Is Software Engineering Dead? Both Camps Are Wrong

Is software engineering dead: magazine cutouts of a laptop with a dark screen, a mechanical keyboard and a coffee mug on a green cutting mat

TLDR: Is software engineering dead? No, but the job you learned is. AI writes most of the code I ship now, and my day is mostly reviewing and fixing what it wrote. Typing code, digging through Stack Overflow, and the easy tickets juniors learned on are gone for good. What’s left, and worth more than ever, is knowing the domain, designing for what comes next, and being able to tell when code that looks right is wrong.

AI writes most of the code I ship now. My days are also more tiring than when I wrote it all myself. Both of those things are true, and that’s exactly why the whole “is software engineering dead” argument drives me nuts.

I’ve been writing software for over ten years, mostly iOS. I’ve watched a lot of “this changes everything” moments come and go. This one is real. It just isn’t the one either side keeps describing.

Is Software Engineering Dead? Two Camps, Both Wrong

Camp one says it’s over. AI writes the code, so why pay someone to write the code? Learn a trade, the robots have the keyboard now.

Camp two says it’s hype. Fancy autocomplete, a bubble, nothing real changes, and everyone will be back to writing code by hand once the funding dries up.

The data doesn’t back either camp cleanly. The 2025 Stack Overflow Developer Survey found 84% of developers use or plan to use AI tools, and 46% don’t trust the accuracy of what those tools produce. Almost everyone is using it. Almost half don’t believe it. That’s not a dead profession or a fad. That’s a profession in the middle of a messy rewrite.

And my favorite stat: in a randomized trial by METR, experienced open source developers took 19% longer on real tasks when they were allowed to use AI, but afterwards they estimated AI had made them about 20% faster. METR has since said its follow-up study got muddied because developers started refusing to work without AI at all. Which, honestly? That tells you more than the speed number does.

What Actually Died (the Doomers Are Right About This Part)

Let me give the doomers their due, because they’re not making it up. A bunch of what I did five or ten years ago is gone, and it isn’t coming back:

  • Typing code as the job. Being valuable because you can crank out a lot of code by hand. That skill on its own is worth a lot less than it used to be.
  • Stack Overflow digging. The hours spent hunting for the right syntax, the right forum answer, the one doc page that explains the weird flag. Gone.
  • The easy entry-level work. The simple tickets juniors used to learn on now go to AI first. I wrote a whole post on why AI replacing junior developers is the wrong question, but the tickets really did move.
  • Specialist gatekeeping. “I’m an iOS developer” used to mean I stayed in my lane. With AI, crossing into web or backend work stopped being scary.

If your identity as an engineer was built on any of those four things, I get why it feels like the end. For that version of the job, it kind of is.

What My Day Looks Like Now That AI Writes Most of My Code

Magazine cutouts of a magnifying glass, a closed laptop and a red pen on a green cutting mat, standing in for reviewing AI written code

Here’s what nobody warned me about. When AI writes most of your code, you don’t stop working. You become a full time code reviewer.

My day is mostly reading what the AI wrote, catching the stuff that’s almost right, and fixing it. Almost right is the dangerous part. Wrong code breaks loudly. Almost right code passes the build, looks clean in the diff, and quietly does the wrong thing.

And reading someone else’s code all day is more tiring than writing your own. When I wrote it, I knew why every line was there. Now I have to reverse engineer the reasoning of something that doesn’t actually have reasoning. It’s a different kind of tired.

I’ll also admit something I didn’t expect: I miss writing code. There was a flow state in building something by hand, line by line, that doesn’t really exist when you’re steering and reviewing. The output is way up. Some of the fun went with the typing.

What Didn’t Die: Knowing Why the Code Exists

Magazine cutouts of a steel building frame, a yellow hard hat and a business handshake on a green cutting mat, standing in for architecture and domain knowledge

This is the part the doomers skip. AI can write the code. It can’t tell you whether the code should exist, or whether it reflects how the business actually works.

The stuff I catch in review is rarely a typo or a crash. It’s domain knowledge and design. The AI doesn’t know the industry I work in. It doesn’t know which edge case happens every single day for real customers and which one never happens. It doesn’t know that the “simple” fix it wrote puts logic in a layer that’ll make the next three features painful. It solves the ticket in front of it. It doesn’t set the codebase up for what’s coming next.

That’s the job now. Knowing the industry. Making sure the code reflects actual business cases. Making sure it’s built so the next person (or the next AI session) isn’t fighting it six months from now.

The biggest surprise for me: I expected AI to level the playing field between experienced engineers and everyone else. It did the opposite. Years of context about how systems fail, how businesses actually operate, and what “done” really means became almost the entire value. The typing was never the hard part. We just used to spend most of our hours on it.

The Numbers Don’t Say Dead Either

If software engineering were dying, you’d expect the jobs outlook to show it. The Bureau of Labor Statistics projects employment of software developers, QA analysts and testers to grow 10% from 2025 to 2035, with about 106,100 openings a year.

I don’t treat that as gospel. Projections are guesses with a government logo on them, and a flat number can hide a lot of churn underneath. The entry-level door is narrower than it was. But “narrower door” and “dead profession” are very different claims, and the people yelling the second one are often selling a course.

Will AI Replace Software Engineers? My Honest Answer

AI replaced the typing. It didn’t replace the engineer. The engineer was always the person who understood the problem, the domain and the tradeoffs. We just used to also be the person doing all the typing.

So when someone asks me whether they should still get into software engineering, my answer is yes. But differently. Don’t just learn a language. Learn to read code critically, because that’s most of the job now. And learn a domain, an actual industry, well enough to know when correct looking code is wrong for the business. That combination is what AI can’t fake.

If you want the longer version of what a decade in this industry taught me, I wrote it up in 10 Years in Tech. This Is What I Learned. And if you’re curious how I actually split work between AI tools, here’s what I use Claude Code vs Cowork for.

Stop Asking If It’s Dead

Here’s your challenge. Next time AI writes something for you, don’t just ship it because the tests pass. Pick one piece of it and write down, in plain English, the business rule it’s supposed to enforce. Then check whether the code actually enforces it.

If it does, great. If it doesn’t, congratulations. You just did the part of the job that isn’t dead.