There’s a section near the start of Four Thousand Weeks that I can’t stop thinking about.
Oliver Burkeman opens the book with a bold prediction made by the famous economist John Maynard Keynes. Back in 1930, Keynes stood in front of a room and told everyone that technology was about to make us roughly eight times richer. The machines would do the work. And within a hundred years, his grandchildren’s generation (that’s us, right now) would have so little left to do that the hardest problem we’d face was figuring out what to do with all our free time.
He was serious. He thought leisure was going to be the crisis of the day.
We already know how that turned out.
Yes, we got the productivity increase. But the 15-hour workweek never materialized.
Why? Because we took every ounce of that efficiency and somehow filled the time right back up.
Burkeman tells this story to make a point that our unbridled ambition keeps pushing us to chase ever more productivity.
I want to borrow that argument to make a point about something else, because the exact same prediction is being made right now, in real time, about AI.
And I think it’s wrong in the exact same way.
Expert predictions always subtract the human
Throughout history, every time a disruptive technology show up, the experts predict the humans will eventually be removed.
Not only are those experts always wrong, they’re wrong in the same way: the machine takes a slice of the work, so the human is needed a little less, and eventually, not at all.
It feels obvious. It seems like simple math. And it is reliably, almost comically, wrong.
In fact, the historical record is almost absurdly one-sided. Harvard economist James Bessen went through the 270-some occupations the U.S. Census tracked in 1950 and found exactly one whose disappearance could be blamed on automation: the elevator operator. Plenty of other jobs were impacted (i.e., telegraph operators, boardinghouse keepers, etc.), but in every other case, the technology changed the work rather than eliminating it.
Take the ATM. When cash machines started showing up everywhere, everyone assumed they’d eliminate bank tellers After all, the machine literally did the teller’s main job. In 2011, a sitting U.S. president said exactly that out loud, pointing to ATMs as proof that technology lets businesses run with fewer workers.
Reasonable. Obvious. And dead wrong.
Turns out, during the years ATMs were spreading across the country, the number of bank tellers actually went up. ATMs made each branch more efficient, so the banks opened far more branches, and the tellers staffing them stopped being human cash dispensers and became the people who actually helped you. They ended up doing the relationship work that no machine could touch.
That’s the arc every single time. The tool eats a task, and the people doing only that task get hit, but the work itself doesn’t vanish. It changes. It evolves. It moves. But the human doesn’t get replaced.
Nobody’s mad at the spreadsheet
In 1979, two guys named Dan Bricklin and Bob Frankston shipped a program called VisiCalc for the Apple II. It cost ninety-nine dollars, and it was, more or less, the first real spreadsheet.
It did instantly what used to take a person a full day with a pencil, a paper ledger, and an adding machine.
The first official customer was an accountant named Allen Sneider. The story goes that he walked into a computer store, saw it running, and was floored when he saw a day’s worth of work done in seconds, right in front of him.
Now run the Keynes subtraction math: if a $99 program can do a day of an accountant’s work in seconds, then accountants are finished, right? The whole profession is about to get deleted.
That’s exactly what people predicted. But here’s what actually happened.
The clerks did take a hit. But the accountants didn’t. For the people whose entire job was running the numbers (the human calculators, the bookkeeping clerks), there are now something like 400,000 fewer of those jobs than there were in 1980. The subtraction was real for them. The machine took the task, and the task was the whole job.
But accountants? There are roughly 600,000 more of them, meaning there are now more accountants than ever.
Read that again, because it’s critically important. The tool that was supposed to delete the profession actually grew it.
Why? Because once the arithmetic got cheap, people wanted a lot more of it. Suddenly, you could ask questions you never could’ve afforded to ask before: What if we raised prices 4%? What if we opened a second location? What if we modeled the next five years three different ways? Every one of those requests used to take a week of somebody’s time to answer. But suddenly it could be done in an afternoon.
So businesses started asking a lot more of those questions.
And answering these questions requires more than just arithmetic. It’s judgment. It’s taste. It’s knowing which questions are worth asking in the first place. The spreadsheet ate the calculation. But it didn’t replace the human. It couldn’t touch the judgment. So the human job moved out of the math and into the meaning.
