Getting Started: Git

Episode 1 folded three things into one sentence: Git lets you see the diff, undo a bad change, and keep an actual history. Each one earns more than a clause.

The diff

Claude Code edits real files in a real folder, sometimes several at once, and not always exactly the way you would have typed it yourself. git diff shows the exact lines that changed, nothing more, before you decide whether to keep them. Reading the diff instead of just skimming the result is most of what trusting AI-written code actually looks like, day to day.

The undo

A bad change is not permanent. git checkout (or git restore on newer Git) puts a file back to how it looked at the last commit, and there is always a way back further than that if the mistake is older. Working without this, every change is a one-way door. Working with it, none of them are.

The history

A commit is a snapshot with a reason attached, and git log is the record of every one of them, in order, forever. Six months from now, that record answers "why is this line here" better than memory does, yours or Claude Code's.

What else it does

Those three cover daily use. A few more capabilities exist for when the work gets more involved: branches let you try something without touching the version that already works, staging lets you choose exactly what goes into the next commit instead of all of it, and a remote, most commonly GitHub for me, keeps a copy of the whole history somewhere other than the one machine.

None of that is required to start. The three ideas above are.

Be Better - Read the diff before you trust it.