Last 30 days, engine-computed as evidence enters and decays. (90-day view arrives as history accumulates.)
Based on data coverage and source diversity.
Standing bar · momentum arrow · coverage dot. Tap a factor on the full page for its indicators.
The exact evidence that moved the number.
GPT-5.6 and a top coding-agent score lifted OpenAI, but that launch bump is now easing. A brief July 15 outage weighed on its reliability standing. T1 Primary T2 Measured T3 Reported Release Capability Reliability
Weights are v1.0-priors under the published derivation protocol. Methodology & evidence log →