Writing from the build: how systems are designed, how they fail, and what it takes to keep intelligent software honest. Fewer hot takes, more working notes.
Demo article: where autonomous agents actually go wrong, from tool-call injection to scope creep, and the boundaries worth enforcing.
Demo article: designing LLM guardrails as system components, with policy engines, output contracts, and failure modes you can observe.
Demo article: what security-critical UX borrows from classic interface design, and why trust is a layout decision too.
