procedural platformer and NES levels. Six controls trained identically (Erdős–Rényi, degree-preserving rewire, dense, no recurrence, MLP, FNO) and the analysis written as code before the data. On fluids the fly beats no recurrence (−4% rollout error) but loses to random (+2.2%) and rewired (+2.7%), 3 seeds each; on 200 unseen platform levels it loses to random (0.46 vs 0.57). One draw per random control; encoder, decoder and degree-picked ports are engineered. FlyWire data CC BY-NC 4.0.
Nobody has read this one closely enough to attach evidence to a level. Until they have, it stays unrated — a guess would be worse than a gap.
Propose a rating →The evidence
No evidence, no level.Nothing filed yet. The description above is carried over from townie/awesome-fruit-fly, which is a source review — not an independent reproduction, and not enough on its own to earn a level.
Record
Nearby
All projects →Research code for the Shiu et al. connectome-based leaky integrate-and-fire model, including activation/silencing experiments, notebooks, and FlyWire data configuration.
FlyWire whole-brain LIF implementation based on Shiu et al., with multiple simulation backends and benchmarking tools. This repository supplies the neural model; it is not a complete embodied demo package.
Ratings on thelearningfly.com are proposals, not verdicts. Every one of them is arguable in public.