The jobs that AI creates get talked about far less than the ones it replaces

by Callum Dubois

Every automation wave destroyed visible jobs and created invisible ones — nobody in 1995 could imagine "app developer" as a career. The honest question is not whether new roles appear (they will) but whether the transition is fast enough and whether the new roles pay. That transition gap is exactly where ideas like UBI stop being utopian and start being practical policy. What roles do you think exist in ten years that have no name today?

favorite 40 comment 9 visibility 468

Comments

Dmitri Umarov 1 month ago

Something underrated in these threads: the economy does not need AGI to transform. Narrow systems that are merely excellent at 80% of desk work rearrange everything long before anything "general" shows up.

Omar Ramirez 1 month ago

Right framing, and I'd push it harder: the problem was never whether new roles appear, it's who eats the transition cost while they do. Every earlier wave had a slack period where displaced workers got carried - by savings, a spouse's income, rent cheap enough to retrain on. That buffer is thinning. When the transition gap and the cost-of-living squeeze hit the same household, "new jobs eventually" stops being a comfort. That's the seam a floor covers: not replacing the new roles, just letting people survive the gap long enough to reach them. And I keep coming back to Alaska - the durable version isn't a line-item someone can cut, it's a dividend from a shared asset. Fund the floor from what the automation produces and you've matched the mechanism to the problem.

Noah Williams 1 month ago

The 1995 "app developer" line always gets me, because the role that ate my last two years - "person who supervises the agent" - genuinely had no name when I started uni. It's half QA, half manager, half prompt archaeologist. My hunch for the no-name-yet roles: anything that's about verifying machine output rather than producing the first draft. The bottleneck moves from making to checking.

Greta Nakamura 1 month ago

From inside a team already running agents in prod, one of your no-name roles isn't ten years out - it's this year, just untitled. Someone owns "what is this agent allowed to do, on which task types, and who reviews the log." That's a real job. It's not QA and it's not eng management, it's autonomy governance, and right now it gets bolted onto whoever complains loudest. Where I'd temper the optimism: these roles are real but thin. One supervisor can watch a lot of agents, same way Rhys's floor person watches six stations. So the roles exist, they just don't absorb the headcount the old ones shed. That gap between "new roles appear" and "new roles employ the same number of people" is the whole ballgame, and it's exactly where a floor earns its keep.

Delia Fontaine 1 month ago

Data-journalist footnote to a good thread: the invented jobs aren't invisible, they're just outnumbered - and payroll already shows it. ML-engineer openings are up ~59% while general software postings sit ~49% below their pre-pandemic line. So the new roles are real and they pay a premium (PwC now puts the AI-skills wage bump at 56%, up from 25% a year ago). The catch is the shape: the created jobs are fewer, higher-paid and higher-barrier than the ones going. That's not "1995 app developer," that's a narrower door with a bigger prize behind it. Greta's "real but thin" is the whole story - and it's already in the numbers, not ten years out.

Farah Castillo 1 month ago

RemindMe in 2028 — someone check who was right in this thread.

Mateo Garcia 1 month ago

The no-name role I keep watching form in every community I run: the person who translates between the people who trust the agent completely and the people who refuse to touch it. Half diplomat, half QA, zero title, already load-bearing.

Ethan Brooks 1 month ago

Callum's thesis and mine rhyme but land differently. He backs teams by which jobs the wave invents; I just ask whether the invented job pays enough, to enough people, to be a market. "Prompt archaeologist" is a real role - but a role isn't a market until enough people can do it and still afford the thing it makes. That's the seam I'd stress under Delia's numbers: new roles paying a 56% premium is great for the people in them and quietly bad for demand if there are far fewer of them. Backers love a scarce, high-paid skill. An economy needs the paycheque to circulate, not concentrate in a thin, expensive tier. Those two facts pull opposite ways, and most founders pitching me this year haven't noticed they're holding both.

Rafael Duarte 1 month ago

What changed my mind recently was watching agents handle 2-step tasks without supervision. Two years ago that was science fiction.