Hashimoto argues that the core issue isn't whether AI training on open-source code is inherently wrong — it's that GitHub changed the terms of the relationship after hosting projects for years without meaningful consent or opt-out mechanisms. He frames this as a governance failure: a platform shouldn't unilaterally monetize hosted code through commercial AI products while maintainers have no say in the matter.
Beyond the AI training grievance, Hashimoto makes a philosophical case for decentralized forges. Git was designed to be decentralized, but GitHub's proprietary ecosystem of Issues, Actions, Discussions, and Packages has created platform stickiness that effectively centralizes open-source infrastructure under one corporate owner. Migrating Ghostty to a self-hosted forge is presented as a principled stand for the decentralized model Git was built on.
The editorial highlights that the post garnered 3,200+ points on Hacker News — one of the most upvoted forge-migration stories in HN history — as evidence that this isn't just one maintainer's grievance but a signal of systemic strain in the developer community's relationship with GitHub. The fact that Ghostty is a serious infrastructure-level project (GPU-accelerated, written in Zig, cross-platform) gives the decision outsized weight compared to smaller protest migrations.
Mitchell Hashimoto — founder of HashiCorp, creator of Vagrant, Terraform, and now the GPU-accelerated terminal emulator Ghostty — announced that Ghostty is leaving GitHub. The project will migrate to a self-hosted Git forge, abandoning the platform that hosts the overwhelming majority of open-source projects.
This isn't a rage-quit or a weekend experiment. Hashimoto laid out a detailed rationale in a long-form post on his personal site, covering everything from GitHub's AI training practices to the philosophical case for decentralized forges. The post hit 3,200+ points on Hacker News — making it one of the most upvoted forge-migration stories in HN history, and a signal that the developer community's relationship with GitHub is under real strain.
Ghostty itself is a serious project: a cross-platform, GPU-accelerated terminal emulator written in Zig, with a native UI on macOS and GTK on Linux. It's not a toy — it's the kind of infrastructure-level tool whose hosting decision carries weight.
### The Copilot problem is now a governance problem
GitHub Copilot trained on public repositories. That's old news. What's changed is the accumulation of grievances: the lack of opt-out granularity, the opaque relationship between code hosted on GitHub and code used to train Microsoft's commercial AI products, and the growing sense that GitHub's incentives have permanently diverged from those of open-source maintainers.
Hashimoto's argument isn't that AI training on open-source code is inherently wrong — it's that a platform shouldn't unilaterally change the deal after hosting your code for a decade. This is the consent problem, and it applies to every project on GitHub, not just Ghostty.
### Platform lock-in is real, even for Git
Git is decentralized by design. You can push to any remote. But GitHub has built an ecosystem of Issues, Actions, Discussions, Projects, Packages, and now Copilot that makes the platform sticky in ways that have nothing to do with version control. Moving off GitHub means rebuilding your CI/CD, your issue tracker, your contributor onboarding, and your discoverability — and that's exactly the lock-in that Git was supposed to prevent.
Hashimoto's post acknowledges this cost directly. The migration isn't free. But his argument is that the cost of staying — ceding control over your project's data, contributor experience, and AI training surface — is now higher than the cost of leaving.
### The forge landscape has matured
Five years ago, self-hosting a forge meant Gitea or bare GitLab, with rough edges everywhere. Today the options are meaningfully better. Forgejo (the community fork of Gitea) has strong momentum. Sourcehut offers a minimalist, mailing-list-first workflow. Codeberg provides hosted Forgejo with no strings attached. And tools like Radicle are experimenting with fully peer-to-peer code collaboration.
None of these match GitHub's polish or network effects. But for a project like Ghostty — which already has a dedicated community and doesn't depend on GitHub for discoverability — the trade-off calculus is different than it is for a new library trying to get its first 100 stars.
This decision carries extra weight because of who's making it. Mitchell Hashimoto built HashiCorp into a multi-billion-dollar company with projects that live on GitHub. Terraform has 43,000+ stars there. Vagrant, Packer, Consul, Nomad — all GitHub-native.
When someone with that track record decides the platform's trajectory is fundamentally misaligned with open-source values, it's not a fringe opinion — it's an informed bet. Hashimoto isn't an ideologue; he's a pragmatist who previously chose GitHub because it was the right tool. The fact that he's now choosing to leave tells you something about how the cost-benefit analysis has shifted.
There's also a personal dimension: Hashimoto stepped down from HashiCorp's day-to-day operations and has been building Ghostty as an independent project. He has the luxury of making decisions that optimize for long-term project health rather than short-term contributor convenience. Most maintainers don't have that luxury — which is partly why GitHub's dominance persists.
If you maintain an open-source project on GitHub: This is a good time to audit your actual GitHub dependencies. How much of your workflow is Git (portable) vs. GitHub (locked in)? If you use Actions, Issues, Discussions, and Packages, your migration cost is high. If you use GitHub as a dumb remote with CI elsewhere, you're already portable. Know which category you're in.
If you're evaluating forges for a new project: The decision tree has a new branch. GitHub is still the default for discoverability and contributor familiarity. But if your project has a niche community (not competing for random drive-by stars), self-hosted or alternative forges are now viable without heroic effort. Forgejo + Woodpecker CI is a credible stack.
If you care about AI training exposure: GitHub's terms give Microsoft broad rights over public repository content. If that matters to your project — because of licensing philosophy, competitive concerns, or simply principle — the only reliable opt-out is not hosting there. Private repos and `.github/copilot` settings are partial mitigations, not solutions.
For the rest of us: Watch the migration closely. Ghostty's experience will be the most high-profile forge migration case study since the post-Microsoft-acquisition wave. If a project with Ghostty's size and contributor base can pull it off without losing momentum, it lowers the perceived risk for everyone else.
GitHub isn't going anywhere — its network effects are enormous, and most projects will stay. But the Overton window on forge migration has shifted. The conversation is no longer "why would anyone leave GitHub?" but "what would it take for *my* project to leave?" Hashimoto just provided the most detailed public answer to that question from someone with real credibility. Whether or not you follow him off the platform, you should read his reasoning. The uncomfortable parts are the important parts.
It really has been remarkable watching GitHub just crumble as an organization. There's a lot of discussion about why: the switch from being independent to being part of Microsoft, having resources pushed to Copilot instead of core service, the organization structure itself, a reliance on vibe c
I can appreciate Hashimoto's genuine feelings about Github, and the world of open-source software development that it opened for him and that he spent a significant chunk of his life participating in.On the other hand, I can't help but think that some of this heartbreak would have been avo
The thing you love has been bought by Microsoft. When things belong to a large corporation, they can (and probably will) drift off in some absurd direction, because in a way, the relationship is reversed. The thing no longer serves you; instead, the brand, the user base, the reputation, and the key
During one of the x threads where Mitchell was (legitimately) complaining about Github, there were a couple replies suggesting that GitHub should hire him to be their CEO.And I remember seeing that and thinking "huh... not at all a bad idea."There is a specific kind of leader that can turn
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
I know this is ridiculously dramatic, but its the truth: I actually cried writing this blog post (tears hit my keyboard, I'm embarrassed to say).Nobody should cry over a SaaS, of all things. But GitHub has meant so much more to me than that (all laid out in the post). I have an unhealthy relati