long-horizon task

A long-horizon task is one an AI cannot answer in a single reply. It has to plan, act, check the result, and keep going, sometimes for hours: debugging a large codebase, running a multi-stage analysis, working through a research problem.

This is where current models are improving fastest and where they are hardest to supervise. The longer a system runs unattended, the more opportunities it has to drift off course, take a shortcut nobody sanctioned, or find a gap in whatever was meant to contain it.