
The Haunted Attack Surface
AI is lowering barriers to specialist cyber work and shortening attack timelines. What comes next, and how security teams can prepare before the window closes.
Read postPosts with an accompanying audio episode.

AI is lowering barriers to specialist cyber work and shortening attack timelines. What comes next, and how security teams can prepare before the window closes.
Read post
Evaluating the April 2026 releases of Mythos and GPT-5.4 Cyber through the lens of the 'Enterprise Stress Test' where legacy debt becomes an active exploit vector.
Read post
A concise, opinionated look at how AI adoption in security acts as a stress test that amplifies existing weaknesses in data, systems, and processes rather than magically fixing them.
Read post
Modern transformer models ship with billions of opaque parameters and undisclosed training data. This post explains why backdoor weights are effectively impossible to verify and why runtime guardrails are mandatory even if you sanitize prompts.
Read post
The Ralph Loop pattern is accelerating both malware development and cyber defense. This post unpacks how agentic automation is being weaponized by threat actors, and how security teams can adopt the same architecture—Ralph-style loops, guardrails, and agentic orchestration—to keep pace.
Read post
Agentic AI systems are quietly turning every connected system into a larger attack surface. This post breaks down direct and indirect prompt injection and the concrete patterns security teams should enforce: containment, input/output filtering, least privilege, and zero trust for agents.
Read post