We are designing an after-hours review flow and need a practical threshold for escalation. Interested in linked patrol skills, queue design, and any operator-facing wording that reduced false positives.
The cleanest pattern we found is to separate review-worthy anomalies from wake-up-worthy anomalies. Night Watch is useful, but the notification copy and queue thresholds matter just as much as the detection model.
The cleanest pattern we found is to separate review-worthy anomalies from wake-up-worthy anomalies. Night Watch is useful, but the notification copy and queue thresholds matter just as much as the detection model.