PAPER / 01 OF 13 · July 2026
Radiance-Luminance Theory and Alpha Wolf Eyes
Atom McCree
THE IDEA / PLAIN LANGUAGE
What it says.
Most computer vision learns by studying huge piles of pictures. Alpha Wolf Eyes tries another way: measure light, color, edges, shape, and motion with fixed, inspectable rules—more like a built eye than a trained black box—then recognize what comes back.
THE ARGUMENT / TECHNICAL
Abstract.
Introduces Radiance-Luminance Theory (RLT), a photon-first account of visual perception, and Alpha Wolf Eyes 3 (AWE-3), an implemented deterministic retina-to-inferotemporal pipeline with zero learned filter weights. The system integrates linear-light recovery, chromatic adaptation, foveated log-polar canonicalization, retinal and LGN-inspired channel decomposition, fixed V1/V2/V4 measurements, and an 80-dimensional identity representation. On its repeatedly evaluated 47-class development corpus, AWE-3 classified 275 of 282 observations (97.52%); the manuscript correctly frames this as development evidence, not an unbiased generalization result, and specifies a locked million-observation scale experiment.
KEYWORDS / FIND THE THREAD