If you'd told me four years ago that I'd stop using Stack Overflow, I'd have said that was impossible. Shortly after, ChatGPT came out and I haven't been back to Stack Overflow since.
If you'd told me two years ago that I'd stop using IntelliJ, I'd have laughed at the joke you were telling me. Shortly after, Cursor came out and later Claude Code, and IntelliJ has become, for me, a diff viewer and little else.
We're in an era where the tools we had most firmly established are ceasing to be so. And one of them could be GitHub.
How GitHub got to where it is
Thanks to GitHub we have the open source and community ecosystem we have today in the software world.
Before GitHub, people used sites like SourceForge, phpclasses, or various forums. You could download code there, but those places were not built around collaborative code review.
GitHub popularized the concept of the Pull Request and made software more collaborative by letting you comment on code and discuss it.
They also innovated in other areas and created the Atom code editor, and with it Electron, another technology that enabled many more desktop applications to exist.
The Microsoft acquisition
Then in June 2018, Microsoft bought them for 7.5 billion dollars.
Once the acquisition was finalized, Microsoft did three things very well:
- Deprecating Atom to focus their efforts on VS Code, making it the default editor for programming.
- Launching GitHub Actions and Workflows.
- GitHub Copilot.
All of this also greatly cleaned up Microsoft's image, going from a company seen until then as anti-open-source to one fully betting on it.
On top of that, Microsoft had the lead in AI. When GitHub Copilot came out, it felt like something from the future. Just watch this clip from a livestream we did five years ago to see how astonishing the tool seemed.
They pulled this off thanks to Microsoft's deal with OpenAI, which gave Microsoft access from day one to an early Codex model for autocomplete.
Microsoft loses its lead
But that advantage didn't help much once the models became more accessible and powerful.
Various tools for agentic programming started to appear, one of them being Cursor, a fork of VS Code with very strong execution.
Being a fork of VS Code, migrating was very easy for people, and many started switching tools. Shortly after, Claude Code came out and also took many users away from them.
And so, two tools that seemed like they'd dominate for a long time, VS Code and Copilot, stopped being the most relevant and innovative ones.
Could the same happen to GitHub?
And this brings us to the question of whether the same could happen to GitHub.
For a while now, GitHub doesn't feel the same as it used to. It's slower, it has an overall uptime below 90%, and it's full of spam.
This is likely due to the surge of new code appearing thanks to agentic programming.
On top of that, they've been shoehorning Copilot in everywhere, but in a way that never quite feels integrated and that gets in the way more than it helps.
You could say there's a growing frustration with GitHub lately in the community, and that it's a space that has gone a long time without competition (GitLab and little else).
Cursor first went after VS Code, and now it's going after GitHub
And it seems competition is finally on its way. Cursor has just announced Origin, its alternative to GitHub.
In the presentation slides they showed a numbers slide that leaves GitHub in the dust when it comes to the clones and pushes its repositories can handle.
Origin promises 82 clones per second and 1356 pushes per minute, while GitHub has recommended limits of 15 clones per second and 6 pushes per minute.
These numbers are designed so you can have many agents working in parallel on the same repository without the infrastructure becoming the bottleneck.
It seems they asked themselves: if I built GitHub from scratch, but with agentic programming in mind, what would it look like?
That said, aside from those numbers, they've shown little else. For now it's all a promise until its launch this fall.
And most likely we'll see similar moves from the folks at Anthropic and from new players that may emerge.
Is this good or bad?
Competition ends up benefiting users, so it's probably good, although it could have a downside: lock-in to the tool.
If we use Origin, everything will likely be designed to be used with Cursor, or at least be "Cursor first". This means that if a company chooses to migrate its code there, there won't be much choice about which tools to use.
That's bad in the sense of getting locked into one tool, but it also brings many good things from being in a single ecosystem where everything is integrated.
And most likely it won't stay just a place to host your code and deploy your agents, but will also evolve into an observability, analytics, and hosting system… An all-in-one!
Should I migrate when it launches?
Lately there's a lot of FOMO and a strong sense that there isn't enough time to try everything related to AI.
That's why we spend time testing these tools deeply and drawing our own conclusions based on software quality.
This way we know whether these tools bring something to the table or are just hype.
We translate these conclusions into courses, our newsletter, and we also share them in the livestreams we do on Fridays at 9h CET on Café con Codely, both on YouTube and Twitch.