You know that moment? When a teammate casually says, “Already asked the agent to do it,” and you realize you’re the only one still typing actual code? If that thought has kept you up at night — this is for you.
You haven’t fallen behind. The profession split. And the side you’re on has a future most people can’t see yet.
Let’s be honest: managing AI agents sounds like a promotion to some. To you, it sounds like babysitting. You’d rather quit. There is nothing wrong with you. There is something wrong with the picture of “the future of work” you’ve been sold.
The middle ground is disappearing. The old world is being pulled apart into two extremes: fast, chaotic agent-supervision, and slow, compliance-heavy, deeply human coding.
You’ve probably seen the job posts. “AI Agent Manager.” “Prompt Engineer.” “LLM Orchestrator.” And you’ve thought: this can’t be the career I trained for. You’re right. It isn’t. It’s a different career wearing your job title.
Meanwhile, on Hacker News, a developer asks the exact question you’re afraid to ask out loud: “Anyone still coding like 2021? Where do you work?”
Here’s what comes back. A developer on defense and classified projects says they’re still old school. Another building U.S. medical products says their processes change slowly, legal reviews gate everything, and not a single AI tool has made it through. A third notes that every so often there’s a skeptical, independently-thinking CTO at a “boring” enterprise quietly ignoring the hype.
That’s not nostalgia. That’s a map.
AI won’t replace you where the cost of being wrong is a human being.
Think about it. Medical devices. Aircraft controls. Financial audit systems. Legal compliance platforms. In those worlds, the cost of a hallucinated line of code isn’t a failed demo — it’s a lawsuit, a recall, a corpse. Regulators don’t accept “the agent suggested it.” Liability is a feature, and it’s the moat that keeps you employed.
So take a side: stop apologizing for writing code the old way. Stop treating your resistance as a personal failure. The real problem is that you’re looking for safety in the loudest, least stable corner of the industry — and ignoring the unsexy sectors that will still be hiring humans to write and review every single line, for decades.
Defense. Medical. Legal. Industrial control. Insurance. Aerospace. These industries don’t move at the speed of AI adoption. They move at the speed of consequences. Your patience, your careful craftsmanship, your willingness to read a full spec before touching a keyboard — those aren’t weaknesses. They’re requirements.
Your skills didn’t become obsolete. The tech companies you worked for did.
This is the twist everyone gets wrong. The safe place for a traditional, hands-on coder isn’t a startup that proudly refuses AI. It’s not some reprogrammed retreat. It’s the heavily regulated, deeply “boring” company down the road that has a skeptical CTO and a legal team with real veto power. That’s where your knowledge of mutexes and memory management and test coverage is literally irreplaceable.
One engineer in the HN thread put it plainly: their company hasn’t changed much on the AI front, and they’re pushing hard to change that. But pushing hard doesn’t matter. Legal review matters. Compliance matters. Regulatory approval matters. Those processes don’t fall to a vibe. They fall to people who know exactly what a line of code does — and can defend it in front of a board, or a court.
So no, you are not the only one coding like it’s 2021. You’re just not looking in the right place. Stop measuring yourself against the AI-obsessed tech bubble. Start measuring yourself against the industries where innovation is kept on a leash.
The future of hand-written software isn’t dead. It’s just boring, well-paid, and structurally protected.
Stop asking where the old-school coders went. They’re where the rules are. And if you want to keep your hands on the keyboard, that’s exactly where you need to be.
FAQ
Q: Isn't this just a few nostalgic anecdotes from Hacker News?
A: The comments point to a structural pattern, not nostalgia: industries with regulatory liability, legal review, and physical consequences cannot hand decision-making to probabilistic models. That's why defense and medical still run on human-reviewed code. It's not about feelings; it's about who gets blamed when something fails.
Q: What should I actually do if I want to keep coding the old way?
A: Stop applying at AI-hyped consumer tech companies. Look for roles in defense, medical devices, industrial automation, legal tech, and enterprise compliance. In interviews, emphasize your experience with safety-critical systems, code review rigor, and comfort with slow, documented processes. Those sectors are actively looking for people who can defend every line.
Q: Isn't the contrarian take that 'boring' companies will also adopt AI soon?
A: Some will, but selectively. A skeptical CTO isn't necessarily anti-AI — they're anti-liability. They'll adopt tools for research, testing, and documentation long before letting an agent write production code that requires regulatory sign-off. That makes your traditional coding skills even more valuable as the human firewall between AI output and real-world consequences.