"Doing my job through the computer, it feels different," the man said. "It is like you're riding a big, powerful horse, but someone is sitting behind you on the saddle holding the reins."
He wasn't talking about artificial intelligence. He'd never heard the term, and wouldn't for another forty years. He was a pulp mill operator in the American South, and the year was sometime in the early 1980s, when the mill replaced hands-on control with a bank of screens. A researcher named Shoshana Zuboff was there, watching factories automate, and she wrote his sentence down. It was the line that stayed with her.
He used to judge pulp by touch — slick or sticky, a texture he could read with his hands before he could have explained it in words. Now he watched numbers scroll. The mill ran faster, cleaner, safer. It also, in his own words, took the reins out of his hands and gave them to somebody riding behind him.
None of it is about artificial intelligence. It's older than computers being called superintelligent, older than the word "software". The mill didn't lose a piece of technology. It lost what the job used to demand of the person doing it. And that was the very thing he used to point to when someone asked what he was worth.
Which is why it's worth sitting with Sam for a moment. Sam is a 29-year-old software engineer, and his job used to be one of the best-paid, most secure professions in the country. In the last six months, most of his day has moved from writing code to reviewing code an AI has already written. He said the thing that used to be "the creative, fun part of the job" has been replaced with the worst part of it: "reviewing code I didn't write." Another engineer — let's call him Matt — spends his four-hour train commute each day writing a browser game entirely by hand, every line, every character model, done without AI, on purpose. "I am actively trying to keep my axe sharp," he said. Asked what's changed at work, he said: "I spent so much time and energy investing into knowledge of the craft and now it's escaped from me."
That is the whole complaint, stated exactly. Not that the work got worse. Not that the pay dropped. That doing the work was how the value got proven, and now the doing is gone, and what's left — reviewing, approving, directing — doesn't look like anything. It looks like less.
Here is the part almost nobody tells you about that mill, because the man with the reins is the only one who ever gets quoted. Go back to his sentence for a second — someone else holding the reins while he just sits there. Now meet another operator Zuboff interviewed: same mill, same year, same new job, describing the identical change completely differently. "Sitting in this room and just thinking has become part of my job."
Two men, doing the same new job, and only one of them lost anything.
Researchers have a name for the difference between them. Some people stay in the loop — they keep questioning what the machine gives them, keep testing it, stay switched on. Others go on the loop — they just sign off on what comes out, present in the room but not really paying attention. From the outside, both look like supervising. Only one of them actually still is. The reins operator was on the loop before anyone had a word for it. The other man was in it.
Ho and Yeung, in a paper for the California Management Review, looked at workers doing exactly this — splitting a task up, directing the AI, checking what comes back. It found they don't just feel differently about the work. They produce more. They earn more. That's not there to make anyone feel better about the situation. The difference between rubber-stamping and actually running something isn't just a feeling. But producing more isn't the same as being seen to produce more — and being seen was always the actual thing on the line.
None of which means the shift is automatic, or that everyone survives it — and it isn't only about attention. Structural forces — layoffs, wage suppression, the leverage employers hold over workers — are real drivers of this grief too. Shriram Krishnamurthi, who teaches computer science at Brown University, said it plainly: "Some software engineers trained well for this, and many did not. Those who did will thrive; those who didn't are going to have to retool."
For the knowledge worker who has gone on the loop — approving output without interrogating it, present without being engaged — the grief is correct. Something is dying, and it deserves to be mourned rather than argued out of. You cannot orchestrate what you've stopped paying attention to, and calling that "a different kind of craft" would just be a lie, told more kindly.
But for the one who's still in the loop — who treats every AI output the way the pulp mill's second operator treated the numbers on his screen, as something to be read, questioned, anticipated, caught before it goes wrong — the grief is a misdiagnosis. They're mourning a death that hasn't happened. The love of the craft didn't die. It moved — except this time it moved faster, and nobody sent a notice.
The reason this keeps getting missed is that we built exactly one instrument for measuring whether someone still has the craft: can we watch them do it. That instrument is old — older than any of us can put a precise date on. It stopped working sometime in the last four years — in software development first; everywhere else is next. It was never actually measuring value. It was measuring visibility — proof you could hold up. Now the proof looks different, or it doesn't look like anything at all, and the people who can't produce it the old way assume, reasonably, that it isn't there.
It might not be. That's the uncomfortable part. The difference between the operator with the reins and the one who started thinking for a living wasn't talent, or seniority, or luck. Both of them lost the exact same task on the exact same day. What one of them kept, and the other let go without noticing, was attention — the plain, unglamorous discipline of still caring whether the number on the screen was actually right.
So before you decide which of the two people in that mill you are, ask the smaller question first, the one that's actually answerable this week: the last time you accepted what the machine gave you, could you have told someone exactly why it was right — the way the old operator could tell you, without hesitating, whether the pulp felt slick or sticky in his hands?