Debian's LLM Ban Debate and the 493-Comment Grudge Match on HN

Everyone keeps saying AI coding assistants are inevitable, that resistance is just nostalgia for a slower, dumber era of software. Fogus's blog post Born Against says otherwise, and 415 upvotes plus nearly 500 comments on Hacker News suggest a lot of people needed to hear it.

The argument isn't complicated, which is probably why it landed so hard. Hobby programming communities aren't rejecting LLMs because the tools don't work. They're rejecting them because the point was never the code. The point was the struggle.

That line is going to get quoted for years, mostly by people trying to win arguments on Discord servers. But it deserves the attention. It reframes the whole debate. This was never really about whether GPT-5-whatever can write a working sorting algorithm. Of course it can. It's about whether you writing that sorting algorithm badly, six times, cursing at your terminal, is actually the valuable part.

The Elephant in the Room

Here's what nobody wants to say out loud: this isn't just a hobbyist quarrel happening in some cozy IRC channel nobody important reads. Debian — yes, the operating system running who-knows-how-many servers on the planet — is actively debating a General Resolution on LLM usage. One proposal on the table would flatly prohibit LLM-assisted contributions from inclusion in Debian, full stop.

That's not culture-war theater. That's governance. That's a major infrastructure project deciding whether AI-generated patches are even reviewable in a way that matters for a codebase people will be running in production a decade from now.

And that's the actual tension buried under all the Hacker News snark: this isn't purely a nostalgia argument. It's a trust and provenance argument. If a contributor can't explain their own patch because a model wrote it, who maintains it in five years when the model that generated it is deprecated and the contributor never actually understood the logic?

The pro-LLM crowd isn't wrong that these tools accelerate boilerplate, debugging, and onboarding. Nobody's arguing GitHub Copilot should be illegal. But there's a difference between using a tool selectively and outsourcing the competence itself. Debian's proposal isn't anti-progress. It's risk management dressed up as ideology.

Where This Actually Lands

• Hobbyists lose nothing by writing code manually — that's literally the hobby.
• Open-source maintainers gain a real headache: how do you review code the submitter doesn't fully understand?
• Companies building dev tools should notice: communities built on trust and craftsmanship are exactly the ones most likely to push back hard, and they're influential ones.

The critics calling this gatekeeping aren't entirely wrong either. Rejecting a useful tool out of purity can curdle into elitism fast — ask anyone who's been told their Stack Overflow-copied code doesn't count as "real" programming.

But Fogus's piece isn't really elitist. It's more honest than that. It says: if the goal is mastery, the shortcut is the failure mode. If the goal is shipping a feature by Friday, that's a different conversation entirely — and most of this debate happens because people keep pretending it's the same one.