The notebook / Articles
Ideas, under examination.
Published research, a practical handbook, and notes from the tools I build.
Published preprints
Research, with the limits intact.
Research / 29 Sep 2026
The fly that stopped repeating itself
A reward-free, fruit-fly-inspired scheduling study reduced repeated actions in measurable lab runs; it did not prove better vulnerability discovery.
Read the research note ↗Research / 29 Sep 2026When an AI pentest agent should stop claiming it found a bug
What a pre-registered verifier study changed in an LLM security agent's reports, and what the evidence does not establish.
Read the research note ↗Field notes
Build, test, question.
Build an autonomous security agent you can explain
The Autonomous Offensive LLM Handbook now teaches bounded agent construction through runnable offline lessons, a lab, and an honest evidence register.
Better notes, better enumeration
How I think about a useful OSCP reference: organize observations, keep context, and make unfinished work visible.
What your JavaScript leaves behind
The idea behind jsrip: turn scattered client-side code into a reviewable inventory, without mistaking a match for a vulnerability.