When AI agents learn to engineer themselves

A primer on self-improving agents: Moving beyond the human-coded harness