Cloud breaches are rarely exotic exploits. They are identity and configuration decisions. For a security leader, the cloud shifts where risk concentrates and where control spending pays off, and a threat model that ignores that shift protects the wrong things.
6 min readThe mechanics of threat modeling are well understood. The reason most programs fail is not technique; it is the operating model around it. What a security leader has to get right for threat modeling to survive contact with a delivery schedule.
6 min readMost expensive breaches trace back to a design decision nobody challenged in a meeting. Threat modeling is how leaders catch that class of risk early, and how to tell whether your organisation actually does it or just performs it.
5 min readNVIDIA announced NemoClaw at GTC 2026 as the security layer OpenClaw always needed. I evaluated both thoroughly; here is my honest verdict on the architecture, where NemoClaw genuinely improves the security posture, and where it still falls short.
15 min readI deployed 6 specialised AI agents for bug bounty hunting using OpenClaw. Here's the honest account: initial scores, what broke, the triage turning point, and what actually improved.
9 min readSame prompt, same context, two different LLMs, two completely different decisions. A real experiment that exposes why inconsistent AI guardrails are themselves a security vulnerability.
11 min read