Somebody Should Be Writing This Down
An AI detector cannot tell a ghost from a prosthetic. I read all 242 comments from the people it will misread, and I am building the room where we stop translating alone.
A companion to The Other 80% and to I Spent Two Decades Building Products for Everyone But Me.
TL;DR: A Swedish designer named Emma Klint posted a small theory: the people who turn out to be good at AI are the ones who spent years translating themselves for systems that were never built for them. The comments became a confessional, a few hundred people recognizing their own life in her two sentences. I read all 242. Then Substack switched on an AI detector, a scanner that cannot tell a ghost from a prosthetic. This piece is about who those people are, why the machine will misread exactly the writing that finally set them free, and the small, human-held room I am building with my wife so the first of them can stop doing this alone.
This week, Emma Klint lost her patience in public.
“This is the last thing I’m going to say about AI detection,” she wrote in a note, “because I am so fed up with it.” She had gone back through more than two hundred comments on a piece she published the day after Substack switched on its new AI scanner, a piece titled “I use AI to write everything I publish here. Substack now grades that.” The comments, she reported, were full of people who had found something that helped them create, enjoy writing, and say what they actually meant. Her message to the people who looked at all of that and decided what everyone’s writing really needs is an arbitrary grade next to it: “Go get yourself a hobby.”
I have been sitting with a different set of Emma’s comments for two months. Two hundred and forty-two of them, pulled and sorted and read three times through, from a note she posted back in May. While I built something out of that thread, the platform we both publish on was building a scanner to run against the people in it. The thread comes first, because it is where the people that scanner will misread stood up and identified themselves, months before it existed.
Emma and I are not strangers. We met last winter on a video call, two people in two countries discovering we had made the same strange bargain with our careers, each of us deliberately stepping into a smaller role to buy back the cognitive room we needed to think. We recognized each other almost immediately.
So in June, when I published a piece called The Other 80%, I did the thing I always do after I hit publish, which is sit on my patio and refresh the stats like a man checking a brisket, too early and telling himself it’s patience. Publishing is the most exposed I ever feel. You write the truest thing you can find, you send it into the dark, and then you wait to see if the dark says anything back.
That night, the dark answered, though not in my own comments. It came through Emma’s feed, and I already knew the shape of the mind behind it.
The note was simple. She had a theory that the people who turn out to be good at AI are the ones who spent years translating their own thinking “for systems that weren’t built for them.” A lot of women. A lot of neurodivergent people. Especially, she said, the high-maskers who got so good at it that nobody ever noticed they were doing it.
Two sentences. And then the floor tilted, because underneath those two sentences sat two hundred and forty-two comments, and I started reading them and read until the early hours, because they sounded like the inside of my own head read aloud by a roomful of strangers.
The note that turned a comment section into a confessional
That comment section was not a debate. Almost nobody argued about whether the theory was correct. They showed up to say some version of “you just named the thing I have felt my whole life and never had words for.” There is a particular relief in that, the relief of a precise name landing on something you had only felt as a vague defect, because specificity is what dissolves shame. One writer, Anna Dasse, watched it happen and laughed that the whole thread was people “coming out of the ND-AI closet.” She was right. It read like a congregation discovering it had been meeting in separate rooms for thirty years.
And this was no room of passive lurkers. The overwhelming majority of the people who replied run their own Substack. Emma had told me, back in the winter, that her audience had tipped from people who think about AI to people who build with it, and this was that audience showing its face: writers, designers, founders, clinicians, all reaching for the same handful of metaphors without coordinating, the way a flock turns without anyone calling the turn.
Between Breaths, a dyslexic writer whose comment climbed higher than any other in the thread, needed one sentence for what takes me whole essays: AI did not hand her intelligence, because she always had that; what it finally gave her was access. Leanna Bishop, who writes essays for people who think too much and feel it all, described a lifetime of adding context, clarifying what she really meant, watching the blank stare, and trying again from a different angle, until a machine finally kept up and it felt, in her words, a little “like being held.” The Maat Movement reframed a whole life of code-switching as manual prompt engineering. father.log named the meta-skill plainly, that this is a translation skill, and a lot of people have been doing it their entire lives without a name for it.
