video1.0<iframe src="https://www.loom.com/embed/cd7a8a6d239b42c4857c4f1e3d6109ed" frameborder="0" width="1662" height="1246" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>12461662Loomhttps://www.loom.com12461662https://cdn.loom.com/sessions/thumbnails/cd7a8a6d239b42c4857c4f1e3d6109ed-6750914ac59d6851.gif101.939This Loom explains how to train and deploy an owned judgment layer for cyber security intelligence using River and Jibrian. The speaker reports training on 560,000 findings and solving the problem for about 8 cents, then notes that the approach can be about 3.5 million times cheaper than prior costs and yield 2.5 times fewer catastrophic misses. They describe converting 900 findings into about 300 labeled pairs, running in 7 minutes on average with 8 cents total per run, and using a 3-step CLI to make deployment easy while still performing well on cases the model never saw. They also cite results like a 20 point increase in close findings and convergence of loss from 19 down to 10 toward 0 over training runs.