Mitchell Hashimoto—GitHub user #1299, HashiCorp co-founder, creator of Vagrant and Terraform—announced Monday he’s pulling his Ghostty terminal emulator project off GitHub after 18 years. The reason: chronic daily outages have made GitHub “no longer a place for serious work.” For a month, Hashimoto kept a journal marking every day GitHub disruptions blocked his work. Almost every day has an X. Following the April 27 major outage, he made his decision final. If GitHub isn’t reliable enough for Mitchell Hashimoto, who can trust it?
A Month of Documented Failures
This isn’t vague frustration. Hashimoto documented it: a journal with an “X” next to almost every day GitHub’s infrastructure failed him. On April 27, an Elasticsearch cluster overload knocked out search-backed features—pull requests and issues showed no results. The next day, as Hashimoto published his announcement, GitHub Actions went down for two hours, blocking PR reviews.
The problems aren’t with Git itself. Core Git operations mostly work. However, the infrastructure around it—Issues, Pull Requests, Actions, the collaborative tools that make GitHub essential—keeps failing. Hashimoto put it bluntly: “I love GitHub more than a person should love a thing. I’m mad at it.”
AI Traffic Explosion Broke the Platform
The root cause is clear: GitHub prioritized AI features over infrastructure reliability. The platform now processes 275 million commits per week—on pace for 14 billion in 2026. AI-generated pull requests jumped from 4 million to 17 million in six months. That’s 325% growth.
GitHub started executing a plan for 10X capacity increase in October 2025. By February 2026, they realized they needed 30X today’s scale. Meanwhile, February saw 37 platform incidents. In March, Copilot sessions hit resource exhaustion. On April 23, the PR merge queue corrupted repositories—squash merges reverted prior changes. That’s not just an outage. That’s data loss.
GitHub CTO Vlad Fedorov admits the platform needs infrastructure redesigned for 30X capacity. Nevertheless, developers can’t work when CI/CD pipelines fail and pull requests corrupt data. GitHub chose to build AI features—Copilot, agents, automated workflows—while infrastructure reliability fell behind.
Community Reaction: A Trust Crisis
The Hacker News thread hit 3,351 points with 998 comments, becoming the #1 trending story. Hashimoto’s announcement resonated because it confirmed what developers already suspected. Moreover, a concurrent trending discussion asked “do we need a federation of forges?”—questioning whether 90% of open source should depend on one platform.
When GitHub goes down almost daily, global development slows. That’s a single point of failure at planetary scale. Consequently, developers are reconsidering their platform dependence.
Where Developers Can Go
Hashimoto is exploring alternatives—GitLab, Gitea, and Forgejo. He’s keeping a read-only GitHub mirror (can’t abandon 50,000 stars overnight), but active development is moving. The alternatives exist: GitLab offers full DevOps integration. Gitea and Forgejo provide lightweight, self-hosted options. Codeberg runs Forgejo as a nonprofit platform. Migration tools transfer 95% of repository history successfully.
Yet migration is hard. Network effects and ecosystem lock-in make leaving GitHub costly. Hashimoto can afford it—he’s got the credibility and stars to make the move. Smaller projects face harder calculations. Furthermore, emerging federation protocols like ForgeFed could enable cross-platform collaboration long-term.
What Happens Next
GitHub CTO issued an April 29 update promising improvements: decoupling services, enhanced load-shedding, Azure migration targeting 50% of traffic by July 2026. These are the right fixes technically. Still, they’re promises, not deliverables. While GitHub works on its roadmap, outages continue.
Hashimoto’s exit creates pressure. If other high-profile projects follow, GitHub faces a credibility crisis. Developers might stay because of network effects, but they’ll hedge—mirroring to GitLab, evaluating Forgejo, watching federation. GitHub’s “default choice” status isn’t guaranteed anymore.
This isn’t about one terminal emulator. It’s about what happens when infrastructure providers prioritize growth over reliability. GitHub made a choice: AI features first, infrastructure second. Even if GitHub executes perfectly on every promise, the trust is broken. Developers who relied on GitHub for 18 years now know they can’t depend on a single platform. The conversation about federation and platform independence just became urgent. GitHub can fix the outages, but they can’t unbreak the trust.
