I watched it happen on LinkedIn, between a post about morning routines and another about crypto. Some founder in a headset, eyes lit up like he’d seen God, said his company cut engineering by sixty percent. “AI does the rest now,” he wrote. The comments filled up fast people changing their headlines to “AI Consultant,” “Prompt Engineer,” like if you rename yourself fast enough, the ground can’t swallow you.
I sat there and felt nothing. Not fear. Not anger. Just the same flat recognition I felt in my second year of IT, when I walked into software engineering class and realized they’d sold me a different map than the one we were standing on.
First year, they showed us the word “generic” on a syllabus. I thought: this is it. This is where we build. I pictured screens lighting up, code compiling, something I made from nothing running in the world. Second year hit, and they gave us diagrams. UML. System design. Architecture. How to think about software before your fingers ever touched a keyboard. No code. Just theory. Endless theory. I remember sitting there feeling cheated, like I’d paid for a concert and gotten a lecture on music theory instead.
I didn’t know it then, but that was the first time the map shifted under me. I just didn’t have the language for it yet.
The numbers are what they are. I don’t need to quote them like a report. You can feel them if you’re looking. Junior roles are vanishing. The under-twenty-five crowd in development is down twenty percent from where it was three years ago. Fresh graduates with CS degrees are facing unemployment higher than the national average. Companies aren’t hiding it anymore—Block cut half their engineering team and the CEO said it straight: AI tools changed what it means to build a company. The seniors are getting promoted faster, paid more, given titles like “AI Engineering Lead.” The juniors are getting ghosted.
And nobody asks the quiet part out loud: where do the next seniors come from when nobody’s hiring the people who need to become them?
I know what the research says because I’ve read it, same as anyone else who’s trying to understand what just happened to their profession. AI writes code four times more likely to be duplicated. Forty-five percent of it still carries OWASP vulnerabilities. The best models the ones with the billion-dollar training budgets produce secure code barely over half the time. When researchers actually scanned real “vibe-coded” apps out in the wild, they found thousands of vulnerabilities, hundreds of exposed secrets, systems that looked professional until you poked them. The security tools are getting better, sure. But they’re cleaning up after a party nobody invited them to. They’re finding fractures in foundations that were never laid.
The experts know this, even if they don’t say it in the press releases. The World Economic Forum says AI will create more jobs than it kills but not the same jobs, and not for the same people. Gartner warns that companies leaning too hard on AI-generated code without human oversight are stacking technical debt they’ll pay for later. McKinsey talks about trillions in economic value but calls the transition “disruptive” which is what a consultant says when they mean some people are going to be destroyed. And the developers themselves? Seventy-five percent manually review every AI-generated snippet. Only twenty-nine percent trust it. Seventy percent of what AI writes gets rejected. The machine is fast. The machine is confident. The machine does not understand.
I think about that hospital management system example sometimes. Someone asks an AI to build it from scratch, just types it in like a prayer. The AI gives back schemas, APIs, a login page that looks clean and professional. It does not give you the knowledge that a hospital system has to survive regulatory audits, load spikes, human error at three in the morning, the slow erosion of time. It does not give you the weight of knowing why this system has to fail this way and not that way. It gives you the how. It has never once given anyone the why.
That is the part the map doesn’t show anymore. The intuition. The scar tissue. The ten years of sitting in rooms where bad decisions were dissected in real time, of breaking production and learning what it costs, of carrying the memory of failures that shaped your judgment. AI doesn’t have those nights. And if we stop letting juniors have them, we stop making the people who can tell when the machine is lying.
I write less raw code now than I used to. I don’t think that’s surrender. I think it’s clarity. The keyboard was never the job. The job was knowing what to do when the keyboard isn’t enough. The build process. The deployment pipeline. The difference between a commit and something that actually runs in the world. That boring, unglamorous knowledge that turns a person who types into a person who keeps systems alive.
The war ended quietly. No treaty. No surrender. One day the map just looked different, and most people were too busy updating their headlines to notice we’d moved. I’m not dead. I’m not even angry. I’m just here, on the new ground, recognizing that I was trained for this long before I knew I’d need to be. The borders changed. The work didn’t disappear. It just stopped wearing the costume we sold it.
And in ten years, when the current generation of seniors is gone, when the people who learned by breaking things have retired, when the judgment that comes from scars has left the building someone is going to need the ones who sat through the theory. The ones who know why the data flows this way. The ones who understand that software was never about the code. It was about the thinking before the code, and the blood after it breaks, and the wisdom that grows in the space between.