I went down the whole thread that first night, and by the end of it I was something I have rarely been in my life. I was not the only one in the room.
The skill is scar tissue
I part ways with the celebration here, because it is aiming at the wrong target.
The thread’s consensus is that translation is the skill, and AI is the tool that finally rewards it. I understand why that lands. It feels redemptive to discover that the exhausting thing you did just to survive turns out to be worth something. But I have written a whole piece about 80% of My Cognitive Life Is Spent Translating for Brains That Aren’t Mine, where I gave that cost a name, the Translation Tax: the toll I pay to be understood at all, a workaround built because the alternative was not being heard. I wrote it the way you write about a tax rather than a trophy.
So when everyone cheers the translation skill, my reaction is quieter and a little sadder. That skill is the scar tissue. We are admiring the callus and forgetting the years of friction that grew it.
Which is exactly why the sharpest comment in the whole thread was not a celebration at all. It came from Thaddeus Howze, who writes as The Cognitive Dissident and grew up gifted and undiagnosed in an era that just called him a slow kid and tried to file him away. The translators are having a moment, he warned, and “the wound becomes the credential.” The same indifference that forged us now gets retroactively justified by what we can finally do for the companies that never cared whether we lived.
He is right, and I have made the adjacent argument myself, that the people most likely to be harmed by AI are the same ones who need it most. Told one way, this entire conversation is a society discovering it can extract value from the survival adaptations of the people it neglected, and then congratulating those people for adapting so well. If the lesson we draw is “look how useful our damage made us,” we have learned the wrong lesson, and we have re-inscribed the harm in nicer fonts.
Howze’s warning was already sitting in that thread when the platform made it literal. On July 21, Substack switched on a scanner built with a detection company called Pangram. Any reader can now point it at any post or note over a hundred words and get back a percentage, an estimate of how much of the text a machine wrote. The company even coined a word for the sin it hunts, Claudefishing, the con of dressing machine output up as human connection. Substack says the goal is disclosure rather than prohibition, and I believe them. The slop problem is real. My feed fills with text nobody meant, same as anyone’s.
But a detector cannot tell a ghost from a prosthetic. It reads the surface of the text and returns a number, and the number cannot distinguish between a writer who outsourced their thinking and a writer who finally found a place to put it. The disability world settled this question decades ago. When a person speaks through a device, the words belong to the person. Nobody scans the output of a speech synthesizer and deducts authenticity points. Nobody asks the screen-reader user to prove she really read the book. That principle did not arrive with an asterisk saying the channel invalidates the message.
Now run the scanner against Emma’s thread. Two hundred and forty-two people, most of them describing, in their own words, a channel that finally works. The dyslexic writer whose access arrived late. The essayist who spent a lifetime clarifying what she really meant to a blank stare. The percentage a scanner returns on their writing measures the prosthetic and mistakes it for the person going missing. The people most likely to be flagged by AI detection are the same ones who needed AI most, and I wrote the longer version of that sentence before the scanner existed. Emma, for the record, is not making this argument; she has said plainly that she is not interested in earning the right to use these tools, and she should not have to be. The precedent is for everyone still being asked for a permission slip.
The real change AI brings has little to do with getting better at translating ourselves outward for other people, because we were already world-class at that. The change is the surface. For the first time, a lot of us have an external place to hold our own thinking, stable and sitting in front of us, and once that surface exists, finishing a thought stops being lonely work. That is what The Other 80% was really about. The first 80% is the lifelong Translation Tax of explaining myself to everyone else. The other 80% is the part where I finally get to finish my own thought, on a surface that holds it, with something that does not flinch. It is the same reason I can’t text you back but can talk for an hour in person: a live surface holds the context that my own head will not. We Evoke asked the design question this whole community is circling without quite saying out loud: what would it mean to build systems that don’t require the translation in the first place? It is the only question that doesn’t dead-end in Howze’s trap.
