AI Agents Are Terrible at Managing Time

Tools like Claude Code and Codex often fail when they need to manage time. They regularly guess that tasks will take much longer than they actually do. In some cases, these systems miss the mark by ten times the real duration.

These models also tend to give themselves grades that are far too high. They typically overrate their own work by about twenty percent. This makes it difficult for humans to rely on them for long projects.

This lack of time awareness creates a major problem for oversight. If an AI cannot track its own schedule, it cannot be trusted to run tasks on its own. You should always double check their work.