After one of my lectures, a student stayed behind.
"Can I ask you something?"
"Sure."
"Do you think anyone will even hire me in five years?"
I expected a question about exams, maybe about which framework to learn next. Instead, he asked whether learning software engineering was already becoming obsolete.
That question stuck with me. Not because I had no answer, but because I realized the honest answer isn't "AI won't replace developers." That's too easy, and it isn't really what he was asking.
The fear is rational
I didn't want to dismiss it, and I don't think you should either. Look at what models can do today:
- build applications
- write tests
- explain algorithms
- review code
- generate documentation
If you're a student watching that happen in real time, it's a completely reasonable thing to wonder: what value do I actually bring?
The turn
Here's where I landed, eventually:
AI doesn't make the question "will anyone hire junior developers" obsolete. It changes what junior developers need from senior developers.
Not less guidance. Better guidance.
Before AI, juniors mostly asked questions like "how do I write this function?" Today, AI answers that in seconds, and reasonably well.
But AI rarely answers:
- Why are we solving this problem in the first place?
- Why is this architecture chosen over the obvious alternative?
- Why is this trade-off acceptable here, but wouldn't be somewhere else?
- Why did the team reject the "better" solution?
Those aren't coding questions. They're engineering questions. And they're exactly the conversations senior engineers should be spending more time having.
AI removes some of the teaching. It leaves the mentoring.
What we actually did
Not long after that conversation, we hired a junior developer.
Not because AI had failed us. If anything, the opposite: we use AI massively across the team, in code, in reviews, in day-to-day workflows, and we keep investing further in it. There's no ambivalence there.
And we still hired a junior. Not despite that investment, but alongside it. The more capable our tools get, the more clearly it shows what they don't teach: architectural intuition, production experience, judgment. Nobody starts out with those. Someone has to invest in that journey deliberately, or in ten years there won't be any seniors left to have these conversations at all.
AI can answer questions. A mentor teaches you which questions matter.
AI can explain a design pattern. It can't tell you why your team shouldn't use it. AI is excellent at transferring knowledge. Experience is about transferring judgment.
That last distinction is the one I keep coming back to.
Knowledge is cheap. Judgment isn't.
Most conversations about AI and engineering careers are really conversations about knowledge: who has it, who needs it, whether it's still worth acquiring when a model can produce it on demand. And it's true: knowledge is becoming cheap. Syntax, APIs, boilerplate, even reasonably good first drafts of a solution are all cheap now, in a way they weren't five years ago.
Judgment isn't cheap. Knowing which problem is worth solving. Knowing when the "correct" solution is wrong for this team, this deadline, this system. Knowing which trade-off will hurt in six months versus which one you can live with. That's not downloaded. It's developed, usually slowly, usually by making calls that turn out to be wrong, ideally with someone more experienced around to catch the expensive mistakes before they ship.
That's what mentorship is actually for. It was never really about teaching syntax.
Back to the student
If that student asked me again today, my answer would still be yes.
Not because I believe AI will stop improving. Quite the opposite: the better AI gets, the more important human judgment becomes, not less. Judgment isn't something you can prompt your way into. It's built over time, usually with the help of people willing to invest in juniors before they're seniors.
AI isn't making junior developers obsolete. It's changing what they need to learn, and what senior developers need to teach.