Why Hobby Programmers Are the Angriest People in the Room About AI Coding

Every conversation about AI coding tools eventually becomes a conversation about speed. How many hours saved, how many lines generated, how many tickets closed. Then you wander into a hobbyist forum — a Clojure mailing list, a retrocomputing subreddit, an IRC channel that’s been running since 2004 — and the mood curdles. Clojure developer fogus hit this nerve with an essay called “Born Against.” The people who don’t get paid to write code turn out to be the most hostile to LLMs writing it for them, and once you understand why, it stops looking irrational.

One caveat up front: this isn’t a snapshot of a fresh news cycle. Community chatter on this specific topic has been thin over the past month. What follows is a synthesis of an argument that’s been recurring for two years, not a real-time read on sentiment.

The Code Is a Means. Except When It Isn’t.

When you write code at work, the purpose is obvious. Solve the problem, ship the product, collect the paycheck. The code is instrumental. Anything that makes it appear faster is a straightforward win. If an LLM writes your CRUD boilerplate and your Terraform modules, that is not a loss — that is Tuesday.

Hobby programming inverts this completely. Writing the code is the point. This is why telling a weekend hacker “you know there’s a library for that” gets you a flat stare. They didn’t need a parser. They needed three hours of building a parser.

Miss this distinction and every conversation goes sideways. One side asks why you’d work so inefficiently. The other side wonders why you’re talking about efficiency at all, at a lathe, on a Saturday.

What Got Taken Was the Process, Not the Product

This is fogus’s sharpest observation. An LLM hands you the artifact. But the thing hobbyists actually value isn’t the artifact — it’s the getting-stuck-and-then-unstuck.

You know the feeling. Three hours into a debugging session, the cause finally surfaces and something clicks into place. Or you’re spelunking through documentation for one thing and stumble onto a concept you didn’t know existed. Those moments are the entire reward structure of amateur programming. The LLM deletes that stretch of road.

It’s like offering a hiker a helicopter. If summiting was the goal, the helicopter is unambiguously better. If the climb was the goal, the helicopter destroys the thing you came for. This is the same argument that plays out every time someone proposes a cable car up a trail people love walking — and hikers lose that argument roughly as often as they win it.

Open Source Traded on Knowing Who Wrote It

Here’s where open source culture gets tangled up in this. The asset in an open source project was never just the code. It was the fact that a human being was attached to that code — someone who understood it and would answer for it.

LLM-generated pull requests strain that assumption. Maintainers increasingly ask “why did you write it this way?” and get nothing back. The code looks plausible. The contributor can’t defend a single decision in it. Review cost stays exactly the same while generation cost on the other side collapses toward zero. That asymmetry is the whole problem, and it’s why projects like Curl have publicly grappled with AI-generated bug reports flooding their security triage.

Then there’s licensing. The copyright status of output from a model trained on GPL code remains a genuine grey area, and neither the US Copyright Office’s guidance nor the EU AI Act has cleanly resolved it. For a maintainer running a project for free, on nights and weekends, that’s one more legal exposure they didn’t sign up for. Nothing kills a hobby faster than needing a lawyer for it.

And Yes, It’s Also About Identity

Let’s be blunt about the part nobody wants to say out loud: some of this is ego, in the least pejorative sense of the word.

You spent years accumulating a skill. Pointer arithmetic, recursion that actually terminates, the instinct for where a memory leak is hiding. That’s not a line on a resume — it’s a piece of who you are. Watching a three-line prompt produce something comparable is a genuinely complicated feeling.

Professional developers feel this less acutely, because the company only ever cared about the output. Better tool, use the tool, move on. For the hobbyist, “the tools got better” reads as “the thing I loved is being deprecated.” Getting defensive about that is not a character flaw. It’s a reasonable response to watching your craft get commoditized in real time.

This Doesn’t Get Resolved. It Gets Sorted.

The fight won’t end in a winner. It’ll end in a split, and the split is already visible. Some projects now explicitly ban AI-generated contributions in their CONTRIBUTING.md. Others actively encourage agents and treat AI-assisted PRs as normal. Two sets of norms are growing under the same open source banner, and they’re diverging fast.

It looks like a technical dispute. It’s closer to a dispute about taste. Taste disputes don’t get settled — they get separate rooms. Film photography and digital photography. Handmade furniture and IKEA. Vinyl and streaming. Everyone eventually stops arguing and goes to their own room, and both rooms turn out to be fine.

One thing genuinely nags at me. If a new developer starts their career having skipped the stuck-and-then-unstuck part entirely, what does programming become for them? If the joy was a byproduct of the struggle, then changing how people learn may quietly change whether there’s anything to love. Nobody has run that experiment long enough to know.

Is coding the end, or the means? Every position in this argument traces back to how you answer that — and the honest answer, for most people, is that it depends on the day.