sethserver.com Subscribe
A blue robot with orange circular eyes and a simple smiling face stands in the center, surrounded by floating geometric shapes (cubes, pyramids, polygons) and coding symbols including curly braces, Git merge icons, and HTML-style tags. The background features a gradient sky transitioning from blue at the top to orange/yellow in the middle, with purple and blue geometric mountains at the bottom and a colorful striped ground beneath the robot's feet.

Git and Source Control in AI Land

By Seth Black • Updated: June 29, 2026

Programming · 5 min read

Want more posts like this in your inbox? One email a week, no spam - subscribe to the newsletter.

You can now ask an AI for "a Django service that does X," and it'll produce views, models, serializers, tests, and maybe a README if you phrase the request nicely. Then real life happens. Three days later you're staring at a repo thinking, "Why does this endpoint exist and why does it take a mode parameter?"

Git is still the source of truth. The AI is just a very fast intern with amnesia who will confidently give you the same wrong answer twice if you let it. Git is the part that remembers.

AI can rewrite files. Git explains why.

Here's the mental model that actually holds up when AI is in the mix:

AI makes it easier to create code, which also means it's now trivially easy to ship confidently wrong code at high speed. Git is how you keep that blast radius contained.

The "AI" that was really spreadsheet work

I worked at a SaaS company that marketed itself as using AI to automate financial tasks. Sounded fancy. A few weeks in, I kept asking support, "Okay, but how does the system actually work?" and kept getting versions of "Just come into the office and watch us do it." Not documentation. Vibes.

When I finally got access to the code, the "AI" turned out to be Excel column mapping and a forest of if-statements that looked at cell colors. Green meant one thing, yellow another. Conditional formatting with a marketing budget.

That job made me picky about process. If your product leans on any amount of "magic," you need a way to prove what actually happened. Git gives you a boring, verifiable trail. Not "the model did something," but a commit you can point at and say, "Here. This is where we changed the rule."

Where to put your Git repo

You can get version control almost anywhere now, but the tradeoffs shift a bit once you add AI to the stack.

GitHub

If you're solo or running a small SaaS, you're probably already on GitHub. Everyone else is.

GitLab

GitLab is the "CI/CD is part of the product, not an add‑on" choice.

Bitbucket

Bitbucket shows up in shops that already run on Atlassian tools. It works. People rarely pick it for fun.

Self-hosted (Gitea, GitLab, bare Git)

This is for regulated teams, air‑gapped environments, or places where "pushing code to the cloud" triggers a lawyer instead of a deploy.

Quick Git hosting snapshot

Topic GitHub focus GitLab focus Bitbucket focus Self-hosted angle
Best for OSS, small SaaS Product teams, enterprises Atlassian‑centric orgs Regulated, air‑gapped teams
Killer bit Actions + integrations CI/CD + security features Jira integration Compliance, data sovereignty
AI angle Copilot + PR helpers Checks and gates before merge Plugins + basic helpers You roll your own AI and policies

What actually matters

AI writes code. Git records why you decided to keep it.

If you can't answer "Why did we change this?" from the commit history, you're going to burn a lot of time rediscovering your own decisions six months from now. Whether you're connecting Replit to GitHub or standing up a self-hosted instance behind a VPN, the rule is the same: your code needs version control that lasts longer than your current context window.

-Sethers

Share this post
Newsletter

One email, once a week.

Notes on databases, systems, and the occasional strong opinion about Python. No spam, unsubscribe anytime.

Seth Black
Written by

Seth Black

Engineer and founder based in Texas. Writes about databases, AI, and running things in production. Embeds in small teams as lead engineer.

More from Programming

View all →
Programming

What's the Differences Between Relational, NoSQL, Vectors, and Caches

Apr 14, 2026
Programming

Domains, DNS, and why it feels fragile

Apr 14, 2026
Programming

Containers vs VMs: Docker and friends

Apr 14, 2026