A post on GitHub’s blog, published on February 12, 2026 and updated the following day, by GitHub’s Director of Open Source Programs. It borrows the Usenet-era term Eternal September to describe the present state of open source (see Eternal September): collaboration runs on trust, trust used to be protected by the friction of contributing, and generative AI has removed much of that friction, so that a pull request, issue or security report can be generated in seconds while the cost of reviewing it has not dropped.
The post is careful to say that most contributors act in good faith and that noisy inbound is not new; the problem it identifies is volume that grows faster than maintainers’ review capacity. It then lists what GitHub has shipped and is exploring for maintainers, surveys approaches the community is building itself, and argues that the answer is better signals and tools rather than walls.
Key Points
- Friction is framed as a balancing act: too much keeps people and ideas out, too little strains the trust open source depends on. The pull request and “Good First Issues” lowered friction, which the author calls a good thing.
- The cost to create a contribution has dropped with generative AI but the cost to review it has not; when volume outpaces review capacity, even well-intentioned submissions can overwhelm maintainers.
- The author argues low-quality or “AI slop” contributions are not a new phenomenon, citing the Linux kernel’s web-of-trust philosophy and Developer Certificate of Origin, Mozilla’s and GNOME’s formal triage systems, and earlier waves of automated scanner reports.
- The imbalance the post identifies is one of benefit: the contributor may get the credit, the CVE or the visibility, while the maintainer gets the maintenance burden. It cites curl ending its bug bounty after AI-generated security reports surged, Ghostty moving to invitation-only contribution, and projects adopting explicit rules on AI-generated contributions as rational responses.
- Features GitHub says it has shipped include repository-level controls to limit pull request creation to collaborators or disable pull requests, pinned comments on issues, banners discouraging “+1” comments, faster pull request diffs and issue navigation, and temporary interaction limits; deleting pull requests from the UI is described as coming soon.
- Directions GitHub says it is exploring with maintainers are criteria-based gating, such as requiring
a linked issue before a pull request can be opened, and improved triage tools that might use
automated triage to check contributions against a project’s own guidelines such as
CONTRIBUTING.md. The post says these should support decisions, not replace them, and that the controls are optional because restrictions can fall hardest on good-faith first-time contributors. - Community responses cited include invitation-only workflows, custom GitHub Actions for contributor triage and reputation scoring, the Vouch trust-management project, and education and onboarding work in communities such as Python and Kubernetes.
- The author argues against building only blocks and bans — “a fortress, not a bazaar” — and says GitHub wants to surface non-code contributions such as issue triage and documentation as trust signals, pointing to WordPress’s “props” credit as a model.
Context
The post is written from GitHub’s position as a platform vendor and describes GitHub’s own product roadmap; its list of shipped features and explored directions is GitHub’s account of its own work. It opens a community discussion to gather maintainers’ feedback. The explicit rules on AI-generated contributions it mentions are the subject of AI contribution policy.