Bayesian Threat Board?How this worksEvery tracked threat has a current probability that moves with evidence. Each dispatch we ingest is scored against every threat; if it raises or lowers the odds of that threat materializing, it nudges the probability via Bayesian update (prior + evidence → posterior).
Confidence intervals tighten as evidence count grows. Not forecasting — continuous re-weighting of a claim against all available reporting.
Confidence intervals tighten as evidence count grows. Not forecasting — continuous re-weighting of a claim against all available reporting.
Tracking nuclear threats.
Dimension index 83 · 2 threats · 13,002 pieces of evidence. Cross-sectional average of the member threats’ current probabilities.
2 threats shown
- 01North Korea provocationsnuclearNorth Korea· 5,722 evidence85CI 84–86%→ ±0.0pp7dElevated
Detail → - 02Iran nuclear escalationnuclearIran· 121,811 evidence80CI 80–80%↓ -9.6pp7dElevated
Detail →
Methodology
Each tracked threat begins with a prior probability based on historical base rates and expert elicitation. New evidence — every dispatch scored against the threat — nudges that probability up or down. Bayesian posterior updates compound; confidence intervals tighten as evidence accumulates.
Read the full approach at /about/methodology. Individual threat pages show the full evidence stream including direction, magnitude, and the reasoning our AI engine used for each nudge.