What Jon and I were actually talking about
A companion note for anyone who watched our Substack Live and came away none the wiser
Mary wrote to me after the video went up and said, with admirable directness, that she had no idea what we were talking about. She is right, and it is our fault rather than hers. Jon and I have had versions of that conversation several times now, and we started this one already three flights up the building, having never shown anyone the stairs.
So here are the stairs.
Who Jon Mick is
Jon is forty-five, based in Texas, and spent fifteen years as a product manager building software for companies most of us have handed money to at some point: Deloitte, General Motors, The Home Depot, Keller Williams, Bazaarvoice. He now holds a deliberately unambitious role at his day job so that his good hours belong to his own work.
At forty-three he was identified as twice-exceptional. ADHD-inattentive, autistic, and intellectually gifted, which is a combination that hides itself well. Smart enough to compensate, different enough to be worn out by the compensating. He describes forty years of trying to fix his brain before working out that it did not need fixing so much as scaffolding.
He writes a Substack called AI Gave Me Autism, he is building a company called AIs & Shine, and he maintains a website that is less a website than an exposed nervous system. That website is the thing Mary was asking about.
The thing Mary asked about: the life model
Jon has spent about three years feeding his own data into AI. Not a diary, not a set of notes. Everything he can get into a machine-readable state.
● Sixteen years of text messages, sixty-three thousand of them, with the images described and indexed
● Five years of Whoop biometrics: sleep, recovery, heart rate variability, resting heart rate, strain
● Blood work, genetic testing, a structural brain MRI, and QEEG data from his neurofeedback sessions
● Fourteen years of YouTube watch history, Reddit activity, calendar, email, and transcribed audio from his own conversations
He calls the assembled thing his life model. It sits in a database he owns rather than in someone else’s platform, and it means that when he talks to an AI, the AI already knows his energy patterns, his triggers, his relationships and his history. He does not have to explain himself first. For a brain with fragile working memory, that is the whole point. He is not automating his life. He is externalising the context his memory will not hold, so that his thinking has somewhere to stand.
The YouTube example Mary picked up on is a good one, and it is a small piece of a larger habit. Fourteen years of watch history is a record of every rabbit hole he has ever gone down. Handed to an AI, it does two things. It shows him patterns in his own curiosity that he could never have seen from inside it. And months later, when he is chasing some new idea, it can tell him that he was already here in 2015 and found three people worth reading.
If you want to see the whole apparatus, it is public.
His site:
https://jonmick.ai/
The personal background: https://jonmick.ai/jon.html
How it is built, in some technical detail: https://jonmick.ai/about.html
A downloadable file of who he is, which you can paste into any AI to give it instant context: https://jonmick.ai/data/jon-mick-profile.md
There is also a phone number on his site, 1-928-706-9386, which connects you to an AI voice agent that has read all of it. He rings it himself while out walking with a weighted pack, because that is when his best thinking happens and he had nowhere to put it.
The conversation, in order
The recording runs a little over an hour. If you want to dip rather than swim, these are the turns it takes.
01:16 Jon has a slide deck generated for him every Friday summarising what he actually achieved that week, because his attention runs permanently forward and he needed something pointing back.
02:41 Voice agents he can ring from a trail. His phrase for the collection of them: an army of cognition. I said it sounded like a Star Trek tricorder, which dates me precisely.
04:40 His rule for what to feed the model: everything. The interesting findings come from correlations nobody would have thought to look for, such as what his overnight recovery score does to how quickly he answers a text.
05:37 The YouTube watch history.
07:35 Where insights actually arrive. A whiteboard marker on the bathroom mirror, the hot tub, and ten miles on a trail with a weight in the pack. The voice agent exists because the insights were arriving somewhere he had no surface to write on.
10:36 Recording devices in public, and what a glowing pendant does to a conversation. Also the Sony recorder in his backpack with four hours on it he has never imported, which is the most honest moment in the first half.
14:16 The one that will divide people. Jon starts recording when he notices an argument with his wife beginning. Not to win it. To go back afterwards and find the moment he could have turned it, and to learn the shape of the thing rather than only the ending.
19:09 My contribution: primacy and recency. A long conversation might carry six points, and you will reliably remember the first two and the last two. The middle ones vanish. A capture device is not a memory aid so much as a defence against the shape of memory itself.
21:20 My own book on this territory, The Augmented Psychologist, and the argument in it: used properly, this shortens the distance between something surfacing in a session and the person understanding what it means, without taking a single scrap of agency off the human. The counsellor in the chair is working extremely hard to follow a life they have only just been handed, and to hold all of it at once. Some of that load does not need to be carried by a memory.
24:37 Jon hires six therapists at once and A/B tests them, giving each the same material every week and feeding the responses back to AI to see which approach fitted his neurotype. He does not recommend this. He is now down to one therapist, one neurocomplexity coach, and a support structure he has deliberately triangulated so that any one of them can raise a hand if he starts drifting.
27:10 Where I get on my hobby horse. People call me a psychotherapist. I am a counselling psychologist, and the gap in training is enormous. The awkward part, which I say in the video, is that the research is clear that credentials are not what makes therapy work. The relationship is. The difference shows up somewhere else entirely: in what happens when someone is harmed.
32:42 The frustrating stretch. I wrote a book some years back examining the concept of rational suicide, and I am partway through a nine-book novel series about a psychologist who declines. Claude noticed both, drew a straight line to me, and has been running welfare checks on me ever since. I have told it repeatedly that the novels are fiction and that I am the happiest I have ever been. Jon gets flagged too, in his case for having enough genetic data to look like he is designing a person.
