Harvard professor builds Jev + Opus 5.5 system to triage security alerts
The system monitors logs and network events in real time, with Jev deciding which signals are worth investigating and Opus 5.5 digging into suspicious cases and producing investigation reports, handling 50,000 alerts in 18 minutes.
a Harvard cybersecurity professor built a Jev + Opus 5.5 system that can investigate 50,000 security alerts every 18 min the system monitors incoming logs, network activity, and system events in real time, while Jev decides which signals are actually worth investigating. Opus 5.5 then takes the suspicious cases, connects evidence from different sources, reconstructs what happened, and generates an investigation report. Jev filters the noise and decides where the investigation should go next. Opus does the deep investigation once a case actually matters. the system essentially turns millions of raw security events into a small number of investigations worth looking at. I found a breakdown of the architecture they used, and the way Jev and Opus split the investigation is surprisingly clever. going to post how the system works next, so stay tuned and send it to someone building AI infrastructure. would you let an AI investigate a security breach before a human sees it?
This post quotes another post: x.com/me_barnyx/status/2101976764779999381