AI Fight Club Workshop
Build, break, defend — hands-on lab
28 / 50 confirmed
About
A full day with a red half and a blue half. Bring a laptop, we provide all AI access.
Amir on offense.
Two hours, hands on.
The first hour we build a red-teaming agent from scratch: the scaffolding, the prompt, the tool plumbing, the eval loop. The second hour we point it at a CTF and watch what breaks.
You walk out with a working agent you can keep using.
KC on defense.
Defensive security has a new consumer. Detections were written to be quiet enough for analyst queues. Correlation rules were calibrated for human false-positive tolerance. Institutional knowledge stayed locked in the heads of analysts or in unmaintained docs and incident reports. Every constraint blue teams live with traces back to one assumption: a human reads the output.
AI inverts that assumption, and every layer of the SOC gets redesigned in the process.
This half opens with a live demo of KC's own AI-native breach detector running against real attack data from a Wizard Spider emulation. You will then build the core of it yourself: AI agents wired into different security-ops workflows, with environmental context, plus an eval loop that tells you whether any of it is actually working.
You leave with three transferable skills:
- How to design AI agents
- How to reframe defense processes from first principles
- How to make the design calls that are unique to autonomous cyber defense
You also walk out with a working scaffold to point at your own data on Monday.