Divergence
Where reference assumptions stop agreeing
Two views of the same synthetic harness: a map of how far the reference models drift apart as the horizon extends, and the recommendation that drift produces at each step.
Demo modelSynthetic dataResearch preview
Widest divergence
—
computing
Action threshold
0.35
above this, admission control stops treating the state as routine
Cells above threshold
0
of 0 domain × horizon cells
Model-disagreement heatmap
Rows are domains, columns are how far ahead the scenario clock has run. Darker means the reference models disagree more about the same world.
computing reference disagreement…
Average divergence by domain
Across the whole horizon. The dashed mark is the action threshold, not a coin flip — anything to its right means the references stopped agreeing on average.
Earth Observation
0.00
Financial Services
0.00
Critical Infrastructure
0.00
action threshold (0.35)
click a row to see the variables behind it
Regime timeline — how the recommendation changes with horizon
Admission control re-decides at every step. This is the sequence of those decisions and the evidence behind each shift.
computing admission decisions…