Local AI Can Red Team. Part 2: Now, Can You Trust the Scoreboard?
TL;DR Part 1 ended on a confession. I’d ranked 33 local models for red-team work with a single local judge that, on some runs, was grading …
Be The Adversary
Offense informs defense.
Recent posts
Engineer's Heart, No Engineer's Training
The model is becoming a commodity and code is cheap. The scarce thing is knowing what to judge and how, and that is exactly the part we are automating away.
Two DGX Sparks, 33 Local Models, One Question: Can Local AI Actually Red Team?
TL;DR: I built a two-node NVIDIA DGX Spark cluster and benchmarked 33 local, open-weight models for real red-team work: BloodHound …
Recent talks
From Snowflake to Snowstorm: Navigating Breaches and Detections
The Snowflake talk taken to x33fcon, a year after CODE BLUE. Same campaign, one that reached more than 165 companies through misconfiguration rather than any …
Beyond Simulations: Advancing Detection Validation in Cybersecurity
A room arguing about whether the way we test detections resembles anything an attacker actually does, and what to do instead without burning a quarter of the …