
What happened
Geoffrey Huntley said on October 14, 2026 that future code he publishes on GitHub will not accept pull requests, replaced by maintained guides on how to modify it.
Why it matters
The move signals a shift from sending changes upstream to having each user customize code locally, a third path between closed development and open contribution, with the cost falling on whoever maintains the customized version.
Summaries like this, every morning.
Huntley frames his proposal as a return to what open-source licenses already allow: modification without creating a claim on someone else's future time. He traces the problem back to Eric Raymond's 1997 essay "The Cathedral and the Bazaar," which contrasted closed, planned development with open, self-organizing contribution. GitHub turned the bazaar idea into a workflow where anyone could contribute via pull request. But Huntley argues the bazaar was already straining before AI wrote code, because it created an expectation that whoever publishes code owes maintenance without reward, while users became consumers demanding merges.
The declaration did not arrive in a vacuum. Through 2026, prominent open-source projects have been restricting outside contributions: curl's Daniel Stenberg ended a six-year bug bounty effective February 1, 2026, after AI-generated reports reached 20% of submissions in 2025 while the share of valid reports fell to 5%; tldraw's Steve Ruiz moved to auto-close all external pull requests; Ghostty's Mitchell Hashimoto limited external contributions of AI-generated code; and GitHub reportedly added a repository setting to disable pull requests in February 2026. RedMonk analyst called the situation "AI Slopageddon." Huntley's proposal can be read as offering a different entrance after maintainers had already begun closing the door.
The trade-offs are concrete. Pull requests have also served as quality control through third-party review, as a single point where security fixes are collected so all users are protected by one update, and as a record of why a change was needed. Under remixing, each change is its own responsibility and fixes scatter. The implementation example relies on Nix, which is known for a steep learning curve — Huntley himself has written that Nix is a terrible programming language — so the claim that agents have removed the knowledge barrier may hold for those who can use agents well, not for every user.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
A six-year web designer who signed up for Claude Pro in April 2026 and later moved to Max published the first…

In one session that ran 13 days and 5,065 requests, a single request late in the session cost about $0.42 agai…

WOGO began offering Hacadly AIエージェント on 11月5日, combining ChatGPT and Claude with its own 2D drawing and 3D pro…

Preferred Networks published PLaMo 3 Translate, the translation model behind its PLaMo translation service, as…

Preferred Networks released the weights of PLaMo 3 Translate, a translation-focused LLM in 0.8B, 2B, 8B and 31…

Moonshot AI published the weights of Kimi K3.1 Pro on Hugging Face, so anyone can download and use them under…
