Notes on agents that keep going.

Long-horizon agentic systems, multi-agent alignment, observability, and evaluation.