Sign in to your workspace

Prodigii

Cybernetics Control System

Prodigii · Ecosystems of Intelligence™
Belief state Homeostatic
Perception sources
—
—
Evidence / sec
—
packets · gpil windows
Model evidencegpil belief fit — running model evidence: the mean probability each Behavioral Digital Twin assigned to what it actually observed, exp(EWMA(log P(observed | belief))), smoothed over time.

High → the twins predict the traffic they see (homeostatic). A sustained drop → the models are being surprised → distributional drift. It's the inverse of Bayesian surprise; ~60–70% is healthy across 5 regimes — watch the trend, not the absolute.

Live = fit on the current traffic. Learned = the trained twin models' fit over their accumulated learning window (1 day up to the 30-day retention max). Live well below learned → the models are being surprised (drift).
—
mean gpil belief fit
Control loop
PDA · closed
Z3 admitting actions
—
—
—

Connection

Where Prodigii reaches this source's alert/search API.

Require mutual TLSPin server + present client cert
Verify event signaturesReject unsigned or tampered evidence

Delivery

Push is preferred for the closed loop; poll is the degraded fallback.

Normalization · source → OCSF → HSML evidence

Field map applied at the ingestion gateway before the belief state updates.