The question in the title has a question mark for a reason. GitHub is not sinking — it just added 36 million developers in 2025 alone, and its 630 million repositories represent a gravitational pull no alternative can match today. But a small, vocal, and technically credible cohort of developers has started migrating, and their reasons are concrete enough to take seriously. Understanding those reasons — and being honest about where the alternatives fall short — is more useful than either "GitHub is dying" panic or reflexive dismissal.
Let's look at what's actually happening.
The Three Real Complaints
1. The AI Training Policy Shift
The sharpest recent trigger was GitHub's March 2026 announcement: starting April 24, all Copilot Free, Pro, and Pro+ users would have their interaction data — prompts, code snippets, suggestions, file context — used to train AI models by default. Business and Enterprise tiers are exempt. Free and lower-paid users opt out manually or not at all.
This is a policy reversal. When Copilot launched in 2021, using public training data drew backlash. The new change extends that logic further, treating the interaction layer itself as training material unless you actively object. The Register covered the announcement with the blunt headline "GitHub: We going to train on your data after all."
Note the scope: this is Copilot interaction data, not your repository code. Public repos have always been fair game for AI training by anyone who can read them. But the combination of "opt-out by default" and "applies to the free tier most developers actually use" landed badly. GitLab, by contrast, has contractually prohibited AI vendors from using customer inputs or outputs for training purposes — a genuine differentiator for teams where data governance matters.
2. Reliability Has Genuinely Degraded
This one is harder to argue away. Between May 2025 and April 2026, IncidentHub tracked 257 separate incidents on GitHub, 48 of which were classified as major outages. In 2024, the platform saw 119 incidents including 26 major disruptions. The trajectory is the wrong direction.
The root cause is the AI-driven development boom. GitHub started a 10x capacity expansion in October 2025 — then realized by February 2026 that 30x was closer to what was actually needed, as agentic development workflows sharply accelerated in late 2025. The infrastructure simply did not keep pace.
Two high-profile departures landed in quick succession. The Zig programming language project moved to Codeberg in December 2025 after a critical GitHub Actions bug that hung build servers indefinitely sat unresolved for months. Ghostty's creator Mitchell Hashimoto — 18 years on GitHub — pulled his project in April 2026, describing the platform as "no longer a place for serious work." These are not random complaints; they are maintainers of serious, actively-developed projects who hit concrete, reproducible problems and chose to leave rather than wait.
3. Centralization and Jurisdictional Risk
Less dramatic but structurally significant: GitHub is a Microsoft subsidiary hosting the majority of the world's open-source code on US-controlled infrastructure. For European government agencies and privacy-sensitive projects, that is an uncomfortable dependency.
The Netherlands soft-launched code.overheid.nl on April 27, 2026 — a self-hosted Forgejo instance for government agencies. The rationale was explicit: external platforms like GitHub are outside government control and not fully free software. The Dutch government classified that as an unacceptable risk for public-sector code. This is not a fringe position; it is the first major government to act on it at a national infrastructure level.
Forgejo is a hard fork of Gitea maintained by Codeberg e.V., a democratic nonprofit based in Germany. It is licensed GPLv3+, has no enterprise tier or proprietary upsell, and is what powers both Codeberg.org and the Dutch government's code.overheid.nl. If you want a self-hosted forge with genuine community governance and no single commercial owner, Forgejo is currently the strongest candidate.
What the Alternatives Actually Offer
GitLab is the pragmatic choice for teams that want feature parity and do not want to give anything up. Its CI/CD pipeline system is more expressive than GitHub Actions for complex multi-stage workflows, its built-in container registry, security scanning, and package management are more complete, and its self-hosted Community Edition is free — no license cost, just infrastructure. GitLab does not train on customer code at any tier. The tradeoff: GitLab SaaS runners cost roughly 67% more per CI minute than GitHub's January 2026 pricing ($0.01 vs $0.006 per minute on Linux), and the integration ecosystem is about 700 entries deep compared to GitHub Marketplace's 20,000+. Copilot versus GitLab Duo is not a close contest for AI assistance features at this point.
