Skip to content

Reports

These pages are records, not explanations of how the pipeline works. They keep the evidence a claim rests on alongside the claim.

Detection reports

Incidents whose diffs the watcher raised, collected under one index so each campaign keeps its own page:

  • In the Wild: the watcher's incident records, how to read them, and what a flag did and did not mean.

Measurement reports

Point-in-time measurements of the rule set on the locked benign corpus, and the methodology and gates behind them:

  • Fire Rates: the false-positive rate of each rule, the 30 % gate, and how a new rule is measured.
  • Benchmarks and Methodology: per-class separation, the CI gates that re-measure it, and the current figures.
  • Hot-path baseline: the measured cost of a review, and the first optimization applied to it.

None of these records is a certificate. A clean measurement describes the corpus and configuration it was taken from, not an AUR update you have not read.