Sustained γ + σ divergence detector.
Zeta sits on the Gamma (capability) and Sigma (behaviour) signal space and forecasts where a project's state is heading. It flags sustained divergence — anti-spike, not single-event noise. Alpha-stage: forecast accuracy is measured on a codebase's own history, not external prediction. For security and platform teams: an early-warning signal that a module or dependency is drifting toward a sabotage or backdoor shape — before it lands in a release.
Trajectory forecast measured against a codebase's own subsequent history (in-sample). This is an alpha-stage signal; the mechanism stays sealed.
Zeta classifies divergence that persists across the trajectory, filtering single-event noise. The forward forecast reproduces the codebase's own next states with cosine ≈ 0.99999 at H=1, holding real-time speed across many forward horizons.
Forecast accuracy is measured on the codebase's own subsequent history — not external prediction of unseen systems. It is not an AGI-alignment claim. Out-of-sample generalisation and the mechanism are disclosed under NDA.