Adrian De Gendt

I'm building CYBRET AI, where we're working on autonomous security while maintaining human in the loop.

Previously, I've worked across cybersecurity, artificial intelligence and space systems. I've built things involving satellites, machine learning, security tooling and offensive security. I previously worked at Telenor Cyberdefence.

These days I'm particularly interested in AI agents, offensive security, AI safety, interpretability, mathematics, and how increasingly capable AI systems change the way software is built and secured.

CYBRET

CYBRET is an attempt to rethink application security around autonomous systems rather than scanners.

We're building a digital twin of an application from its behaviour and API traffic, then letting agents reason about and attack that environment using black box, grey box and white box techniques.

The thing I care about most is finding vulnerabilities that actually matter. I don't want to produce another list of theoretically suspicious things for engineers to triage.

Writing

Essays and notes on AI, security, systems and what I think is changing.

Things worth reading

Papers and essays I've found interesting. Mostly AI, security, science, systems and building things.

• Dario Amodei, Machines of Loving Grace
• Dario Amodei, The Adolescence of Technology
• Dario Amodei, The Urgency of Interpretability
• Anthropic, Mapping the Mind of a Large Language Model
• Daniel Kokotajlo, Scott Alexander, Thomas Larsen, Eli Lifland and Romeo Dean, AI 2027
• James Mickens, The Night Watch
• Paul Graham, Maker's Schedule, Manager's Schedule
• Paul Graham, Schlep Blindness

principles.md

Elsewhere

GitHub    LinkedIn