That’s why, forty-five years later, nobody is mad at the spreadsheet. Nobody feels betrayed when their accountant uses Excel. Nobody asks, “Did you crunch these numbers by hand?” The machine took the part that was never the point, and left the human standing exactly where they mattered.
“Is this AI?”
Now let’s apply this to something we see all the time online: the “is this AI or is this human?” argument.
I think that binary choice completely misses the point.
Here’s the clearest version of it I’ve run into lately. I was listening to a tech podcast where a developer mentioned he’d built a new app “with a little help from Claude.” But I also keep hearing non-developers describe the exact opposite experience — that they’d “vibe-coded an app, and Claude basically did the whole thing.”
In both cases, it’s impossible to answer “Is this human or AI?” Because in both cases, it’s human and AI.
And in both cases, the human had to decide what the thing was even supposed to be. AI coding tools dropped the technical barrier to the floor. But the thinking barrier didn’t move an inch.
You still have to know what problem you’re solving. You still have to picture how it should work. You still have to say what you want clearly enough that the machine can actually build it.
So “how much did Claude do?” is unanswerable, and frankly, completely beside the point. One of them would say ten percent, the other ninety, but neither number tells you whether the app is any good.
AI does the same thing the spreadsheet did, just to creative work instead of arithmetic. The spreadsheet ate the calculation. AI eats the generative grunt work — the rough draft, the first fifteen versions, the blank-page friction, the “just give me something to react to.” That stuff used to cost you a week. Now you can do it in an afternoon. So you can do a lot more of it.
And just like the accountant, once the labor falls away, what’s left isn’t nothing. It’s the part that was always the actual job: taste. Judgment. Deciding what’s even worth making. Looking at what came back and knowing in your gut, faster than you can explain it, that this part is true and that part is hollow and this whole section has to go. That’s the work. That was always the real creative work. The blank page was just the labor that hid it.
So here’s where I think the real discomfort is coming from. Underneath all the hand-wringing about what’s AI-generated content and what isn’t, I don’t think we’re actually scared of the machine doing too much. I think we’re scared there’s nothing left that’s truly ours. That if the tool can draft and rewrite and polish, then the labor was the only thing we were ever really contributing — and once it’s gone, so are we. That this time, the subtraction really does go all the way to zero.
But the fear has it exactly backwards. The labor was never the part that was ours. When drafting took eight hours, it was easy to think the typing was the work. Now that drafting takes eight seconds, you can finally see what was doing the work the whole time, and it was never the typing.
It was the taste. The judgment. The voice. Knowing what’s true, what’s hollow, what’s worth keeping, and why.
No machine can achieve that, because the one thing AI is built to produce is the median, and the median is, by definition, average. Point it at a blank page, and that’s exactly what you’ll get back.
But here’s the PKM takeaway: point it at your own material — your notes, your reactions, the stuff only you have — and it amplifies something only you could have made. That’s the difference between generation and connection, and it’s honestly the whole point.
The Bottom Line: Put Your ❤️ Into It
The whole “made by a human vs. made by AI” question is the wrong framing.
Think about what that label is actually trying to certify. When someone wants to know “did a human make this?”, what they really want to know is: did anyone care? Did a person with taste stand behind this and decide it was good? Or did somebody just dump the raw output and ship it cold?
That’s the real distinction. Not human or machine. Not the amount of AI. Whether a human brought their taste to the part that needed it, and stood behind the result.
As a creator, I believe AI can, and should, be used to make what you create better. I also believe that soon every piece of content will be touched in some way by AI. Even if you never open Claude or ChatGPT, if you run a writing assistant like Grammarly, you’re technically already using it.
So let’s not get caught up on labels like “made by a human” or “made by AI.”
If I had to choose a label for the things I create, I’d want it to be Made with ❤️ (+ AI).
The heart is the taste, the human judgment on what’s worth making and whether it’s any good. The “(+ AI)” is just being honest: it’s the spreadsheet doing the arithmetic so you can do the thinking. Don’t hide it, and don’t apologize for it. Stand behind the heart.
Keynes wasn’t wrong because the technology let him down. He was wrong because he thought the tool would replace the work. But it never does. It strips away the labor and leaves us holding the part that was ours all along.
So stop asking whether work was made by a human or made by a machine. It's probably both.
Ask whether you showed up for what you make and put your heart into it.
