git Is Amazing

Listen, git is the most important tool in my stack. Not Cursor, not Claude Code. git.

And it doesn't matter if you're coding alone or with AI - git was critical ten years ago and it's critical today. What has changed is why.

Before AI

Linus Torvalds wrote git in 2005 because he needed to manage Linux development - a massive project with thousands of developers working in parallel on the same codebase. He needed something that could handle branches, merges, and let people work without stepping on each other.

And it caught on. Not just for Linux but for the entire development world. git became the standard for code management, team collaboration, and review processes. The entire modern development workflow was built around it.

After AI

Today AI needs a place to work. A playground where it can modify files, delete, add, refactor - without us worrying it'll break something.

And that's exactly what git provides. A branch is a sandbox. Everything the AI does there, stays there until we decide otherwise. git reset and git diff are always there. You can let AI run free because you can always roll back.

git went from being a collaboration tool between developers to being the AI's playground. And without it? Nobody would let an agent touch their codebase.

Worktrees

git worktree is a feature that's been gaining serious traction lately, and it's no coincidence. Both Cursor and Claude Code use worktrees behind the scenes to run agents on code without affecting the main working directory. And developers around me are starting to use it more and more - once you try it, there's no going back.

So what is it? The ability to have multiple branches open simultaneously, each in a separate directory, on the same repository. No additional clone needed.

git worktree add ../my-project-experiment feature/ai-refactor

Now you have two folders - the regular project and a separate experiment. Both share the same git history but each is on its own branch.

Why is this so relevant with AI?

Parallel experiments. Let AI do a refactor in one worktree while you keep working in the main one. Worked? Merge. Didn't work? Delete it and nothing is harmed.

Comparing approaches. Ask the AI for two different solutions, each in a separate worktree. Run both, check performance, choose.

Main stays clean, always. When AI works in a separate worktree, your main is never at risk. Test, review, and only then decide.

The Oldest Tool, The Most Relevant

There's an irony here that I love. The whole world is talking about AI, new tools every week, models improving every month. And in the end, the tool holding it all together is git. A tool from 2005.

git didn't become less relevant. It became more important than ever. It's what lets us move fast without fear. And if you haven't discovered worktrees yet - now is the time.