# The GitHub Platform Era Is Over

> Source: <https://julien.danjou.info/blog/the-github-platform-era-is-over/>
> Published: 2026-07-28 00:00:00+00:00

# The GitHub Platform Era Is Over

For fifteen years, the smart move was to rent your primitives from GitHub. A decade of neglect and a wave of AI just flipped that math. You don't displace GitHub. You stop depending on it.

For fifteen years, the smart move in developer tooling was to build on top of GitHub. That’s over.

Not because GitHub is dying. It isn’t, and I won’t pretend otherwise. GitHub’s real moat was never its features or even its engineering. It’s custody. Your code lives there, your history, your identity, your webhooks, your two hundred integrations. Nobody clones their way out of that, and no AI touches it. GitHub will be fine for a very long time. I [wondered a couple of years ago whether it was becoming obsolete](/blog/is-github-the-future-or-becoming); that was the wrong question. What’s obsolete isn’t GitHub. It’s the reason to build on top of it.

But custody is a moat around *hosting*. It was never a moat around everything built on top of hosting, and that’s the part that just came loose.

I don’t say this from the stands. We’ve built on GitHub’s API for years. Mergify lives inside it, every merge, every check, every queue. We’re a technology partner, and I’ll say the awkward part out loud: we compete with GitHub in exactly this lane. That’s not a reason to discount what follows, it’s the reason I saw it early. We don’t see GitHub through a UI that mostly works. We see it through the API, the edge cases, the undocumented behavior, the tickets. And yes, we see the edges, not the median, because the edges are where a competitor lives. But the edges are also where maintenance stops first, which makes them the leading indicator. From out here, the indicator looks bad.

## We rented our primitives, and it was the right call

For fifteen years, the correct move wasn’t just to host your code on GitHub. It was to rent your *primitives* from GitHub too: the merge logic, the review model, the CI orchestration, the plumbing under every workflow. Building your own was expensive and pointless when GitHub’s were excellent and sitting right there. So a whole ecosystem, us included, built on top of primitives we didn’t own, and trusted GitHub to keep them solid.

That trust was the asset. That trust is what’s gone.

## What the crack looks like from inside

Take the stacked pull requests preview. Half a dozen of our customers got in and started using it. We, the partner whose entire product is merging code, are still on the waitlist. That part is just access politics, and I can live with politics. The problem is what came next: you can’t merge a stack through the API, and there’s no documentation on any of it, so “working as intended” and “broken, and nobody wrote it down” look identical from the outside. Ship a feature to real users with no way to tell those apart, and that’s not a preview problem. It’s a primitive nobody’s tending.

And it isn’t one bad week. We now open a support ticket against GitHub roughly once a week: a bug, a limitation, a doc that’s wrong, a behavior that makes you stop and go *what*. That cadence keeps climbing. What comes back keeps shrinking: less information, slower resolution, more of the polite non-answers you get from a team stretched past caring. A single anecdote is easy to wave off. A rate is a trend, and this one points down.

Meanwhile the energy goes somewhere else entirely. Copilot everywhere, more piled onto Actions, feature after feature added on top of primitives that were already shaky. Actions was never a well-designed CI system, it was a good-enough one that won on distribution: YAML pushed into doing the work of a real programming language, a secrets model everyone works around, reruns you can’t target. Nobody funds fixing the base. They fund the next floor.

## The math that made renting smart just flipped

Renting your primitives made sense on one condition: that building your own was harder than tolerating someone else’s. Two things broke that condition at once.

One, GitHub stopped tending the primitives you were renting. That’s everything above.

Two, AI collapsed the cost of building your own. I want to be careful here, because the easy version of this claim is wrong. AI does not make a merge engine free. Cloning a Copilot panel over a weekend is not the same as running merge logic correctly across a hundred million repositories, and anyone who’s built the reliable version knows the demo was never the hard part. What AI collapses is the cost of *starting*, of getting your own primitives to the point where owning them beats renting degrading ones. That used to be a research project. Now it’s a quarter of work.

And notice AI cuts both ways. It doesn’t politely erode GitHub’s surface and leave the deep stuff alone. It makes everything more cloneable, GitHub’s primitives included. Which is exactly the point: if nothing is protected by being hard to copy anymore, then “hard to copy” was never the moat. The moat is being *hard to displace once you’re running on it*. Custody is that, for hosting. Nobody holds it yet for the primitives, because for fifteen years nobody needed to own them.

## The strategy inverted

So the move flipped, and if you build on GitHub’s primitives you should feel it. You will not displace GitHub. Custody makes that a fantasy, and it was never the goal. The goal is narrower and very real: stop *renting* the primitives you depend on. Own your copy. Run them beside GitHub instead of on top of it. Your code keeps living on GitHub. Your merge logic, your review model, your orchestration, the parts that actually decide how your engineering works, don’t have to.

This doesn’t escape platform dependency, let’s be honest about that. You still rent compute, you still rent git hosting. You’re subletting a different floor. But you move the dependency to a layer where the switching costs start working for you instead of for GitHub. That’s the whole game now.

Yes, that is exactly what we’re doing at Mergify, and I’m not going to be coy about it. We spent years building on top of GitHub’s primitives, and we’re moving the ones that matter off, because we watched them stop being maintained from a vantage point most people don’t have. I’m not telling you this because it’s good for us. I’m telling you because if you build on GitHub’s primitives, the ground shifted under you and most of the ecosystem hasn’t looked up yet.

GitHub keeps the code. That was never in question. The question is whether you keep renting the parts that decide how you build from a landlord who stopped fixing the plumbing, or whether you finally own them. I know which side of that bet I’m on.

## Related posts

### [ GitHub Actions Pricing: The Platform Reality Check ](/blog/github-actions-isnt-getting-greedy)

GitHub Actions Pricing: The Platform Reality Check

[Read more →](/blog/github-actions-isnt-getting-greedy)

### [ Why Engineers Shouldn’t Decide Your Cloud Strategy ](/blog/why-engineers-shouldnt-decide-your)

Growth Is the Battlefield

[Read more →](/blog/why-engineers-shouldnt-decide-your)
