Predictive history, run as a machine
Put a number on what happens next — and show the working.
Ask a question about the future. The machine turns it into something that can be scored, gathers the evidence, and has six independent analysts read it — none of them able to see what the others said. Deterministic code combines their answers into one probability, then a seventh agent attacks it.
Six readings, taken blind
Actors, incentives, history, belief, structure, resilience. Each lens answers the same question without seeing the others.
Combined by arithmetic
Log-odds pooling, with lenses that share a method discounted so a crowded view cannot vote twice.
Attacked before publishing
A skeptic argues the other side. Accepted challenges move the number, and the whole run stays on the record.
Checking session…
API health (via the BFF, over the private network): /api/health