AI can change a lot of files quickly. I want a clear way to see those changes and return to an earlier working version when needed.
That is why Git is part of my setup.
Git records the changes I choose to save in a project. Each saved checkpoint is called a commit. GitHub is a service where I can keep a remote copy of that history and work with other people.
Those are different jobs. Saving a commit on my laptop does not automatically upload it. Uploading the changes is a separate step, often called a push.
The habit I want is simple: save a working starting point, make a small change, review it, test it, and save another checkpoint.
A branch lets me work on a change separately from the main saved version. Before bringing it back into the main version, I can inspect the differences. Developers call that comparison a diff.
For AI-assisted work, I want to read the actual changes, even when the tool provides a reassuring summary. If a change is unclear, I can ask for an explanation and then check the relevant files.
I plan to start practice repositories as private. Private access reduces who can see the work, but it does not make a repository a suitable place for passwords or private customer information.
Each project also needs its own ignore file, named .gitignore. It tells Git which untracked files to leave out. It does not hide files from an AI tool, and it does not erase something already saved in Git history.
I will keep the active code project outside ordinary file-sync folders to avoid competing systems trying to manage the same files. Separate backups still matter for work Git does not contain.
Try this with a disposable file: save a checkpoint, make a change, review the difference, and practice recovering the earlier version. That is a much better place to learn recovery than an important project.