I wrote this book because AI, used correctly, is a force multiplier with no real precedent, and arguing it as either a miracle or a menace is lazy. Treat it as a tool and you’ll do fine. Let it do your thinking for you and you’ll have problems you won’t see coming, because nothing about the output will warn you.
It amplifies thinking and replaces none of it. You can increase your productivity, improve the quality of what you produce, and do work you wouldn’t previously have had the capacity to attempt. The approach is simple in its principle and requires mostly the will to maintain it.
What does AI amplification look like?
The removal of cost from the parts that were never the valuable bit. Drafting isn’t where the value in writing lives. Neither is restructuring, or assembling research, or producing the fourth version of something that needed to exist before you could see what was wrong with it. Those are all necessary and none of them is the thing anybody is paying for.
Take the cost out of those and somebody with judgment produces a great deal more of what they’re good at.
What it requires is that the judgment already exists. Hand these tools to somebody who can’t tell good work from bad and they’ll turn out bad work faster, with cleaner grammar. I think that’s how most people are getting hurt by AI right now, and it’s dishonest of the people selling the tools to stay quiet about it.
Why does the other path not announce itself?
Because fluent output produces the sensation of understanding, and the sensation arrives whether or not the understanding does. Nothing goes wrong visibly. There’s no moment where you notice that you’ve stopped forming a view before asking. The work continues to come out, it continues to be acceptable, and the internal experience is of being more productive instead of less capable. By the time the difference matters, the capacity that would have detected it hasn’t been exercised in some time.
That silence worries me more than anything else in the book. A writer who notices the slide can correct it. A writer who doesn’t keeps publishing, and the readers find out first, by not coming back.
That’s why I called the introduction The Other Path Doesn’t Announce Itself. It’s a description of what makes this particular failure hard to catch.
How do you tell which path you are on?
One question, asked regularly instead of once. Could you tell if the output were wrong?
If the answer is yes, you’re amplifying. You have a view, the tool produced something, and you’re in a position to evaluate it. That’s the whole safe version and it doesn’t require any particular restraint. If the answer isn’t sure, you’ve handed over more than you intended. Not catastrophically and not permanently, but the direction outweighs the position, and the drift is silent.
Not catastrophically and not permanently, but the direction outweighs the position, and the drift is silent.Share on X
The book on this: The Birth of the Augmented Human is 31 chapters on using these tools as an amplifier instead of a replacement, and it’s a companion volume arguing what happens when you get that backwards.
Why write a companion volume that argues the other way?
Because both cases are true and I couldn’t write one without the other. These tools multiply capability. They also erode it. Which one happens depends on the method and not on the technology, and a book arguing only one side would be selling something.
The two volumes are the same argument seen from the two ends. One is about what becomes possible when the mechanical cost disappears. The other is about what disappears alongside it if you’re not deliberate.
I use these tools every day. I also think the second book is the more urgent one, and holding both is the real position.
What does the will to maintain it mean?
The unglamorous part, where most people fail.
What it takes is the willingness to keep doing something that’s become optional. Forming your own view before asking. Checking output you could reasonably assume is fine. Keeping one area where you never reach for help, purely to know that you still can.
None of that is difficult. All of it is skippable, and it’s skippable every single time. That’s a different problem from difficulty and considerably harder to solve.
The productivity crowd gets this badly wrong. They tell you to automate everything you can, and they measure the hours you saved this quarter. Nobody’s measuring what you lost the ability to do by next year, and you’re the one who pays for that.
How do you keep AI working as an amplifier?
Form a view first, then ask, then compare. The comparison is the part that keeps the capacity alive.
Give away the mechanical work freely. Drafting, formatting, restructuring. None of that was ever the value.
Never give away the judgment. That’s the one thing the whole arrangement rests on.
Ask whether you could catch an error. Regularly, and answer straight.
Keep one thing unaided. A single area where you work without help, to know the capacity is still there.
The productivity gain is real, and I want every writer to have it. What I hate is watching people trade their judgment for it without knowing they made the trade. The companies selling the tools have no reason to warn them, and by the time the loss shows, getting the judgment back takes far longer than keeping it would have.
