The Predictive SI Runtime: Dashboard Variables, Output States, and the Falsification Ledger

This paper develops the runtime layer for Predictive Structural Intelligence. It translates the predictive architecture of breach hazard, buffer exhaustion, and the time-structure of answerability into an executable instrument for public use. The central claim is that predictive SI becomes dependable only when its forecasts are made inspectable through domain placement, observation lock, time horizon, state classification, dashboard variables, output states, downgrade rules, recovery verification, and a falsification ledger. The paper defines the Predictive Runtime Sheet; specifies a state sequence from Answerable, Compensating, Fluency-masked, Debt-financed, Theater-stabilized, Brittle, and Flickering through Crisis, Repairing, and Fragmenting; and introduces dashboard variables including drift velocity, contact half-life, contradiction recovery lag, operational variance, flicker count, repair velocity, binding strength, maintenance inflation, hidden-holder dependence, burden export velocity, maneuver saturation, hysteresis, synthetic trace risk, reader-capture risk, and active falsification timing. It also defines output states—Verified, Partial, Provisional, Blocked, Containment-first, and No predictive reading justified—and formalizes downgrade rules such as the Buffer-Dependency Downgrade Rule, Recovery Verification Rule, Synthetic Trace Downgrade Rule, and Active Falsification Timer. The contribution is practical rather than ontological: the paper gives Predictive SI refusal power by requiring forecasts to declare their evidence, anchor variables, falsifiers, review dates, and ledger obligations. It argues that prediction becomes answerable only when failed predictions leave a trace.