A short history, an open question
We used to be
the computers.
The real question isn’t whether AI changes work — it always has. It’s whether this time is different.
At NASA, “computer” was a job title — a person, usually a woman, who did the math by hand so the rocket would fly. When the electronic machines arrived, the human computers were reassigned, and the division was disbanded in 1958.
We don’t pay people to do arithmetic anymore.
That’s the pattern with work: we abandon it the moment we stop needing it, and somehow we’ve always found more.
The strongest argument for calm
Every wave before this one left more work than it took.
Automation has always come for work one job at a time, and people kept landing on their feet.
Switchboard operators
~1M → 7K
Over a million in 1950 — one in every thirteen working American women. AT&T automated the network. They didn't end up jobless.
Power-loom weaving
0%
of the labor it took to weave cloth was automated away — and the number of weavers grew.
Exposure today
~half
of US jobs flagged as AI-exposed (Frey & Osborne ~47%; OpenAI ~46%). "Exposed" is not "eliminated."
The roles you have now
0%
of the jobs people hold today didn't exist in 1940.
Where the comparison breaks
This is the first wave aimed at the place everyone ran to.
Every wave before this one came for routine work — calculating, switching, lifting, filing — and left people somewhere to retreat: the thinking. The judgment. Taking scattered information and making something new from it. That was always ours.
But strip most white-collar work down and that’s nearly all of it — moving information around so that someone can decide. The deck, the model, the memo, the research summary. Which is exactly what these tools do.
Call it half.
The numbers are smaller than the internet claims and still large: the Frey/Osborne study flagged about 47% of US jobs as exposed, and OpenAI’s own work landed near 46%.
The shape of the next job
The specialists collapse into a generalist. The doer becomes a director.
Picture that NASA room again. Dozens of people each doing one piece of one equation, passing it down the line. That arrangement is gone, but it didn’t become no one. It became one engineer who understands the whole rocket — the drag, the thrust, how the thing flies — and who uses the tools to build it.
That’s what the near-term future of work probably looks like. One person can now ship an entire product alone. Run a full photo shoot start to finish. Carry a campaign or a piece of software from nothing to done, because the tools hold everything in between.
People have a word for this person: a builder. Not a programmer, not a designer, not a project manager — each of those was one piece of an equation. Just someone who can build the whole thing.
The next decade is a consolidation, not a crisis.
Fewer people, each doing far more, each acting more like a director than a doer. The honest read: this decade rewards the people who learn it faster than it punishes the people who don't.