When I started career in software in 2015, every job application I filled out had the same dropdown.
I picked frontend. Not because I didn't write backend code — I did, throughout my whole career. But the dropdown needed an answer, and frontend was where my mass-applied recruiter keywords lived. You know how it is. Required field. Pick one. Move on.
Except you don't move on. That answer followed me for a decade. The jobs I saw. The filters I passed. The ones I didn't. A dropdown became a label. A label became an identity. And identities, it turns out, are harder to kill than a production bug on a Friday deploy.
Last week I sat down with our R&D chief to talk about the future of software development. Not in the abstract LinkedIn way — in the "what does this actually mean for our team" way. I threw out two directions.
The first is horizontal. The walls between backend, frontend, product, design — they're gone. I watched a PM at our company trace a 35-second production slowdown to a cold buffer pool, calculate that 33,000 rows were being scanned when only 3,300 were needed, and open a PR with the fix. Six months ago he was the guy engineers joked about for "refusing to open PRs." Nobody's laughing now.
The second is vertical. Systems thinking. Architecture. The kind of knowledge that doesn't fit in a prompt and doesn't show up in a diff. The kind of knowledge that prevents the 3am pager from going off in the first place.
He said something that reframed both.
"ICs will always be measured on one thing — how fast they deliver quality. AI doesn't change the metric. It changes the multiplier."
Go wide → you parallelize. Ship 3 things instead of 1.
Go deep → you prevent fires. One architectural call saves 3 months of rework.
Breadth or depth. Both are multipliers. Neither is optional.
I agreed. Then I told him the real problem.
Most developers will pick neither.
A Dead Economist Walks Into a Standup
Stay with me here.
John Maynard Keynes was a British economist who published The General Theory of Employment, Interest and Money in 1936 — right in the middle of the Great Depression. Why should you care? Because before Keynes, every serious economist believed the same thing: markets self-correct. Prices fall, wages adjust, supply meets demand, everything finds equilibrium. Just wait it out. This was the "it works on my machine" of economics — theoretically sound, practically useless.
Keynes looked at millions of people in unemployment lines and said: no. Things get stuck. And they stay stuck.
His big idea was sticky prices. Wages and prices don't adjust to new realities even when they should. Contracts get in the way. Coordination problems get in the way. The simple human refusal to accept less than yesterday gets in the way. The economy doesn't self-correct. It freezes. By the time it unfreezes, the damage is done.
If you're a developer, you've already heard this voice. It sounds like: "I like writing code manually, that's what I was hired for." Or: "I'm not a prompt engineer, I'm a real engineer." Or my personal favorite: "AI can't do what I do" — said by someone mass-applying to the same role they've held for 6 years while the job description quietly changed underneath them.
That's a sticky wage. That's a price that won't adjust.
This was controversial enough to split economics into two schools still arguing about it 90 years later. Keynes told an entire profession "your core assumption is wrong" and proved it with math. Which, if you think about it, is the most senior-engineer move in the history of intellectual work.
Why am I telling you about a dead economist?
Because developer identities are sticky prices.
"I'm a frontend developer" is not a description of what you do. It's a wage that refuses to adjust. I wrote backend code my entire career, but the label said frontend. The market moved — AI dissolved the skill barriers, companies restructured around output — but the label stayed. Comfortable. Familiar. Attached to years of mass-applied recruiter keywords and mass-internalized self-worth.
And it's not just economics. Psychologist Glynis Breakwell developed Identity Process Theory in 1986, which describes exactly why this happens at the individual level. People construct identity around four principles: self-esteem, continuity, distinctiveness, and self-efficacy. Threaten any of them — say, the market telling you your specialization is no longer special — and people don't adapt. They defend. They double down. Erik Erikson, who coined the term "identity crisis", would call what's happening now a mass identity renegotiation. Most people avoid those until the pain of staying exceeds the pain of changing.
Kahneman and Tversky put a number on it: losses feel twice as painful as equivalent gains. To stop calling yourself "a React developer," the alternative has to feel twice as valuable. It doesn't. It feels like opening a codebase you've never seen at 4pm on a Thursday.
So you stay. Not because staying is the right move. Because the cost of moving feels higher than the cost of standing still.
The Market Already Moved. You Didn't.
The numbers are in. Stanford tracked payroll data: employment for developers aged 22–25 dropped 20% from its 2022 peak. Developers over 26? Stable. The two cohorts diverged the moment ChatGPT launched. Like a git branch nobody merged back.
