Writing on AI security, agentic systems, and engineering practice.
Why the security models we built for traditional software fail against LLM-native threats.