state management, observability, retries, permissioning, recovery paths, eval drift, human escalation. the model is only one component now.

This post from Harrison Chase discusses how AI agents have evolved beyond just the underlying model, emphasizing critical infrastructure components including state management, observability, retry mechanisms, permission controls, recovery procedures, evaluation drift detection, and human escalation workflows. It reflects the modern understanding that successful AI agent deployment requires robust systems engineering across multiple dimensions rather than focusing solely on model capability.

Source: Harrison Chase (X) | 2026-05-09