https://sachncs.github.io/writingTechnical writing on AI architecture and engineering.enhttps://sachncs.github.io/writing/designing-reliable-ai-agentshttps://sachncs.github.io/writing/designing-reliable-ai-agentsMon, 21 Sep 2026 00:00:00 GMTIdentity, authorization, durable state, and recovery determine what an agent can safely do after it chooses an action.AgentsSecurityArchitecturehttps://sachncs.github.io/writing/evaluation-as-infrastructurehttps://sachncs.github.io/writing/evaluation-as-infrastructureMon, 21 Sep 2026 00:00:00 GMTA useful evaluation system connects versioned evidence, explicit failure categories, and release decisions across the life of an AI application.EvaluationProduction AIEngineering