Predictability
Dashboard
The first commercial AI trust gauge. For every connected data stream, the dashboard computes the live Lyapunov exponent λ, the effective decision horizon k*, and the current chaos regime — then surfaces drift alerts the moment conditions change.
Requests for horizons k ≤ 14 are allowed. Beyond that, Rebound will route to a safer model or block the call entirely.
k* shifted from 22 → 9. Regime → Chaotic. Blocking horizons > 9 steps.
Drift detected. λ rising 3σ above 24h baseline. Human checkpoint recommended.
What the dashboard tells you
that nothing else can.
Live λ estimation on streaming data
Online Lyapunov exponent on 1-second to 1-hour windows. No batch retraining. Updates as market regime, weather, or task chain shifts.
k* derivation, not guessing
We translate λ into a hard, interpretable forecast horizon — the step beyond which the model's output is statistically indistinguishable from noise.
Regime classification
Every stream is continuously labeled Predictable, Transitional, or Chaotic. Dashboards, webhooks, and the selector all consume the same signal.
Drift-aware alerts
We alert when λ rises 3σ above baseline or when k* compresses materially. Webhook payloads are structured for Slack, PagerDuty, or Opsgenie.
See your own λ and k*.
Bring a dataset. We return your Predictability Audit in 10 business days.