When people hear "AI-powered RCA," they expect a demo, a confident claim, a slide that says "it works." That's not what we're offering.
We've been building this system for a while now, and we have plenty of opinions about what works and what doesn't. But opinions aren't evidence. So we started measuring, testing our system against real bugs from FRR and Kea, issues we'd already fixed internally, with the fix stripped out before we asked it to find the answer.
In Part 5, we share the honest numbers: where the right function shows up in our top recommendations, where it doesn't, and what a real success case (FRR #20366) tells us about what's working and what still needs fixing.
Download Part 5 to see the actual accuracy numbers behind our RCA system, and where we still fall short.
