flymaxxing.
Back to the collection
Tools · BY PixelML

Firefly: wildfire damage

Browser version not available yet

The arena depends on a CUDA neural controller. Its documented endpoint is a private-network address, not a public browser service. The CUDA kernel has not been ported to WebGPU.

Inspect the original source below ↓
Tools / Source filesFirefly: wildfire damageMaleCNS
Project reference card · Flymaxxing
Tools / Code available

About Firefly: wildfire damage

Satellite tiles go through a frozen fly circuit, then a small trained decoder predicts damaged or undamaged structures.

MaleCNSSatellite imageryClassificationNegative result

What’s actually happening

A benchmark with rewiring and pixel-only controls. The committed results report 0.588 balanced accuracy for the fly circuit versus 0.655 for pooled pixels, with no detected advantage of the specific biological wiring. It is an experimental image-classification study, not an operational wildfire warning system.

Sources reviewed September 16, 2026. Browser operation checks do not validate scientific claims.
Inspect the source without leaving Flymaxxing

Pinned source revision · loaded from the original public repository

Choose a file to read its source.