Case 006: autonomous local AI agents took CVE-2026-53988 into an isolated Dockhand lab. They rebuilt the unsigned-webhook chain to root on a simulated managed Docker host, verified v1.0.40, and…

Autonomous local AI agents took CVE-2026-58138 into an isolated @orkesio Conductor lab. They turned a workflow into root inside the victim container, verified the 3.30.2 fix, then built detections…

Autonomous Research Lab | Case 004 AI agents took on CVE-2026-85706, GitLab’s CVSS 10 flaw. They reproduced the file read, followed a fictional CI chain to root, verified the fix, then created…

PaperCut confirmed real customer compromises. I gave the public exploit chain to autonomous AI agents running locally. They rebuilt it, tested the fixes and turned the attack into detections…

Qwen asks builders for real-world workflows. We gave Qwen3.8-Flash-Next one: investigate a critical CVE end to end, fully local. It built the lab, reproduced CVE-2026-60004, verified the fix 22/22,…