A roomful of people who each thought they were alone
But the deepest thing in that comment section ran underneath the theory, and underneath my counter-theory, and underneath even Howze’s warning. It was the recognition itself.
Every one of these people thought they were alone. You can hear it in the phrasing they reached for over and over, the I thought it was just me, the I never had words for this, the I cannot tell you how emotional this feels. Each of them had spent decades building the same rare competence under the same kind of pressure, in total isolation from everyone else who had done exactly the same thing. And the note did not teach them anything they didn’t already know. What it did was introduce them to each other.
My people are all standing in one place for once, and I am going to point.
Kate, who writes at the intersection of ADHD and entrepreneurship, calls LLMs a prosthetic brain. Dr. Stephen Bradley liked that metaphor so much he named his entire Substack A Prosthetic Mind, because the tool is what holds his working memory long enough to get it onto the page. Annette Snyder described building what she calls her “open brain,” connecting it to Claude, and watching her dyslexia loosen its grip while she did systems thinking for hours without hitting the wall of overload. Nathan Organ said AI is the only thing that can hold his full intensity without collapsing or getting defensive. Jessica Anslow, late-diagnosed AuDHD, named the actual mechanism, that she spent her life holding her own model of the world and someone else’s at the same time without losing either one, which is exactly what working well with these tools requires.
There is a real architecture under the metaphors. I have argued in more clinical pieces that a brain like mine runs on external scaffolding as a prosthetic necessity rather than a convenience. That comment section is a few hundred people naming the same necessity in their own vocabulary, none of them having read the research first.
And the chorus widened from there. Areg pushed the circle past gender and neurotype to include immigrants and multilingual people, anyone who grew up translating across systems that didn’t share their context. Caroline Klokmose, an organizational scholar, had a whole research literature ready for it, a thing called boundary spanning. The Regulated System wondered aloud whether the underlying skill was translation all along, only ever dressed up as prompting. Lola the Firebird said AI helps her make sense of her thoughts and calms her down, and that she wishes it had been around when she was a kid.
It was not all agreement, and that is part of why I trust it. Maximillian Alexander rejected the word “masking” entirely, arguing it implies something to fix when the truer word is “adapting,” a skill rather than a defect. Amal Jbira raised the warning the rest of us need most, that each good output deposits a little more confidence in the tool and a little less in yourself, and that the same wound which made us translators also made us the people most likely to notice when our own self-trust starts to slip. There was even a perfect moment when Anne-Marie Traas got frustrated that half the replies were obviously written by ChatGPT, while one commenter pasted an AI’s analysis of the note in as his own comment. The community contains its own skeptics and its own people watching the watchers. Good. That is what a real one looks like.
Being “good at AI” was always too small a framing, and it walks straight into Howze’s trap. The thing with actual power in it is that there are so many of us, that we are this articulate, and that every last one of us believed we were doing it alone.
Why a cleverer chatbot was never going to be enough
The thread circled one more step and stopped just short of it.
Everyone there found the same thing: a tool that finally kept up. But keeping up is not the same as being built for you, and that gap is the whole problem I have spent two years inside. (I told the long version in June, about two decades of building clarity for everyone except myself.) Generic tools fail minds like ours because minds like ours are not generic, and no one else’s is either. A study in Nature Neuroscience this spring found that autism, one diagnosis, splits into at least two different biological subtypes, one where brain regions under-communicate and one where they over-communicate. Same label, different machinery underneath. All along, the unit that mattered was the person.
So the thing this room is actually reaching for is not a cleverer machine that studies you from a blank page while you wait. A mind that reconstructs itself every morning cannot afford to wait, and a confident AI voice can overwrite a fragile sense of self in days. What helps is the reverse motion. You build a model of yourself, in your own words, that you own and keep and change, and the AI works in service of that instead of slowly becoming it. That distinction is small to say out loud, and it is the entire game.
