Breaking and Building Complex Systems
AI / Robotics / Distributed Systems / Digital Assets / Vulnerability Research and Offensive Security
-
Open-Source Models Solving Red-Team Labs at Machine Speed
A refusal-reduced open model on one B300 reached Domain Administrator in an authorized GOAD range in under 19 minutes. Defense now has to operate at machine speed too.
-
Can Frontier LLMs Audit Smart Contracts? Part 2: Making Every Finding Prove Itself
A follow-up to the Part 1 benchmark: Forge Proof, an open-source pipeline that makes every LLM audit finding prove itself with a machine-checked exploit or die. How it works, and 100% detection on Damn Vulnerable DeFi v4 for the price of lunch.
-
Can Frontier LLMs Audit Smart Contracts?
An internal evaluation from EigenLabs security: 6 frontier LLMs on 14 real audit-fix commits from our own repo. Honest numbers on recall, precision, cost, and the operational pain of running this kind of sweep.