Everyone's wringing their hands about juniors getting replaced by AI. They're worried about the wrong people.

The developer most at risk isn't the one still learning. It's the one whose entire job is turning tickets into pull requests.

The Ticket-to-PR Pipeline Has a New Operator

Ivan Turkovic wrote an essay recently that stopped me mid-scroll. His argument is simple and uncomfortable: mid-level engineers are the real casualties of AI, not juniors.

Think about what a typical mid-level engineer does day to day. They take a well-scoped ticket, understand the codebase patterns, and produce working code that follows those patterns. They're reliable. They're consistent. They're exactly the profile AI is getting scary good at replicating.

Juniors, on the other hand, are still in the struggle phase. They're building mental models. They're learning why things work, not just how to ship them. You can't automate the process of becoming a better thinker. You can absolutely automate "convert this Jira ticket to a PR that passes CI."

Why This Isn't a Death Sentence for the Industry

Here's where it gets interesting. Turkovic cites Jevons paradox — the idea that when something becomes cheaper to produce, demand for it doesn't shrink. It explodes.

When steam engines got more efficient, the world didn't use less coal. It used way more. If AI makes software cheaper to build, companies won't hire fewer developers. They'll want to build more software. Total developer employment might actually rise 🤯

But the composition of that employment shifts. The roles that survive look different from the roles that existed before.

The Builder Archetype

The survivor in this story isn't the fastest coder. It's the person Turkovic calls the "builder" — someone with taste and judgment.

→ Builders decide what to build, not just how to build it.
→ Builders smell a bad abstraction before it ships.
→ Builders know when the AI-generated code is subtly wrong in ways that won't show up until production.
→ Builders make decisions that compound over months, not just close tickets that expire in sprints.

This is the part that's hard to hear if you're mid-career and comfortable. The skills that got you promoted from junior to mid — consistency, pattern-following, reliable output — are exactly the skills that AI replicates best.

The skills that protect you are fuzzier. Taste. Judgment. Knowing when to say no. Knowing when a feature request is actually three different problems wearing a trench coat 😄

The Junior Advantage Nobody Talks About

Juniors have something mid-levels often lose: they're still in learning mode. They haven't optimized for throughput yet. They're still asking "why" instead of just shipping.

That questioning mindset is actually the foundation of the builder archetype. A junior who struggles through understanding why a system is designed a certain way is building exactly the judgment muscles that AI can't replace.

It's a mid-level who hasn't asked 'why' in three years and just transcribes specs into code. That's a process, not a skill. And processes get automated.

So What Do You Actually Do?

If you're that mid-level right now and you feel a pit in your stomach, don't worry. Awareness is the first step.

→ Start caring about product outcomes, not just code output.
→ Develop opinions about architecture that go beyond "this is how we've always done it."
→ Practice saying "we shouldn't build this" with a good reason attached.
→ Use AI tools aggressively — not to coast, but to free up time for the judgment work that matters.

The goal isn't to outcode the AI. It's to outsmart it. The bar for 'I wrote functioning clean code' continues to get lower and lower. The bar for 'I made the right judgment about what to build and how to build it' hasn't moved an inch 🎯

The mid-level role is not going away. It's changing. The question is: Will you change with it, or will you find yourself standing in the same place the factory line used to be?

What's the skill you've developed in your career that you're most confident AI can't replicate?