“Where is the programming profession going?” — 150 engineers answer, and they don't agree
An Ask HN about AI's effect on the craft drew the people living it. The recurring split: is the senior engineer's judgment the durable moat, or is everyone just a more efficient middleman for a vendor that can undercut them tomorrow?
One camp says nothing built with an AI exceeds what its operator could have built alone — the CS background is what shapes the prompt, and “knowing Dijkstra by heart” matters less than architecture. Another says the comparison was settled years ago: outsourcing to an LLM is the same trust trap as outsourcing to a dev shop.
A developer in China reframes the cost worry entirely: a $4,699 box already runs a smarter model locally, and a new state-of-the-art lands every two months.
The question the people in the trenches actually argued
The Ask HN cut past hype to a working-engineer question, and the answers divided along a clean fault line: what, exactly, is the durable skill?
The optimist-but-grounded position held that the human still sets the ceiling. One developer offered an axiom from months of heavy use: “there is nothing I can build with Claude that I could not build myself … In my hands it can go as far as I could anyway, and no further.” Another sharpened it into where the value moved:
“There's a huge yawning chasm between the person who says 'Claude make me a $thing' versus the person who … lays down the overall architecture, gives thought to libraries, dependencies, performance trade-offs … Knowing how to implement Dijkstra or a linked list by heart is no longer important. Actual software engineering skills are more important.”
“We've seen this movie — it was called outsourcing”
The most deflating reframe came from a veteran who'd watched the same dynamic play out before LLMs existed, with offshore dev shops:
“there is no fundamental difference between outsourcing a development task to other human developers versus outsourcing it to LLMs. Either way, total and complete understanding is being sacrificed in the name of productivity.”
The skeptics weren't romantics either. One reader said he's “continuously shocked” at people handing critical work to agents “after only a few months of existing,” when the output is still “laughably wrong on a pretty regular basis … in subtle ways.”
The cost panic, answered from the other side of the world
A thread-within-the-thread worried that frontier pricing is unsustainable and the rug will get pulled. A developer in China dismissed the framing with a price tag:
“Two years ago, SOTA was gpt-o1 … Now, for $4,699, you can easily run a much smarter Qwen3.6-35B locally with DGX Spark. … This is an era where a new SOTA arrives every two months.”
The deepest worry wasn't replacement at all — it was capture. One reader asked the question that lingers after the thread: when the value you add collapses to the dollars you pay a lab, “What's stopping the AI provider from coming in and eating your lunch?” That, more than any benchmark, is the profession's open question.
Get breakdowns like this every day
The top Hacker News stories, unpacked — by email, Telegram, or RSS.
Why an LLM feels more tiring than a keyboard — and why introverts in the thread disagree
A good tool disappears into your hands; an LLM makes you negotiate, persuade, and occasionally get angry at it. Ohad Ravid calls that the “social tax” — the cost of a conversation that never reciprocates. The thread split on whether that tax is real or a feature.