Codeberg is the ethical-hosting choice for open-source maintainers. As of late 2025 it hosts over 300,000 repositories and 200,000 registered accounts. It runs on Forgejo, enforces no ads or tracking, and is operated by a German nonprofit on a membership model. It is also genuinely small — storage limits were introduced specifically to manage sustainability. For a personal project or a small open-source library, Codeberg is a coherent choice. For a team that needs SLAs, role-based access, and enterprise SSO, it is not the right tool today.
Self-hosted Forgejo or Gitea threads the needle for teams with infrastructure comfort and specific compliance requirements. A 25-person team can run a capable Forgejo instance on a single $80/month cloud VM and pay zero in licensing. Migration tooling imports issues, PRs, comments, labels, milestones, and releases — the main losses are GitHub Projects v2 boards, Discussions, and GitHub Pages. Most teams report 95%+ of useful history surviving migration.
The Network Effect You Cannot Relocate
Here is the part that honest coverage of GitHub alternatives usually underweights: GitHub's network effect is structural, not cosmetic.
180 million developers. The default upstream for npm packages, PyPI packages, and most other registries. The place where open-source contributors expect to find your project. If you publish a library and it lives on Codeberg, you will lose pull requests from developers who will not make an account on an unfamiliar platform. If your team's hiring pipeline expects GitHub profiles, self-hosting adds friction that has real cost. None of this is Microsoft's doing in some conspiratorial sense — it is just the compounding effect of a decade of network growth.
The 2026 data is clear: 59% of surveyed developers say they want to use GitHub for code collaboration; GitLab comes in at 22%. That gap does not close because one platform has a better AI training policy.
This is why "who should actually move" is a more useful question than "is GitHub dying."
Who Should Actually Consider Moving
You have a genuine case for exploring alternatives if you are in one of these categories:
European government or regulated-sector teams where data residency and software freedom are compliance requirements, not preferences. The Dutch government's move to Forgejo reflects a real and growing policy direction in EU public administration.
Open-source maintainers who have been bitten by reliability in ways that blocked releases or broke CI. If you have filed GitHub bugs that sat unresolved for months, the cost-benefit of migration shifts materially.
Teams with strong DevOps maturity and compliance requirements who want CI/CD, security scanning, and container registries under one roof with no vendor training their models on interaction data. GitLab self-hosted is the answer here, not Codeberg.
Individuals who simply object to the policy direction and are willing to accept the discovery and contribution friction that comes with a smaller platform. Codeberg is a reasonable choice. It is a choice with real tradeoffs, but they are knowable and manageable for a personal or community project.
Who Should Probably Stay
If your primary measure of success is getting contributors, traction, and visibility for an open-source project, GitHub is still dramatically better. Codeberg does not have the pull request from the random developer who found your project through search. GitLab.com has a fraction of the open-source discovery surface.
If you are a startup or a product team and GitHub Actions plus Copilot is part of your workflow, the productivity cost of moving is real and the reliability issues — while genuinely worse than they were — have not crossed the threshold of "GitHub is unusable." The platform handled 630 million repositories in 2025. Most teams are not hitting the agentic workflow scaling wall that burned Zig and Ghostty.
If you are evaluating GitHub alternatives because you read a headline, do the actual assessment: have you personally been blocked by an outage in the last six months? Does your organization have a stated data governance requirement? Is there a specific feature you need that GitHub does not offer? If the answers are no, the switching cost is probably not worth it.
The question mark in the title is the honest framing. GitHub has real problems that real projects have been materially harmed by. It also has network effects and ecosystem depth that no alternative matches in 2026. Both things are true. The answer to "should I move?" depends entirely on which set of problems you actually have.
Originally published at pickuma.com. Subscribe to the RSS or follow @pickuma.bsky.social for new reviews.
Top comments (0)