There is a column on the New York Times called “I’m Begging You: Never Write With A.I.” by Bret Stephens. It is behind a paywall, so I could only read the opening lines before I hit the wall everyone hits. But those opening lines were enough to tell me exactly where this was going, and honestly, they were enough to explain why Bret Stephens is wrong about AI, and wrong in a very specific and very avoidable way.
Stephens is a smart man. His argument is not stupid. It is actually well built, right up until you step outside his office and into the rest of the world. That is where it falls apart.
What Stephens Actually Argues
Let me give him his due first, because his piece is not some lazy panic column.
He starts by clearing the table on ethics. Passing off AI writing as your own without saying so is dishonest, sure, but that is a solved problem. Just disclose it. Easy.
He also does not pretend AI writes badly. He admits it writes fast, and that it will soon outwrite most humans at the sentence level, the same way you cannot outrun a car or beat Stockfish at chess. He is not in denial about the tool’s power. That is actually refreshing.
His real objection is deeper. He calls writing with AI “mentally enfeebling.” His analogy is an escalator next to a flight of stairs. You can ride the escalator and get to the same floor. But if you always ride it, your legs stop working the way they should. Writing, in his view, is the stairs. It is the gym where human reasoning gets built, rep by rep, sentence by sentence. Skip the climb and you skip the muscle.
And he does not draw a line anywhere. A major academic paper and a two line email to a colleague get the same verdict. Let AI outline your notes, draft your opening hook, or suggest a transition, and in his eyes you have already short circuited the actual thinking. Editing an AI draft afterward does not make the ideas yours, because you skipped the wrestling match that produces real thought in the first place.
Where He Is Actually Right
I want to be fair here, because this is not a takedown for the sake of a takedown.
Writing is thinking. That part is true and it is old news, not new news. Long before anyone worried about AI, Socrates worried about writing itself. In Plato’s dialogue Phaedrus, Socrates argues that writing would weaken memory, because people would stop holding knowledge in their own minds and start relying on external marks on a page. He was, in a strange way, making Stephens’ argument two and a half thousand years early, just aimed at a different tool.
He was partly right too. Writing did change how memory works in human beings. Offloading a task does change the muscle you were using for it. If people offload all their written communication, even something as small as everyday emails, their ability to think clearly and independently on the page does erode over time. That is not controversial. That is just how skill atrophy works, in writing or in anything else.
So no, I am not going to pretend that concern is fake. It is real.
Where He Goes Wrong
Here is my problem with the column, and it is the same problem I have with a lot of writing that comes from a place of comfort.
His refusal is absolute. He treats AI as an all or nothing replacement instead of what it actually is for most people, which is a tool. Not a substitute for thought, a tool. A proofreader. A translation assistant. An accessibility aid for someone who did not grow up speaking English. He never makes room for any of that. It is all or nothing, and that is where he loses me.
First, shortcuts are already here, and human beings always take the shortcut. This is not new behavior invented by ChatGPT. It is just what people do, and always have done. When the calculator showed up, people did not keep doing long division by hand out of principle. When spellcheck showed up, people stopped memorizing every exception to “i before e.” When power tools showed up, nobody kept using hand saws to prove their character was intact. Human beings look for shortcuts to make their lives easier. Not because they always should, but because that is simply what we do, every single time a shortcut becomes available.
Second, the world Stephens lives in, the world the West has built over the last century, runs on capitalism, and capitalism has stretched the gap between the haves and the have nots to a scale that is hard to even picture. More than half the people on this planet live below the poverty line. A person trying to make enough money to survive and pay rent is not thinking about brain exercise. He is in survival mode. When someone in that position gets access to a tool that makes them faster and more competitive against publishing houses worth hundreds of millions or billions of dollars, of course he takes the shorter, easier path. Anyone would.
Third, not everyone on earth has had the same access to books, schools, and universities that Stephens has had, or that people like him have had. This is about privilege, plainly. Most of the world does not get to spend an afternoon pondering why the apple fell to the ground. People are in emergency mode most of the time, rightly or wrongly, and that is simply the fact of how most of humanity lives.
Put those three things together and over reliance on AI is not some moral failure. It is the logical, natural response for people who are competing against a stack of disadvantages that has nothing to do with talent. It has to do with the color of someone’s skin, the part of the world they were born in, whether their family had old money or no money, the schools they had or did not have access to, and even something as arbitrary as the color of their passport, which decides how far they are allowed to travel and how seriously the world takes them before they have even opened their mouth.
Why the “Begging” Line Doesn’t Land
This is really where the column loses its footing for me.
It is easy to write an emotional line like “I’m begging you.” It reads well. It sounds principled. But that line means nothing to a person who has to pay for his mother’s cancer medicine. It means nothing to someone trying to put a plain piece of bread on the table for his family tonight. To that person, “please don’t use AI to write your emails faster” sounds like something said from a very comfortable chair, in a very safe room, by someone who has never had to choose between his craft and his rent.
There is a name for this in psychology, even if Stephens never uses it. Abraham Maslow laid it out decades ago in his hierarchy of needs. People do not reach for self actualization, for the refined pursuit of craft and mastery, until their basic needs of food, safety, and stability are met. You cannot ask someone standing on the bottom rungs of that ladder to worry about the long term health of their prose. They are busy trying to survive the week.
Stephens is standing near the top of that ladder and shouting down at everyone still climbing it, telling them to slow down and admire the scenery.
The Verdict
Stephens’ essay comes from a genuinely idealized, academic worldview, one where everyone has the time, the energy, and the resources to treat writing as a craft rather than as a function. That worldview is not wrong for him. It is just not the world most people actually live in.
His point about cognitive atrophy is valid in a vacuum. But an absolute ban, argued with zero nuance and zero acknowledgment of who has access to what, ignores something important. For billions of people, AI is not an engine of intellectual laziness. It is a tool for economic survival and for a kind of democratized opportunity that did not exist before.
And there is one more layer of irony worth sitting with. An essay begging humanity to step away from AI got read, broken down, and argued with, with the help of AI. That is not a gotcha. It is just proof of the point. The tool is already woven into how people think, write, and survive. The question was never whether to use it. The question is whether we use it with our eyes open, or pretend, from a comfortable chair, that everyone else has the luxury to say no.
