I build AI agents
that do the work.
$
I'm a Pentester, YouTuber, and proud father of SwarmAttacker. I spend my days breaking into things on purpose, building autonomous agents that do it for me, and filming the messy middle. Incurable LangGraph fan.
About me
I got here by following the fun. I started out breaking things — the discipline, then the deep end of IT security — and somewhere along the way I realized the most interesting adversary I could build wasn't a script, it was an agent that reasons.
These days I live inside agentic systems. State machines, tool calls, planners, critics, the whole orchestra. I'm a shameless LangGraph fan — I like being able to draw the graph, watch it run, and actually understand why the model did what it did. When something needs to plan, act, and self-correct without me babysitting it, that's my happy place.
My main obsession right now is SwarmAttacker — a project I care about the way you care about a kid. More on that below.
I'm also starting a YouTube channel to show this stuff in the open — breaking things, building agents, and the honest messy middle. Link's coming soon.
SwarmAttacker
A swarm of AI agents that thinks like an attacker.
SwarmAttacker is where my two worlds collide: offensive security and agentic AI. Instead of one model trying to do everything, it's a coordinated swarm — recon agents, planners, exploiters, and critics — each with a job, talking to each other, adapting as they go.
The idea is simple and a little unsettling: give a set of specialized agents a target and a goal, and let them collaborate the way a real red team would. Enumerate, hypothesize, probe, learn, escalate — autonomously, and with a paper trail you can actually read.
It's built the way I like to build: LangGraph at the core, clear state, explicit tools, and every decision traceable. It's the project I'm proudest of and the one I'm putting the most into.