PROJECT SUBSTRATE / CONTROLLED REPLICATION
Freeze the capture code, hardware profile, sampling window, z-score implementation, thresholds, XOR keys, packet widths, render shapes, and success criteria before reviewing new data.
Collect interleaved conditions: idle baseline, controlled CPU load, thermal ramp, storage activity, network activity, scheduler stress, and synthetic streams matched to the observed distribution and autocorrelation.
Run the exact same transform inventory blindly across every condition. Report every attempted key and layout, not only visually striking outputs.
Define success numerically before capture: out-of-sample classification, effect size, false-positive rate after multiple-comparison correction, and replication on a second machine.
External authorship would require more than anomaly: repeatable information content not explained by the host and decoded by a preregistered method.