Six hours in the air, wifi against the odds, and I’m doing something I didn’t plan to do on this flight: actually reading the AI predictions I keep half hearing everywhere, instead of just nodding along to them.
Somewhere over the clouds I kept thinking about the lamplighter. The one from The Little Prince, the tiny planet with a single street lamp that gets lit and put out every minute, because the planet spins so fast a day only lasts sixty seconds. The order to light the lamp was written a long time ago, back when the planet turned slower and the order actually made sense. Nobody ever rewrote it. So the lamplighter keeps lighting a lamp nobody in the town needs anymore, exhausted, faithful, because it’s the job he was given.
That image stayed with me for the rest of the flight. I stopped reading for a while and just sat with it, letting it connect to everything else I’d been turning over about jobs and AI. Are we still following an order that quietly stopped making sense, the way he does, and just haven’t noticed yet?
Start with the version everyone’s actually scared of. Dario Amodei, who runs Anthropic, described a scenario last year that still sits with me [1]. AI cures cancer. The economy grows 10% a year. The government runs a balanced budget. And one in five working adults still can’t find a job. He wasn’t hedging either: half of entry level white collar jobs gone, unemployment between 10 and 20%, inside one to five years.
The IMF has similar numbers with less drama attached [2]. Around 40% of jobs worldwide exposed to AI in some way, 60% in advanced economies. Their Managing Director called it a tsunami hitting the labour market, speaking in Davos this year [3]. Stanford’s Digital Economy Lab has the payroll data behind that word [4]. Entry level hiring in the most exposed roles is down about 19% for workers aged 22 to 25, measured through this summer. Real data, not survey guesses.
So something is genuinely happening. I don’t want to argue that away.
But sit with the fear itself for a second, not the numbers behind it. What are people actually afraid of losing here? I don’t think it’s the job itself. I think it’s the certainty. Somewhere along the way we built an entire identity around the idea that if you learn a skill well enough, it stays yours. You own it. It’s stable. Career as bedrock.
That was probably always a story we told ourselves more than a law of nature. Everything is impermanent. Not a slogan, just how things have always worked, we just don’t like being reminded of it on a timeline we can feel happening to us personally. The scribe thought handwriting was forever. The switchboard operator thought the connection was forever. The only constant, as far as I can tell, is change, and every generation is shocked by that fact as if it’s news.
The World Economic Forum’s Future of Jobs report actually holds this tension inside one document [5]. Same survey, over 1,000 employers, 14 million workers, 55 economies: 92 million roles disappearing by 2030, and 170 million new ones appearing in their place. Net gain of 78 million. You can read that as a horror story or a birth story. It’s the same data either way.
Here’s the part I think gets missed. Most of the loudest fear is coming from an obvious place. So much of what we actually consume today, and so much of what modern work actually is, sits inside a screen: text, images, audio, video, code. When almost everything you touch for a living is digital, the model getting better at digital work stops being an abstract trend. It becomes personal, fast.
So what do I actually think happens.
New jobs get born out of the mess, not despite it. Vibe coding wasn’t a real concept two years ago. Now there are people whose entire job is fixing what vibe coding broke, vibe coding cleanup specialists, a title that didn’t exist as a sentence until very recently [6]. Prompt engineer didn’t exist three years ago. Context engineer didn’t exist a year ago, and by multiple accounts it’s already replacing prompt engineer as the standard title [7]. AI red teamer, AI trainer, forward deployed engineer: all real titles now, all invented in the last two or three years because of the exact same mess and the exact same opportunity, arriving at once. AI doesn’t just remove roles. It manufactures brand new categories of work as a side effect of being implemented badly, quickly, and everywhere at once.
Entry level roles will probably shrink first and fastest inside companies built AI-first from day one, the ones with no old infrastructure to protect. Older organisations, carrying legacy systems and legacy approval chains, will likely keep hiring juniors longer. Not out of loyalty to the past. Because business continuity means you can’t just stop and rebuild mid-flight. You keep the machine running while you build the next one. Rebuilding the systems, in the end, is the manageable part. Rebuilding the mindset inside a large organisation, the habits a few thousand people carry into work every Monday, is slower and far more tedious. That’s the lamplighter again. Still lighting a lamp the town has outgrown, because nobody has rewritten the order yet, and somebody still has to keep the lights on while they do.
History doesn’t repeat the details. It repeats the shape. The Industrial Revolution didn’t leave humanity jobless, it moved us from fields to factories. ATMs didn’t end the bank teller, they made branches cheaper to open, so banks opened more of them, with people doing different work inside. SEO as a career didn’t exist until search engines did.
Universities won’t keep pace either, and that’s not really their fault. Curriculum committees run on a multi year clock. The skills gap moves on a multi month one. So the next generation will do what every generation does when the institution is too slow: teach itself, on the job, with AI as the tutor instead of the lecture hall. I think they end up faster than we were. Not smarter. Just less starved for information than we were at their age.
None of this makes the next few years easy for everyone. Entry level, especially in digital production, is going to feel it first. I’m not going to pretend otherwise.
But zoom out far enough and this stops looking like an ending. It looks like what it has always looked like, right before we had the language to name it: more jobs than before, categories of work nobody had a name for yet, and a slightly different answer to the question of what “career” even means, one generation from now.
My grandfather told me once that if you don’t learn something every single day, you’ve lost that day of your life. I think that’s the actual answer hiding in all of this. Stay curious, keep learning, keep evolving alongside whatever comes next, and there’s genuinely nothing to worry about. The people this hurts are the ones who stop. Not the ones who keep moving.
Not AI taking work away. AI finally rewriting the order on the lamp. What we do with the extra light is still up to us.
Sources
Axios — Amodei’s warning on entry level jobs and unemployment (May 2025)
IMF Blog — Georgieva, “AI Will Transform the Global Economy” (Jan 2024)
Stanford Digital Economy Lab — “Canaries in the Coal Mine?” (August 2026 update)
World Economic Forum — Future of Jobs Report 2025
Forbes — ed “The New Job Of Being A Vibe Coding Cleanup Specialist” (Sept 2025)
The Interview Guys — “What Is a Context Engineer?” (2026)