The GitHub Setup Every AI Builder Needs

Most people treat GitHub as something engineers handle. Every AI build starts from a repo, so skipping it leaves everything you make stuck in the chat.

Dejan Krstic, Chief Product Officer at Product School (ex-Spotify), starts with no account. Live, he creates one, sets up a repo, works in it, makes a change, pushes it, and connects a model to work inside it. Six steps, one session, screen shared the entire time.

No theory. You follow along and leave with a working setup you can reuse in every build that follows.

What you'll walk away with:

1. A repo is where your AI work lives. Chats disappear, but a repo keeps every change, so you can go back to any version when a build breaks.
2. Pushing is what makes AI builds safe. Every push is a save point you can return to, which matters most when a model rewrites files faster than you can read them.
3. Connecting a model turns the repo into a workspace. Once a model can see your repo, it works on your real files instead of pasted snippets.