With over 25 years of experience building software systems, now focused on AI agents, harness-driven execution systems, and auditable AI infrastructure. Currently building PonyBunny, an AI Agent OS designed to move beyond black-box demos toward systems that can plan, act, recover, and explain themselves under real constraints. Particularly interested in agent reliability, evaluation, observability, and production-grade AI system design. More interested in AI as infrastructure than AI as hype.