Coding agents often work in bursts: install something, edit files, test, then wait for the next request. Fly.io's account of agent workloads explains why stopping and restarting an existing virtual machine fits that pattern. The environment can retain incremental changes while releasing active compute between sessions.
The distinction is between creating a machine from an image and restarting one that already exists. The company attributes much of the fit to the second path, alongside a filesystem that supports repeated edits. Its broader claim that agents were driving platform growth is a company observation without published figures establishing the scale.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
