AI Coding - Don't Ya Forget about dem good'ol Source Control

AI coding assistants are awesome. They're fast. They're confident. They happily generate more code in 10 minutes than I write before my first coffee ☕.

But here's the thing nobody talks about: AI makes it way too easy to create absolute chaos in your repo.

Can watch the Video here:

The Problem Nobody Warned Us About

Picture this (fictional, but painfully familiar) scenario:

A senior developer—let's call them Alex—has been coding for 8 years. Alex knows design patterns. Alex can debug a race condition in their sleep. Alex asks Claude to "add pagination to the API endpoint."

Claude delivers. 200 lines of beautiful, working code.

Alex commits it with the message: updates

Then asks Claude to "make it faster."

Another commit: fixes

Then "add caching."

Commit: more fixes

Then "actually, make the page size configurable."

Commit: final

By lunch, there are 12 commits, the diff is 800 lines, and nobody—including Alex—can figure out what actually changed or why.

AI Didn't Break Your Workflow. It Just Accelerated Your Bad Habits.

Here's what happens when you combine AI-speed code generation with human-speed source control discipline:

Commits every 3 minutes - Because the AI is fast and you're in flow state and who has time for thoughtful commit messages?

Diffs that look like novels - AI doesn't know you wanted that feature split across three logical commits.

Commit messages that read like:

• "test"
• "actually test"
• "ok this time for real"
• "AI suggestion"
• "reverted AI suggestion"
• "re-applied AI suggestion but different"

Branch names like fix-thing and temp and asdf because you were just "trying something quick"

The classic: Someone asks the AI to refactor the entire auth system, commits all 47 files at once with the message "improvements", and now the code review is a novel nobody wants to read.

The Uncomfortable Truth

AI can write your code. It can't:

• Write your commit messages
• Create meaningful branches
• Break changes into reviewable chunks
• Explain to your team lead why production is on fire
• Remember what you were thinking when you made that change
• Save you from git push --force on main

So What Do We Do?

Maybe—just maybe—before we let AI write even more code, we need to level up our source control game.

Not because we're junior developers who don't know better.

But because AI makes it so easy to generate code that our old "I'll clean this up later" habits don't scale anymore.

Some ideas:

• Commit before asking AI for the next feature
• Write the commit message before you generate the code (forces you to think about what you're actually doing)
• Use branches like they're free (because they are)
• Resist the urge to commit AI's first draft—review it first
• Remember that future you will have no idea what "updates" means

The Bottom Line

AI coding assistants are incredible tools. But they're accelerants. They make everything faster—including the rate at which you can turn your repo into an archaeological dig site.

Good source control habits were always important.

With AI in the mix? They're absolutely critical.

Because AI will gladly generate a thousand lines of beautiful code.

It won't explain to your team why the commit history looks like a Jackson Pollock painting. 🎨

That part's still on you.

---

Have your own AI + Git horror story? (Fictional or otherwise?) Drop it in the comments. Let's suffer together. 😅