Source: Stanford Digital Economy Lab, "Canaries in the Coal Mine?" (2025)
Harvard went wider — 62 million workers, 285,000 firms. At AI-adopting companies, junior employment fell 9–10% in six quarters. Not layoffs. Companies just stopped posting junior roles. The bottom rungs aren't being kicked out. They're being removed before anyone gets on.
You know the headline numbers. 245,000 laid off in 2025. 95,000 so far in 2026. Salesforce cut support by half, replaced them with AI agents. Snap said 40% of new code is AI-generated. But the same Q1 2026 that saw 52,000 layoffs also had 67,000 active job postings — highest in three years. Same quarter. Same industry. Companies destroying one type of role and creating another, fast enough for both curves to coexist.
The jobs disappearing are defined by labels. The jobs appearing are defined by output.
And yet — go on any job board right now. I went on Anthropic's. The company building Claude — the AI that's supposedly dissolving the boundaries between specializations. Here's what their careers page looks like:
Anthropic's careers page, April 2026. The dropdown is gone. The labels aren't.
The company building the AI that dissolves specialization boundaries is hiring with the same boundaries. Stickiness isn't a bug in the hiring pipeline. It's the pipeline.
The Twist: You Already Know This And It Doesn't Matter
Here's where it gets uncomfortable.
If you've read this far you probably agree. AI changed things. Labels are outdated. Adapt or die. You've seen the LinkedIn posts. You've nodded along. You might have even reshared one with a thoughtful comment like "this resonates."
And then you went back to work. As a frontend developer. Doing frontend things. In a frontend team. With a frontend title.
This is Breakwell's theory in action. You intellectually accept the threat. But your identity defense mechanisms kick in anyway — you minimize the threat, you tell yourself your niche is different, you reframe the layoffs as something that happens to other people. The continuity principle wins. The identity survives. The adaptation doesn't happen.
METR — a nonprofit research group — proved this gap is measurable. They ran a randomized controlled trial in 2025. 16 experienced open-source developers. Randomly assigned tasks with or without AI.
The developers using AI took 19% longer.
Before starting, they predicted a 24% speedup. After finishing — after experiencing the slowdown — they still believed AI had sped them up by 20%.
A 39-point gap between perception and reality. Not a skill problem. A belief system so sticky that lived experience couldn't dislodge it. If that's not downward nominal rigidity applied to self-perception, I don't know what is.
Keynes's whole point. Sticky prices don't persist because people are stupid. They persist because the psychological cost of adjusting is higher than the economic cost of being wrong — right up until the moment it isn't. And by that moment, you're in the LinkedIn "open to work" queue. Which is an unemployment line with better lighting.
Three Things That Break the Freeze
Keynes argued economies don't self-correct. They need intervention — something external that forces movement.
Developer identities work the same way. Three things do it:
Your company stops hiring your label. Already happening. Postings labeled "entry-level software engineer" grew 47%, but actual hiring dropped 73%. The dropdown still has the option. The jobs behind it don't.
Someone crosses the boundary you wouldn't. I watched a PM open PRs, trace production issues, and challenge engineering assumptions that two senior engineers missed. He didn't learn to code — he already knew the domain. AI just removed the last barrier between his knowledge and the codebase. When someone crosses the wall you're standing behind, it stops feeling protective. It starts feeling like a cage.
You realize the label was never a description. It was a limitation. I called myself a frontend developer for years — while writing backend code the whole time. The label didn't reflect what I did. It reflected what I was willing to claim. The gates were already open. The only thing still closed was the dropdown in my head.
The Dropdown
Somewhere right now, a developer is filling out a job application. There's a dropdown. "Select your specialization." They're picking the same thing they picked last time. Not because it's wrong. Because it's familiar. Because the cost of changing feels higher than the cost of staying.
Keynes called this sticky prices. Kahneman called it loss aversion. Breakwell called it identity threat defense. I'm calling it sticky identities.
The biggest risk right now isn't picking the wrong direction — breadth or depth, horizontal or vertical. The biggest risk is not picking at all. Inertia is the default. And the data says the default is getting people left behind.
I picked frontend from a dropdown 11 years ago. It took me most of those years to realize I wasn't choosing a specialty. I was choosing a ceiling.
The dropdown is still there. I stopped picking from it.
Top comments (0)