On August 17, GitHub experienced one of its biggest outages. According to its outage report, the cause was this:

“We failed to scale critical components before demand exceeded their capacity. Since April, monthly commits have grown from 1.4 billion to 2.9 billion.”

That’s more than a 100% increase in commits, and we all know why. AI is changing the world by allowing more people to contribute and create code and projects.

This is a real shift. Coding was previously available only to a subset of people, but now many more are embracing and experimenting with it.

But what does this say about the future and about GitHub?

Who is making these commits?

I assume that this 100% increase comes from:

1. Non-engineers using AI
2. Engineers using AI
3. AI agents

This paints a picture of what the future might look like, as we see the usage of platforms like GitHub grow because coding is now easier.

But these three groups are very different, and I think some of them might change over time.

Non-engineers using AI

When I first started studying computer science, I remember using Gentoo, a Linux distro that was compiled entirely by hand. Getting its most basic form up and running would take hours.

That experience, though, led me to learn a lot about Linux and helped me build the know-how to manage servers, use the CLI, and so on.

That learning was crucial, as were other experiences I had with structuring software, understanding what to change, and debugging bottlenecks.

As the next phase of engineering unfolds, I do think this group of people (disclaimer: I’m excited about what they will bring to the world) will also need to face these scalability and growth problems. Any app will need maintenance, and learning what to do might be critical to its success.

Yes, AI might be able to do that too, so we’ll see how this plays out.

Engineers using AI

If you’re in this group, as I am today, welcome. The world is changing, and adapting can also be fun.

What AI has unlocked for this group of people is limitless. Cleaning up the backlog and rewriting codebases, like the Bun rewrite, are just two examples of how powerful AI can be.

AI agents

AI agents are the final group. These are agents running in “bot” form, like Grok Bot, Codex Cloud, Cowork on Cloud, OpenClaw, Hermes, and others.

They might run because they have been summoned or run repeatedly based on specific input.

This means they can create an infinite amount of code, to the joy of every AI model or AI proxy.

What this means for GitHub

As you can imagine, things are only going to get worse. The number of commits will increase even further because creating code is now so easy. With the advent of good local models like Qwen 3.8, price might no longer be a blocker.

This, in turn, means that things will become harder for GitHub as it scales and as new players emerge, like Cursor Origin.

All in all, I also wonder whether we should find a way to host our code in true Git fashion: truly distributed.

The future

Will we be steering AI agents in the cloud? It’s hard to know, but the benefits are undeniable. The amount of work that code-hosting platforms like GitHub need to handle will only increase as AI becomes commoditized.

This means that there are more challenges ahead for them, but it’ll be interesting to see how code hosting itself might evolve over time as a result of this shift.