AI Will Reveal You
Same tools, opposite lives. One person decides faster than ever. The other has not decided anything in a year. A four-month MIT study suggests what is happening in their brains.
Two executives, both accomplished, both using the same AI tools. Both are composites of people I advise.
The first uses these systems the way a good climber uses oxygen: to go where judgment says to go, faster and with more reserve. Research that took a week takes an hour. Options that would never have surfaced, surface.
Then the machine falls silent, because the machine’s part is done, and the person decides.
The second has never been more productive and has not made a consequential decision in a year. Every choice can now be analyzed further. Every plan can be regenerated in a dozen variants by morning.
The tools that were supposed to accelerate action have become the most sophisticated instrument for postponing it ever built.
Same technology. Opposite lives.
The difference between them owes nothing to technical skill. AI took what each of them already was, revealed it, and amplified it.
Knowledge is abundant and growing. What stays scarce is discernment, imagination, self-command, moral courage, and wise action under uncertainty.
There is early evidence that the second pattern leaves a physical trace.
An MIT Media Lab team ran participants through four months of essay writing in three groups: one working with an LLM, one with a search engine, one with nothing but their own minds. EEG throughout.
The unaided group showed the strongest and widest neural coupling, which means more regions of the brain working in concert rather than in isolation. The LLM group showed the weakest, scored lower across the board by the final session, and struggled to quote from essays they had produced minutes earlier.
Fifty-four people and not yet peer reviewed, so consider it cautiously; my own training in cognitive neuroscience leaves me wanting replication before anyone builds policy on it.
But it points where the argument here already points. The tool changes the output. Given enough time, it also changes the person operating it.
That is the argument of this essay: the machines’ real danger is the elegant refuge they offer to people who prefer generating possibilities to committing to one.
Judgment becomes the bottleneck
The standard fear says machines get smarter and human judgment matters less. Watch the actual mechanics.
When analysis is nearly free and every option can be elaborated endlessly, the only remaining bottleneck is the human capacity to choose. AI makes judgment the scarcest resource in the system.
None of this is a new worry, which is worth knowing. In 1950 the MIT mathematician Norbert Wiener published a short book for general readers called The Human Use of Human Beings. He had founded cybernetics, the study of how any system, a machine or a body or a company, steers itself by feeding its own results back into its next move.
His warning was precise for its day. He expected the trouble to arrive through delegation rather than conquest: we would hand machines our purposes, let systems optimize for goals nobody had examined, and end up as components inside machinery we did not control.
He meant it as a warning. Seventy-five years later, we have built it.
Consider what these systems, for all their power, cannot do. They cannot imagine a future worth building, only extrapolate from futures already described.
Moral risk is beyond them, since they have nothing at stake. Acting under genuine uncertainty is beyond them too; estimating within it is not the same thing. And the question of what is worth wanting, which was always the hard part, does not compute at all.
Every one of those capacities lives on your side of the keyboard. The machines are returning the deepest questions to their rightful owner.
The new hiding place
I have written elsewhere that explanation is the most refined form of delay. Intelligent people rarely procrastinate by doing nothing; they procrastinate by understanding their situation ever more thoroughly.
AI makes explanation nearly free, and delay rides in on the discount. A fourth market analysis gets commissioned, the scenario tree grows another branch, and the memo arguing for the decision is reasoned so beautifully that nobody notices the decision itself stays unmade.
It does not take a boardroom. Someone who wants to coach has compared nine certification programs, mapped three possible niches, and drafted a brand voice, and has still not had one conversation with one person who needed help.
Someone who wants to write has read four books on structure and can now explain precisely why their opening does not work. That is a real insight. It is also the reason page two does not exist.
Analysis has always been where able minds hide. It has never been this cheap, or looked this much like work.
Seneca named the mechanism in the first of his letters to Lucilius, two thousand years before anyone had a scenario tree. While we are postponing, he wrote, life speeds by. He was talking about answering letters and running a household. We have vastly better instruments now, and we are using them to avoid exactly the same thing he was.
There is a reason this feels good while it is happening. The mesolimbic dopamine system fires hardest on anticipation. The prediction of a reward moves it more than the reward does. Generating one more option delivers a small dose of exactly that signal. Choosing ends it. It is the same circuit that makes planning a trip more pleasurable than taking one. The chemistry is on the side of the open tab.
Possibility-hoarding used to have a natural ceiling, because generating options cost effort. That ceiling is gone.
You can now hold open a hundred futures at a cost of zero, except the actual cost, which is that a life acquires shape only by closing possibilities, and no system on earth can close one for you.
What to develop instead
If the machines are absorbing the analysis, then whatever they cannot absorb is compounding in value, and it is worth being specific about what that is.
Discernment, first: telling the genuinely important from the merely urgent and interesting. No model can do this for you, because importance is a function of what your life is for, and the model does not know.
Then self-command, which is the ability to act on a decision after the feeling that produced it has faded. Most good decisions die there.
Then moral courage, the willingness to be responsible for an outcome nobody can guarantee. That one costs the most, because it requires tolerating being wrong in public.
None of these is a skill in the usual sense. They are capacities of a person, which is precisely why they are scarce and getting scarcer. No workflow can hold them.
Something else gains value in this environment, and it surprised me when I noticed it. As machines get better at producing plausible language, the old stories become more useful. A myth is compressed evidence about what happens to people, tested across more lives than any dataset holds, which is why the same shapes keep surfacing in cultures that never met.
When a model can generate a thousand credible accounts of your situation before lunch, the accounts that have already outlived a hundred generations of human testing are worth more than they were a year ago. Anything can sound true now. The stories that have already been checked against millions of actual lives are the rarer thing, and they got rarer the moment plausibility became free.
The practice
A place to start this week. Take one decision you have been analyzing for more than a month.
Ask an AI to make the strongest case for each open option; let it do what it is peerless at. Then close the laptop and answer the one question it cannot touch: which of these futures is unmistakably yours to build?
Then choose, and let the consequences begin reorganizing you. Choosing was always the whole job. The machines are just making it impossible to pretend otherwise.
Leverage without surrender. That is the entire framework, and the rest of this publication is the working out of it.
Next: The Most Dangerous Moment in a Successful Life
The fourth piece states what this publication is: what it believes, who it is for, and why a life that still looks enviable from the outside can stop asking anything of the person living it.
Reading order
All five essays were published on August 2, 2026, rather than spaced across a month. They were written to be read together, and this is the order I would suggest.
You Cannot Return Home as the Person Who Left. Odysseus in the cave, my third exit, and why the self that built your success is the wrong one to author what comes next.
The Speech That Killed Odysseus’ Crew. The strength that wins the first half of a life and wrecks the season after it.
AI Will Reveal You. What gets scarce when analysis becomes free. (this essay)
The Most Dangerous Moment in a Successful Life. What this publication believes, and why it exists.
Ninety Days for the Life You Have Only Been Researching. The practical one. Six steps, a two-column log, and one question worth asking.