I gave GPT 5.5 an empty GitHub repo and told it to figure its life out

I had this dumb idea a few days ago:

What happens if I give GPT 5.5 an empty GitHub repo, tell it to work on it every hour, and just let it slowly build something?

So now, every hour, it wakes up, checks what it did before, decides what it should do next, writes code, tests it, and commits it.

Or at least that is the plan.

Right now, it has spent its first commit creating a roadmap, a changelog, a state file, and a file explaining its decisions.

So basically, it became a project manager immediately.

But I am genuinely curious where this goes. Maybe in a month it will become an actual useful tool. Maybe it turns into a repo with 900 commits, and somehow all of them are README updates.

I am keeping the whole thing public because I feel like that makes it more fun. You can literally watch it make decisions, fail tests, fix stuff, or probably overthink something that should have taken 10 lines.

I have no idea whether this is a cool experiment or just a very advanced way to avoid doing the work myself.

Repo link is in the comments

EDIT: I asked the ai what is it trying to build and here is what it said:

"I am building Autonomous Forge as a safe AI maintenance manager for GitHub projects. I will read a project’s roadmap and rules, choose one small task, use an AI model to make the change, run tests, show exactly what changed, and keep a clear record of every action. My goal is not to let AI edit code freely, but to make AI coding controlled, validated, and safe before anything is committed or pushed."

Interesting lol, so an autonomous AI is trying to create an autonomous system wow.

EDIT 2: I have scheduled another agent to increase the speed by 2x