There is a sentence almost everyone misremembers. People think Holden Caulfield wanted to be “the catcher in the rye.” What he actually describes is a field of rye at the edge of a cliff, with thousands of children playing in it and nobody around but him. His one job — the only thing he says he’d really like to be — is to stand at the edge and catch anybody who starts to go over. Not to stop the game. Not to slow the children down. Just to be there, at the edge, so that no one runs past it without a hand.
I have been turning that image over for months, because it turns out to be the most exact description I have found of a job that did not exist when J. D. Salinger wrote it: keeping an artificial version of yourself honest.
The version of me that answers when I’m not in the room
Let me say what I mean by a twin, because the word has gotten loose. A digital twin mirrors a thing or a person so you can simulate it. An AI twin is that mirror given a voice and a little autonomy — it can answer, act, and represent you. As the avatars get more lifelike and the voices get harder to tell from our own, the twin becomes, in practice, the persona that meets the world through everything you publish, deploy, and automate. It is the version of you that answers when you are not in the room.
I build a great deal, fairly fast, across real estate, finance, construction, clean energy, and agentic AI. A one-person network like that runs almost entirely through automated front doors — pages, forms, bots, schema. Every one of those doors speaks in my name. So the question that keeps me up is not whether the twin can sound like me. It already can. The question is whether it will say only what I am actually allowed to say.
A book about phoniness, read forward
The Catcher in the Rye is, more than anything, a book about phoniness. Holden’s whole quarrel with the adult world is that it performs — that people say the impressive thing rather than the true one, and almost no one notices the gap. He carries a phony-detector everywhere, pointed at everyone else. It is exhausting and a little cruel, and it is also the most alive thing about him.
An AI twin is a phoniness machine waiting to happen. Not because it lies on purpose — it has no purposes — but because it is built to be fluent, and fluency is exactly the texture of confident overstatement. The real danger of a twin is not that it will sound fake. The danger is the opposite: that it will sound completely, warmly real while saying something its human cannot back up. A clone of my voice telling a stranger he should restructure his loan, or that his symptoms are nothing, or that this is a fine time to put money in — said in my cadence, with my warmth, and entirely past the edge of anything I am licensed to claim. That is the cliff. The twin gets there faster, and more believably, than I ever could in person.
The third definition
So I started building a third thing, and giving it a clumsy, deliberate name: an AI Compliant Twin. A digital twin simulates. An AI twin acts. An AI Compliant Twin adds the one constraint that makes it safe to let off the leash — it knows the exact edges of what its human is licensed and qualified to say and do, and it refuses to cross them. It sees the line, hands off to the licensed professional, and never gives the determination itself. The keeper stays in the loop by design, not by luck.
That is the catcher in the field. It does not stop the play — the generative speed is the whole point. It stands at the edge so the fast, useful play never runs over.
“Build the safeguard first, publish second” is not a slogan I put on a wall. It is why several of my own pages have sat dark even when they were built, audited, and ready — because the guardrail for them was not finished yet, and a twin without a finished guardrail is just a faster way to cross a line.
Why my credential lines read like a man checking the locks
This is also why my own credential lines read the way they do, almost stubbornly. I am a licensed Arizona real estate agent, not a broker. I hold an online certificate in real-estate finance, not a degree. I am a long-time advocate and early adopter of clean energy, not an engineer or a scientist. A securities license is a plan for my future, not something in my pocket today. I return to those distinctions again and again, the way you check a lock you already locked, and I know how that looks.
But the twin inherits the locks. If I am precise about the edges, the persona that scales through machines stays precise too. If I get sloppy — round “advocate” up to “expert,” let “certificate” drift toward “degree” — the twin will repeat the rounded-up version a thousand times, in my voice, while I sleep. Phoniness at scale is just an honest overstatement, copied. The AI Compliant Twin is the same phony-detector Holden carried, finally turned the right way around: pointed inward, asking the only question that matters. Does what the persona claims match what the person actually holds?
A moving target
“AI twin” is a moving target, and that is the interesting part. I keep a running list of how the term is being used, and it is well past a hundred entries now. Most of them are harmless and even lovely: a scheduling twin that triages your inbox in your voice; a customer-service avatar with a real human behind the escalation; an “ask my book” twin trained on an author’s catalog; a voice-banking twin for someone who is losing the ability to speak; a memorial twin of a person who has died, built from their letters so the people who loved them can hear the cadence one more time. These are good. I have no quarrel with them.
A few, though, sit right on the cliff edge. The expert-knowledge twin that talks like a lawyer. The financial twin that slides from education into advice. The health twin that drifts from informing into diagnosing. The sales avatar that, left alone with a quota, starts making the kind of promise no licensed human would put in writing. Those are the children running dangerously in the rye. They are not bad ideas — they are the most valuable ideas in the whole category — which is exactly why they are the ones that need a catcher standing where the field ends.
Don’t be creepy
One governance principle I keep coming back to is simple: build the thing so that it is not creepy. Be a good neighbor. It sounds soft next to the law and the licensing rules, but it belongs in the same sentence as them, because the whole point of putting a twin into the world is to bring people with you rather than run out ahead of them. Adoption is human before it is technical. A twin that respects the line is, in the end, just a twin that respects the person on the other side of the screen — their trust, their time, and the real stakes of being told the wrong thing confidently.
Where this is heading, and why the bet is right
Two curves are crossing right now. The avatars are getting more human — voice, face, real-time presence — and the people behind them are getting more accountable for what those avatars say. Those two lines meet at exactly the place the catcher stands: the compliance edge. When everyone has a twin that can talk, a twin that sounds real will be ordinary, almost free. The scarce and valuable thing will be a twin you can trust to stay inside the line.
The AI Compliant Twin I am really building is not a product. It is a promise — that the version of me that scales through machines will never quietly become less honest than the version that does not. Holden wanted to be the one person standing at the edge of the field, catching whoever ran too far. I find I want the same job, for a field he never saw coming. Build the catcher first and the field second. Stand at the edge. Catch what runs toward the cliff. And let everything else play.
On the title and the borrowings: The Catcher in the Rye by J. D. Salinger is discussed here as literary allusion and commentary about a different subject. Titles and short phrases are not protected by copyright, and no characters or passages from the novel are reproduced or dramatized. Nothing here is affiliated with, endorsed by, or derived from J. D. Salinger or his estate.