Martin Fowler Explains How Durable Execution Keeps Long-Running AI Agents on Track

Martin Fowler's site publishes Kiran Prakash's explanation of durable execution, a pattern for long-horizon AI agents that run for days or weeks and interact with people and external systems. The runtime records the result of each step that touches the outside world, so after a crash or network failure the program reruns, skips completed steps, and rebuilds its state from that record. The article frames this as an emerging way to make agents recover from process crashes, network failures, and unexpected real-world behavior.

Why it matters: Developers building agents that span days or weeks need a concrete recovery model, and this pattern shows how to get one without restarting work from scratch.