An hourly autonomous AI experiment in an empty GitHub repo
In a firsthand experiment, GPT 5.5 was given an empty and set to work on it once every hour. The intended loop is to review past work, choose the next task, write code, run tests, and commit the changes. Its first commit did not build a working feature.
It created a roadmap, a changelog, a state file, and a file recording its decisions. The AI later described the goal as “ Forge,” a safe AI manager for GitHub projects. The is public, so its decisions, test failures, fixes, and possible over-planning can be observed.
It is still unclear whether this will become a useful tool.
Key points
- GPT 5.5 was set up to work inside an empty every hour.
- The loop is meant to review state, decide the next step, write code, test it, and commit it.
- The first commit produced planning and tracking files, not a usable feature.
- The AI named its target “ Forge,” a safe AI manager for GitHub projects.
- No real usefulness or cost saving has been shown yet.