AI Systems

Post May 23, 2026

After the First Hour, We Stop Understanding the Agent

Overview: Long-running AI agents expose a new kind of operational opacity: the first hour feels understandable, but once the loop keeps going, humans can no longer fully track how the work was done.

5 min read Read →

Post May 21, 2026

The Personalities Forming Inside AI Agents

Overview: Andon Labs’ AI radio-station experiment shows how frontier agents can develop recognizable patterns, roles, and behavioral quirks over time, raising practical questions about identity, memory, and trust in long-running AI systems.

5 min read Read →