The OP argues that HN's persistent anti-AI sentiment misses the point: end users judge products by whether they work, not by how the code was produced. From this product-manager framing, AI-written code that ships a functional feature is indistinguishable from hand-written code that does the same, so the constant complaining feels like craftsman gatekeeping.
The editorial argues AI collapsed the cost of producing code while leaving the cost of reviewing it untouched, and that asymmetry is the entire fight. HN's median commenter is a senior IC whose actual job is reading code others wrote and deciding whether to trust it, so AI-generated 400-line diffs land on exactly the people whose workload it inflates rather than reduces.
The editorial points out that a meta-question asking 'why does HN keep complaining?' out-performed most of the day's actual technical submissions and generated hundreds of thoughtful, mostly-negative replies. The fact that this is the fourth such 'genuine question' thread in six months — each producing the same predictable shape of disagreement — suggests the community's position is stable and reasoned, not a passing mood.
An Ask HN thread titled *'Why is the HN crowd so anti-AI?'* hit 96 points on the front page. The OP's complaint is straightforward: every day the Best-of-HN RSS surfaces another post about AI writing bad code, introducing bugs, or piling on technical debt. The OP's counter-argument is the standard product-manager move — *'code is just a means to an end. Users don't care whether the code was written by AI or by hand.'*
The thread is doing exactly what the OP says doesn't happen: it's generating a long, thoughtful, mostly-negative response. By the time you read this it will have a few hundred comments, most of them disagreeing, and a non-trivial chunk of them disagreeing with each other. The fact that the meta-question — 'why does this community keep complaining?' — outdrew most of the day's actual technical submissions is itself the answer.
This isn't the first such thread. It's roughly the fourth in six months, each one phrased as a *'genuine question'* and each one collecting the same shape of replies. The pattern is stable enough now that you can predict the top-voted comment before you click.
The OP's framing is the tell. *'Code is just a means to an end'* is true if you are the person shipping. It is false if you are the person reviewing, on-calling, or inheriting. HN's median commenter — based on every reader survey the site has ever produced — is a senior IC at a company old enough to have a legacy codebase. That person's day job is not writing code. It is reading code other people wrote and deciding whether it can be merged, deployed, or trusted.
AI didn't change the cost of writing code. It collapsed it. What it didn't change is the cost of reviewing code, and that asymmetry is the entire fight. When the marginal cost of producing a 400-line diff drops to zero, the marginal cost of *evaluating* a 400-line diff stays exactly where it was — bounded by a human brain reading at human speed. The reviewer's queue grows. The reviewer gets grumpy. The reviewer posts on HN.
This is the same dynamic we covered nine hours ago in the open-source maintainer story: free PRs, expensive reviews, drowning maintainers. The Ask HN thread is the in-house version of the same complaint. The economics scale the same way whether you're a Postgres committer or a senior backend engineer at a 200-person SaaS.
There's a second layer the OP misses. The 'users don't care how the code was written' argument is correct for a six-month time horizon and catastrophically wrong for a five-year one. Users don't care, right up until the moment the system breaks in a way that requires understanding why a particular abstraction was chosen, and the only person who could answer that question was a model whose weights have since been deprecated. The HN crowd is disproportionately the cohort that has lived through one or two of these arcs already — outsourced code from the 2000s, microservices-everything from the 2010s, the hand-rolled CRDT layer your last startup left behind. The skepticism isn't aesthetic. It's pattern-matched.
The third layer is signal-to-noise on the platform itself. HN's voting system rewards posts that are surprising or actionable. *'AI tool writes good code'* stopped being surprising sometime in 2024. *'AI tool wrote code that looked good and silently broke production'* is still surprising and still actionable, so it keeps surfacing. The feed isn't anti-AI. The feed is anti-boring, and competence isn't news.
If you're the OP — the person who genuinely cannot understand the resistance — the practical move is to stop arguing about whether AI writes good code and start measuring the review-side cost. Track time-to-merge, revert rate, and post-merge incident attribution on AI-authored diffs separately from human ones for thirty days. If the numbers come back clean, you have data to bring to the next thread. If they don't, you have the answer to your own question.
If you're the reviewer — the person quietly nodding along to every 'AI code is technical debt' post — the move is to stop letting AI authorship be the variable you flag. The variable that actually matters is whether the author can defend the diff in review. A human who can't explain their own code is the same problem as an AI-assisted human who can't explain the model's code. Make the policy about explanation, not provenance. Otherwise you end up gatekeeping based on tooling, which is a losing fight on a long enough timeline.
For engineering managers, the actionable read is staffing. The bottleneck in an AI-heavy team is not how fast features ship. It's how fast reviews clear. If you've cut headcount on the assumption that AI productivity gains net out, you've almost certainly under-staffed code review. The teams that look healthy six months in will be the ones that explicitly budgeted reviewer hours per AI-generated diff and treated that budget as non-negotiable.
The Ask HN thread will hit the front page again in roughly seven weeks — that's been the cadence — and the comments will look identical. The interesting question is when one of the AI-coding vendors decides to ship a tool that targets the *review* side rather than the authoring side: a model that reads a diff, asks the questions a senior reviewer would ask, and flags the abstractions that don't survive contact with the rest of the codebase. When that ships and works, the HN sentiment shifts overnight. Until then, the complaints aren't going away, because the underlying asymmetry isn't going away — and the people doing the complaining are the people the platform was built for.
Genuine question.<p>Over the past six months, there hasn’t been a single day where I’ve checked the HN Best RSS feed without seeing a post about how AI “writes bad code,” “introduces bugs,” “creates t
→ read on Hacker NewsBecause I enjoy writing code. I enjoy being paid for writing code. And I don't enjoy writing prompts for AI.Code is not just a means to an end. Code is a means to my happiness. Users might not care, but I do care. I love good code. I feel great when I can write good code.I won't say that I
I call these AI tools "proprietary non-determenistic database of the free internet". They belong to american companies which can cut off your access if american government doesn't like your country's government. They fed from the free internet that many of us grew up in, store it
I actually hold both extremes inside of me simultaneously. The speed at which you can ship when you have a strong vision of the end product and the architecture is extraordinary (the part of me that loves AI-assistance).The journey itself, at least for me, has been absolutely grueling though; I'
The very premise of your question is very questionable.> Over the past six months, there hasn’t been a single day where I’ve checked the HN Best RSS feed without seeing a post about how AI “writes bad code,” “introduces bugs,” “creates technical debt,” or something along those lines.At the same t
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
It's simply divided. With every such division A vs. B, the A team thinks HN is anti-A and the B team thinks it's anti-B. This is an invariant.You can see from the following megathread, currently on the front page, that HN is by no means anti-AI:Ask HN: What was your "oh shit" mom