Long-running agents are getting better. But they're at their best after a human finds the prompt that lets the model cook. To get started, try this flow from our team: 1. Find your longest agent run What's the longest stretch you have trusted an agent on autopilot? If you don't know, you can't push it. 2. Extend that runtime with a goal Use long-running modes when the success criteria are clear. Codex and Claude Code have /goals commands that allow agents to pursue objectives across…