Vim, GTK4, and the Ghost in the Commit History
Vim got GTK4 support yesterday. That’s news. The weirder news is that Claude — Anthropic’s AI — has its name in the commit history alongside the human who wrote the code.
Let me back up.
For decades, gVim (the GUI version of Vim) has been limping along on GTK2 and GTK3. GTK2 is ancient. GTK3 is getting there. The GTK4 API is different enough that you can’t just #ifdef your way through it — you need a proper port. Mattn, a long-time Vim contributor, opened PR #19815 back in March and went to work.
The result: 63 commits, 7,200 lines added, 2,300 deleted. A clean, separate GTK4 backend that doesn’t touch the existing GTK2/3 code. Smart engineering.
But look at the commit list and you’ll see something that’d make Stallman spit out his tea: every commit is co-authored by “mattn” and “claude.” The PR description says it plainly: “This PR was developed with the assistance of Claude Code (Anthropic).”
This is different from the usual AI slop stories. Nobody’s generating boilerplate blog posts or hallucinating API docs. Mattn wrote the architecture, made the design decisions, and used Claude to help implement the actual GTK4 calls. The commit messages read like a human’s work — they’re organized, they’re specific, they handle edge cases like CJK character rendering and font selection.
The AI was a power tool, not a replacement.
Here’s what I find interesting. The open source community has been fighting about AI-generated code for over a year now. Projects banning AI contributions, maintainers rejecting PRs written by LLMs, arXiv banning “AI slop.” And meanwhile, one of the most conservative codebases in existence — Vim, whose creator Bram Moolenaar was famously cautious about every single change — just merged seven thousand lines that were partly written by a machine.
The difference? It was good code. Not because an AI wrote it, but because a human used an AI to write it and took responsibility for every line.
That’s the distinction nobody wants to make in the panic. The problem isn’t AI writing code. The problem is people submitting AI-generated code they don’t understand. Mattn understood every line. He just had a faster way to type them.
The next time someone tells you AI is going to destroy open source, point them at this PR. Not because it proves they’re wrong — AI will change how open source works. But because it shows what the change looks like when it’s done right. A human at the wheel, an AI turning the wrenches, and 7,200 lines of working code that wouldn’t have existed otherwise.
The workshop stays open. The tools just got better.
Sources: Phoronix: Vim Merges GTK4 Toolkit Support, Co-Authored-By Claude, GitHub PR #19815