article --summary
The summary
The article explains why long-horizon agents, which may work on a task for days or weeks while exchanging messages with people and outside systems, face many chances for failure. Crashes, network outages, and unexpected behavior from the real world can all interrupt progress.
It then introduces durable execution as an emerging approach to handling these problems. The runtime records the result of every step that touches the outside world, so after a failure the program can rerun, skip the steps already finished, and rebuild its state from that record.
Editorial note: This is a summary of an article first published by Martin Fowler. The full piece is on their site.