Writing

Notes from the field

Research notes, essays, and excerpts from the AI Threat Hunting book in progress.

featured

When half your evidence is a model's chain-of-thought

Threat hunting was hard enough when the evidence was logs. With AI agents, the evidence is also reasoning. Here's what changes.

featured

Blast radius for AI agents: a working framework

When you give an AI agent a tool, you're handing it the keys to everything that tool can reach. Here's how I think about scoping that.