39:43 The best line in the hour, and it is Jon’s. Specificity dissolves shame. The more precisely he understands why he does a thing, the less grip the shame around it has.
43:02 Which is where I come in, because that is affect labelling and it is the same mechanism as my touch work. Naming a feeling with precision takes the heat out of it, and men are handed a vocabulary for touch and emotion roughly the size of a postage stamp.
44:33 Jon on his brain MRI and neurofeedback data as a different kind of vocabulary. Not emotional words, technical ones, doing the same job.
50:09 The fear before hitting publish, and what he calls Facebook trauma, which I suspect a great many people reading this will recognise.
54:51 His agent tells him who he owes a reply to. It also told him that one friend warranted more sentences than another, which annoyed him, and then made him work out his actual protocols for different circles of friendship.
57:07 The idea of wanting unsubscribes rather than subscribers. A smaller number of people who genuinely want to be there.
59:56 My own late diagnosis at sixty-six, and the friends back in Australia who have made it gently clear they are not interested in the new material and are perfectly happy as they are.
01:01:32 The close, and my favourite part. People drift in and out of your orbit like a wave. Silence is not absence. Someone who has not spoken to you in two years may have been reading everything you have published and simply never said so. Jon’s version: his friends are not the ones with the most in common with him, they are the ones who can let him disappear for three months and pick straight back up with no wounded feelings.
Where to go next
Jon
The recording, on his Substack:
His newsletter:
The infrastructure:
https://jonmick.ai/
Start with his piece on working memory fragility, which is his theory of his own mind and the thing everything else on that site is built on.
Mine
Six pieces, each covering ground we crossed in the video. If you only read one, read the first.
What happens when a neurodivergent man builds his own brain?
Jon Mick is 45, lives in Austin, Texas, and has published his brain MRI on the internet.
My assessment of Jon’s project as a psychologist, written in April, before this conversation. It is the long answer to the question Mary asked.
How I use AI as a professional psychologist and writer
There’s a moment in Joanna Stern’s recent conversation with Kara Swisher—the two of them on stage at New York’s 92nd Street Y, discussing Stern’s book I Am Not a Robot—where Stern explains where she drew her line. She’d spent a year handing chunks of her work and personal life over to AI. She let it do the administrative grunt work and reckoned it made …
What I actually do, said plainly, including the part where the tool stopped being a tool. This is the piece for anyone wondering how the audio versions of my books came to exist [see https://www.quiethalf.com/p/library-deepdives]
The tool you keep switching away from
There is a particular kind of browser tab arrangement that has become the natural habitat of the modern content creator. ChatGPT in one tab. Claude in another. Gemini somewhere on the right, because Google bundled it into a subscription you were already paying for and the cancellation flow involves three confirmation screens and a survey. Perplexity in …
The case for picking one AI and staying with it for a year rather than keeping four tabs open and calling it research. Continuity is the thing that makes any of this useful, and it is the reason Jon’s system works at all.
She did the work. He bought the noun
People call me a psychotherapist. On my Substack, in emails, at parties where somebody is generously attempting to explain me to somebody else. They mean it warmly, they mean it as a synonym, and I have largely stopped correcting them, because a man who corrects that sounds like a man with strong feelings about his own business card.
The psychologist and psychotherapist argument I start making at the twenty-seven minute mark, at proper length. Australia regulates the man who lodges your GST more tightly than the man who sits with your grief.
Why does he always want to touch my tits?
Companion piece to my RSD post. After this post, read the follow-on, ‘Why does she hold my spark plug when she sleeps?’
Published 16 July. The title is doing a job; the piece is about grooming rather than sex. Men have two touch systems and have only ever been told about one, and most of them end up with a single relationship licensed for prolonged affectionate contact, filed under the wrong heading. It opens on the same massage table as the video conversation, and it is the long version of the vocabulary argument I make at forty-three minutes in. It runs as a companion to my earlier piece on rejection sensitive dysphoria, so read that one alongside it if the ending catches you.
The body keeps the invoice
I was lying face-down on a massage table in Đà Lạt when it occurred to me that I might be doing something medically significant. Not spiritually significant. Not Instagram-wellness significant. Actually, measurably, physiologically significant.
What decades of accumulated stress do to a nervous system, and what the research says about paying some of it back.
The books post Mary mentioned, with the AI-generated audio walkthroughs: https://www.quiethalf.com/p/books
Longer essays:
https://mindblownpsychology.com/
Midweek and weekend pieces:
If you want to try any of this yourself
You do not need a database, a Whoop strap or a brain scan to get the first ninety per cent of the benefit. The useful move is much smaller than Jon’s version of it.
● Write a page about yourself, the way Jon has, and paste it in at the start of a conversation with any AI. Context is the whole game, and most people never supply any.
● Feed it something you already have. An export of your own writing, a folder of notes, a year of journal entries, and ask what patterns it can see that you cannot.
● Use NotebookLM to turn a long document, or a book, into an audio conversation you can listen to while doing something else. That is how the audio on my books post was made.
● Capture the thought when it arrives rather than trying to hold it. Voice memo, note app, whiteboard marker on the bathroom mirror. The tool matters far less than the number of seconds between the thought and the recording of it.
Mary, thank you for asking. The question was better than the video.