Justin Smith argues that faster AI-assisted coding shifts more pressure onto production operations, where engineers still spend substantial time maintaining systems, investigating incidents, watching deployments, and coordinating fixes.
The proposed background-agent pattern pairs an execution engine with continuously updated production contextContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task.. Agents can run on schedules, respond to deployment or messaging eventsAn always-on agent remains available to respond to events and resume ongoing work beyond one interactive session., work in isolated environments, retain task knowledgeAI agent memory is stored information that an agent can retrieve and use across steps, sessions, or changing contexts., and adapt their checks as systems change.
Four practical workloads illustrate the approach: monitoring releases, running scheduled health and anomaly checksAnomaly detection identifies observations or behavior that differ meaningfully from an expected pattern and may warrant investigation., preparing operational reports and handoffs, and answering routine engineering questions when confidence is high. The central design lesson is to augment established operational systems with context-aware checks rather than replace them wholesale.
Watch on YouTube




