Agnostic Learner

⏺ Supplementary code for "Evolving Observers for Agnostic Pattern Discovery in Spatiotemporal Fields." Contains the complete implementation of the observer model (16-gene genome with evolvable trace-axes bitmask and deterministic entity selection), the two-phase fitness function (spectral discovery f_s and predictive validation f_p), the ecology manager (energy economy, CV-based mode switching, population dynamics), a generalised outer-totalistic CA environment supporting arbitrary B/S rules, noise environments N1--N5, a Configuration Graph algorithm for ground-truth pattern detection, a GPU-accelerated FFT validator (WebGPU/WGSL), and experiment runners for the controlled scenarios S1--S5 and the 8-rule CA comparison. Python 3.10+, NumPy, SciPy.