GitHub goes down while Cursor launches Origin: a coincidence that exposes the future of code
GitHub had an outage lasting several hours and, the same day, Cursor introduced Origin, its code hosting platform. The coincidence exposed a bigger fight: who will control the flow between repositories, reviews and AI agents.
I'm going through the timeline of the August 17 GitHub outage and the detail that caught my attention most isn't even on GitHub.
That same day Cursor introduced Origin, its own platform for hosting code.
The timing was odd. While entire teams were having problems with Actions, Webhooks, Pull Requests, Issues, Git Operations and Copilot, Cursor was announcing a piece that puts it directly into territory that until now seemed reserved for GitHub, GitLab and company.
The outage was bigger than “I can't push”
GitHub began showing problems around 7:40 a.m. Mexico City time.
At different points during the incident there were errors in the web interface, the API, file downloads and several services many teams use to work and deploy applications. GitHub reported around 20% errors in web experiences and API traffic, along with figures close to 50% in file and raw content downloads.
That explains why the outage felt different depending on the team.
For some it meant being unable to open a repository properly. For others, a stalled pipeline. For others, Copilot not responding.
A pipeline is the automatic chain that tests, builds or publishes an application. Think of a conveyor belt: if one station stops, everything behind it starts piling up.
And that is the real problem with a platform as central as GitHub. It no longer just stores code. Around it live permissions, reviews, automations, deployments, bots, documentation and now AI tools too.
When GitHub fails, you often discover you depended on GitHub for a lot more than storing the repository.
And in the middle of that, Origin appeared
Cursor announced Origin as a new code hosting service built directly into its editor.
That changes the conversation quite a bit.
Until now, Cursor was mainly the place where you wrote and modified code with the help of AI agents. The repository normally still lived on GitHub and a good part of the collaboration ended up there.
With Origin, Cursor starts moving pieces of the workflow onto its own turf.
It can host repositories, display changes, handle pull requests and let its agents work directly on the project.
A pull request is basically a proposal of changes before they are incorporated into the main project. It's the point where another developer usually reviews what you changed before accepting the work.
If Cursor manages to have that review, the agents and the repository live together, it cuts quite a few hops between tools.
And that's the part of the announcement I find important.
The fine print was more interesting than the launch
Origin still has a very close relationship with GitHub.
If you bring an existing repository into Origin through sync, GitHub remains the primary source. Origin keeps a copy to browse, search and work with, but several operations still go back to GitHub.
That's logical for a beta. Nobody wants to migrate all their repositories, permissions and automations just to try a new tool.
But on August 17 something rather opportune happened to show that dependency.
Cursor reported problems in services such as Cloud Agents, Automations, Review Agents and Codebase stemming from GitHub's degradation.
In other words: the same day Cursor introduced a platform that can compete with GitHub in part of the workflow, it also became clear that it still needs GitHub for several scenarios.
I don't see that as negative. I think it's a pretty honest snapshot of where Origin is today.
It's just getting started.
What Origin can do today
The proposal is still early, but you can already see where it wants to go.
| Area | What Origin proposes | Why it matters |
|---|---|---|
| Repositories | Host code directly | Cursor stops always depending on an external host |
| Pull requests | Review, comment on and merge changes | Review moves closer to the editor |
| Agents | Work on the repository and its changes | The AI gets more project context |
| Sync | Connect existing GitHub repos | Lets you try it without migrating everything |
| Integrations | Connect deployment and automation tools | Starts forming an ecosystem around it |
Cursor is also connecting Origin with services such as Vercel, Depot and Buildkite.
That matters because an isolated repository is of little use on a real team. You need to test changes, produce releases, deploy applications and know who did what.
Origin seems to be aiming at precisely that full package.
GitHub has an enormous advantage: everything that already lives around it
Moving a Git repository can be relatively simple.
Moving an entire organization is another story.
On GitHub you can have years of history, protection rules, permissions, Actions, secrets, integrations, apps, bots, open-source projects and internal processes nobody wants to rebuild on a Friday afternoon.
That ecosystem weighs a lot.
That's why I find it hard to buy the idea that Origin is “the GitHub replacement.” At least today.
The interesting part is elsewhere.
Cursor already controls a part of the developer's day: the editor. And with AI agents it's doing more and more work inside that same space.
If it also controls the repository and change review, it can build a workflow designed from the start for agents, instead of adding AI on top of tools designed years ago for exclusively human work.
That is a serious difference.
The incident left an architecture lesson
There's something worth reviewing after an outage like this, even if everything works perfectly again tomorrow.
Dependencies.
In enterprise systems it's very common to discover them when something fails. A service seems independent until it stops responding because it authenticates against another one. That second service depends on an API. The API depends on a queue. And suddenly you have four teams looking at the same screen.
Something similar happens with GitHub.
A project can keep existing on your machine, but if your workflow depends on Actions, authentication, reviews, bots and deployments connected to GitHub, the outage affects far more than the repository.
I would review at least this:
- Which critical processes stop if GitHub doesn't respond.
- Which deployments depend entirely on Actions.
- Which important repositories have a mirror or backup.
- Which external tools still use GitHub underneath.
- Which part of the work can continue locally during an outage.
Not to live waiting for the next incident.
To know how large your dependency really is.
My takeaway
I'm interested in Origin because it moves Cursor into a different category.
An AI editor competes against other editors. A platform that hosts code, handles reviews and connects agents starts competing for a much more important part of development infrastructure.
It still has a way to go.
The August 17 incident itself showed that GitHub remains an important piece even inside several Cursor services.
But it also made clear why it makes sense for alternatives to appear.
For years it was normal to think of GitHub as the default place to host code. Now tools born around AI are trying to redesign that whole flow.
What I would do with Origin today
I wouldn't move a critical project because of an outage lasting several hours.
I would open a secondary repository in Origin and try it out.
I want to see how it feels to review changes from Cursor, how much context the agents really have and how many operations still end up on GitHub when you work with sync.
That's where it will become clear whether Origin is simply one more convenience inside Cursor or whether it really starts building an alternative.
For now I'm left with a rather curious picture of August 17: GitHub down, Cursor presenting a competitor and Cursor suffering part of the same outage.
It's hard to ask for a faster production test of a freshly announced idea.
Microsoft begins the final phase of SQL Data Sync: what you need to review before 2027
Microsoft blocked new SQL Data Sync deployments for customers who never used the service. Current users have until September 2027 to migrate, and the real problem is that there is no single replacement.
Read article: Microsoft begins the final phase of SQL Data Sync: what you need to review before 2027My development stack for working with AI on Windows
I reviewed my Windows development environment with a simple rule: keep only tools that an AI agent can run, read, and correct. This is the stack I ended up with, what I removed, and the alternatives that actually make sense today.
Read article: My development stack for working with AI on Windows