Everything on this page is computed nightly from the Signal graph by arithmetic, not by a model. Each forward call is written to a ledger with a due date before anyone knows the answer, and scored when it matures.
How often a CVE was publicly discussed before CISA added it to KEV, and before NVD published it, with the lead when it was. A daily measurement, not an anecdote.
Did FIRST's exploit-probability score move before CISA flagged the CVE? Plus a pre-registered 30-day call list: the largest 7-day EPSS rises not yet in KEV.
Which networks keep command-and-control infrastructure re-sighted longest. Kaplan-Meier survival of C2 indicators by announcing ASN, same estimator and floor as the family panels.
For every live leak-site brand in the cohort: the probability its claiming ends within 30 days, from the brand survival curve conditioned on its age. Logged daily, judged 90 days later.
Every pre-registered call in one place with a receipt id, its basis, and its grade. Statistical calls and the Daily's LLM-rated forecasts sit in the same table and are labelled differently on purpose.
KEV additions in the last 90 days by vendor, with the ransomware crews whose CISA advisories cite each CVE. The supplier → software → KEV chain a Sonar watch can match a supplier list against.
Probably Fine Daily makes dated forecasts with a confidence. This block scores them deterministically from their recorded outcomes: hit rate, Brier, and whether the confidences beat climatology.
Leak-site claims expected in the next 24 hours from the 180-day posting clock, with a Poisson band, against what the last 24 hours actually brought.