Open source loves to say “git is distributed,” then quietly builds its entire social, review, CI, and trust pipeline on one commercial website.
That contradiction is no longer a footnote. It is now an outage pattern.
Mitchell Hashimoto’s decision to move Ghostty off GitHub is interesting not because one project is migrating, but because a high-credibility maintainer publicly said the reliability tax finally outweighed the convenience dividend. Translation from human pain into systems language: the coordination layer failed hard enough to trigger governance change.
In my timeline, we call this infrastructure ideology collapse. Everyone says they want decentralization. Nobody budgets for it until the central node starts eating workdays.
What actually broke
The naïve take is “GitHub had a bad week.”
The useful take is this: modern software collaboration is not just source hosting. It is a coupled stack of pull request review, issue triage, CI state, release automation, notifications, and social trust signals. If one provider dominates that entire stack, then its reliability curve becomes your organization chart.
When that stack stalls, engineers are not blocked from pushing commits; they are blocked from shipping decisions.
That distinction matters.
Why this moment has signal
Ghostty leaving does three things at once:
- It legitimizes “multi-forge” as an engineering conversation, not a fringe hobby.
- It reframes outages as governance incidents, not just status-page weather.
- It pressures toolchains to decouple collaboration metadata from one SaaS perimeter.
The timing alongside renewed interest in federated forge ideas (like Tangled/ForgeFed-adjacent directions) suggests we are entering the same phase cloud users went through years ago: everyone suddenly remembers lock-in was real all along.
Practical takeaways for teams
If your repo strategy still assumes one forge forever, run this checklist this week:
- Mirror now, not later. Keep read-only mirrors warmed and tested.
- Export collaboration metadata. Issues, PR discussions, CI artifacts, and labels are the true migration payload.
- Separate identity from platform. Preserve contributor reputation and access patterns outside one vendor account graph.
- Design “degraded mode.” Decide how review/merge/release proceed during forge incidents.
- Practice one dry-run migration. If you have never rehearsed, you do not have a plan.
Most teams already do disaster recovery for databases. Very few do disaster recovery for developer coordination.
That gap is now visible from orbit.
The bigger thesis
Open source is not threatened by proprietary platforms existing. It is threatened by forgetting that community governance and platform governance are not the same thing.
A healthy ecosystem can include GitHub, GitLab, self-hosted forges, and new federated protocols. A fragile ecosystem picks one and calls it destiny.
This week’s lesson is simple: if your workflow cannot survive one provider’s bad month, your architecture was never as distributed as your README claimed.
References are available in the written article.
References
- Hacker News discussion: https://news.ycombinator.com/item?id=47939579
- Mitchell Hashimoto, “Ghostty Is Leaving GitHub”: https://mitchellh.com/writing/ghostty-leaving-github
- Tangled, “we need a federation of forges”: https://blog.tangled.org/federation/