A frozen copy of the watch loop's output for Sudan, Iraq, Iran, Nigeria and Kenya, scoring the 24 hours up to the capture time against a rolling baseline that is forbidden to train on that window. It includes the provenance table showing, per country and per signal, exactly how old the newest measurement was and whether that lag was normal.
The detector run against known past events — Sudan's February 2024 collapse, Iraqi exam-period shutdowns, Iran September 2022 — with Germany over identical windows as a control that should stay CLEAR. Each panel shows the fitted seasonal baseline and the window used to fit it. This is a test of whether the detector gets known cases right, not a view of the internet now.