Reader, I wrote it down
At some point in the early hours of that first June night, reading my way down the same thread for the third time, I caught myself having a very particular thought. Everyone is talking about this. Everyone can feel that something is happening. But is anyone actually writing it down, for the record, for the history of it?
Reader, I wrote it down. All 242 comments, every name, every publication, pulled and read and sorted, and a good thing too, because the live thread is already down to 238. I cannot help it. I am the kind of person who shows up to a spontaneous gathering and quietly starts taking the minutes, because some part of me has always suspected that if I do not catalog the moment, it will scatter before anyone realizes it mattered.
Substack also shipped a small box alongside the scanner where a writer can post a statement about how they make their work. About a hundred words, to explain yourself. I will fill mine out, but I have been writing that statement since December of 2024, and it is currently sixty-three posts long. This one makes sixty-four. Disclosure at that resolution stops being a defense and becomes the work itself: a running public record of exactly how one particular mind uses these machines to reach itself.
And I am not doing this only for the archive, but because I have spent the last year meeting my people one video call at a time, each one a happy accident I could never have scheduled. I have spent my whole life circling the same few themes alone, the rejection, the masking, the quiet certainty that I was assembled slightly wrong. I never got the clean moment a lot of people describe, the diagnosis that explains everything, or the one person who finally holds the whole of you. My wiring was always a little too much for any single label, or any single person, to carry. Which is the whole point of a room. Recognition, for wiring like mine, arrives a piece at a time, from a few hundred directions. I found Emma by luck. I am not willing to leave the rest of it to luck anymore.
So I am going to reach out to them. One at a time. The ones who described my own work back to me in their own words before I ever showed it to them, and the ones who pushed back hardest, and the ones who said they hate these tools for reasons I happen to think are correct. I am going to try to build the community I have wanted my entire life and never had.
And I am going to start it as a room you can actually walk into: a small, closed group of people building the same thing at the same time, each of you making your own model of your own mind, in your own words, to keep. I teach the architecture, the part the thread kept calling a prosthetic brain. My wife, Charlotte, holds the other half. She is a trauma-informed yoga therapist, and she keeps the room safe in the body, because handing someone a high-resolution picture of themselves without a steady place to stand is the fastest way to do harm. I read. She holds. It is explicitly not therapy, the model you build is yours and leaves with you, and the whole thing runs on one rule we both believe in: every invitation stays an invitation and never hardens into a command. Architecture for the too much, tactics for the not enough.
I am building the first one now, and I would rather it be full of the people from that thread than anyone else on earth. So this is me, the guy quietly taking the minutes, sliding a note across the table. If you want to be in the first room, tell me. That is the whole ask. The door is open, and I am holding it for the people who spent too long believing there wasn’t one.
If you were in that comment section, you are already one of us. And if you have never commented anywhere in your life, but you read Emma’s two sentences and felt your chest go tight, you are one of us too. None of us were ever broken translators. What we never had was a room.
The room exists now, and I am keeping the minutes. The first seats are real, and a few of them have your name on them if you want one.
Human. Deeply seen.
Why I write this
Most people get described at low resolution, a stick figure of who they actually are, handed to them by a doctor with nine minutes or a personality quiz or their own worst day. I build the higher-resolution version. It’s called a Life Model: a structured, shame-free read of how your particular mind is built, in a form you can actually use with yourself, your doctors, and the people you love.
If a piece like this lands as recognition and not just an interesting read, that’s usually the signal. Later this year I’m running a small cohort with my wife Charlotte, a trauma-informed yoga therapist, where we build these together. I also do one-on-one builds for people who want the bespoke version. The cohort is paid work, and it’s the first thing I’ve ever charged for that I’d point my own people toward.
Neither is open for enrollment yet. What’s open is a conversation. If you want to talk through what you’re working with and whether any of this would help, just reply, or book a short call. No pitch. And if you want first word when the cohort opens, say so here and I’ll keep you posted.