← Intelligence

Every number here is a receipt

A forecast written after the fact is not a forecast. So GODENGINE writes each one to an append-only ledger — with its resolution criteria fixed at issue time — before any article about it is composed. Each row is hashed over its own contents and chained to the row before it. Change one probability, one criterion, one timestamp, and the chain stops verifying.

Chain status

Verificationverified
Forecasts in chain3
Resolutions hashed0
Head hashbe8f00f1701cb39a0072733f207b9c828956c22f0bac85ff0ff0d07965ad00b1

Do not take our word for it — hit the verification endpoint directly. It recomputes every hash from the stored row contents on each call.

Track record

Forecast accuracy: not yet graded. no resolved forecasts yet. A percentage here before enough forecasts have resolved would be a number about nothing.

Accuracy is the fraction of resolved forecasts that landed on the right side of 50%. Brier is the mean squared error between the stated probability and what happened — lower is better, and it punishes confident wrong calls far harder than hedged ones. Void forecasts (questions that stopped being answerable) are excluded from both, and counted separately so the exclusion is visible.

What the panel numbers mean

Agents loaded is a census of the runtime registry in the process serving this page — 404 right now — not the number the architecture declares. Active this mode is how many of them the current activation tier recruits. Mean live probability is what the engine currently believes across open forecasts; it says nothing about whether the engine is any good. Forecast accuracyis the graded track record and stays blank until enough forecasts have actually resolved. Those last two are deliberately never merged into a single “confidence” figure.