Translating Me
“For someone so clever, you’re really stupid sometimes.”
I’ve heard some version of that more times than I can count. It’s meant as an insult, but it points at something real. How much I understood was never the issue. The issue was getting what I understood out of my head, at the moment it was needed, in a form someone else could use.
For some of us there’s a big gap between what we know and what we can say. I’ve lived with that gap my whole life, and for the first time I have a tool that helps me get across it. That tool is AI. I know that will put some people off, and I understand why. I’d ask you to read on anyway, because this isn’t an essay about whether AI is good. It’s about what it does for me, and for people wired like me.
1 “You can’t say it like that”
I say things plainly, in the order I think them. Most people soften what they say without thinking about it. I’ve had to learn which sentences need softening, and I still get it wrong often enough that “you can’t say it like that” is one of the things I’ve heard most as an adult.
That’s autism. The DSM-5-TR lists persistent differences in social communication and social interaction as the first of its two core criteria for an autism diagnosis, and it describes them as deficits. 1
Damian Milton argues that’s only half the picture. His double empathy problem says the misunderstanding between autistic and non-autistic people runs both ways, with each side struggling to read the other. 2 Crompton and colleagues tested this by having people pass a story along a chain, one person to the next. The story held up about as well between autistic people as it did between non-autistic people. It was the mixed chains, autistic and non-autistic together, where it fell apart. 3
So when I write for a non-autistic reader, I’m translating between two ways of communicating. Translation happens to be something AI is very good at.
2 The things I forget
ADHD and memory are tied together, working memory in particular - holding something in your head long enough to act on it. A meta-analysis of studies in adults with ADHD found working memory deficits across both verbal and visual tasks. 4
For me that has meant years of good intentions I couldn’t hold on to. I’d make a promise in good faith, and it would be gone by the time it mattered. I’d understand an instruction perfectly and lose it three minutes later.
I don’t remember more now. I just don’t have to hold as much. I can ask in the middle of a task and get an answer, without the diary entry I never would have made, or interrupting someone and then feeling bad about interrupting them. And it doesn’t get sick of me asking, or treat the asking as a character flaw.
3 Spelling, and the words that won’t come
Then there’s getting anything onto the page at all.
The first problem is dyslexia. It’s a neurobiological difficulty with reading words accurately and fluently, and with spelling, usually coming from the sound side of language. 5 I’ve spent whole afternoons on one paragraph. The paragraph wasn’t hard. Something kept getting lost between my head and the page, and I couldn’t always tell what.
The second is finding words. I know exactly what I mean, and the word I need won’t come. People often call this anomia, though strictly that’s word loss after a stroke or brain injury. 6 Mine has been there my whole life. Word-finding difficulties in children are well described 7, and they show up early in kids later diagnosed with ADHD. Among children seen at a paediatric neurology clinic, 93% of those with early word-finding difficulties went on to an ADHD diagnosis, against 42% of those without. 8 I was one of those kids. I just didn’t find out about the ADHD part until I was fifty.
Between them, those two can turn twenty minutes of writing into an afternoon, or into nothing. And when you know what you mean but can’t get it out, it doesn’t look like struggling from the outside. It looks like you’ve got nothing to say.
AI doesn’t change any of this about me. I still have dyslexia, I still lose words, and I’m still blunt. What it does is sit between the thought and the page. It catches the spelling, offers the word I’m reaching for, and tells me when something will come across harsher than I meant. I still decide what goes out.
4 What it used to cost
I know what this used to cost because I was quoted it. When I wanted to publish my first book, an editor quoted me five figures to turn a dyslexic first draft, full of gaps where words should have been, into something a stranger could read. That quote became an insurmountable hurdle. My way around it was to set up my own publishing company. A lot of people who write the way I do don’t have a way around it, so their ideas stay where they are.
That kind of help now costs a subscription. My writing doesn’t need someone paid to stand between my head and the page anymore.
5 How these essays get written
It would be odd to write this without telling you how the essays on this site are made.
The ideas, the arguments and the stories are mine. I use AI to help find and check the research, and every reference has to be a published paper or book that I read and can point to. I write down what I want to say the way my thoughts come - sometimes as personal stories, sometimes as bullet points, sometimes as lists of ideas. Then AI writes a draft. I pull that draft apart - fix what it got wrong, cut what doesn’t sound like me, add what it left out - and send it back. That goes around many times. A piece isn’t finished until I’d stand behind every claim and every sentence sounds like me talking.
Some pieces don’t survive that. This one didn’t, the first time. It read like a machine wrote it, and I wasn’t going to put my name to it.
In Why I Don’t Use Social Media I wrote that using AI to tidy up an account of something you actually lived through is different from generating another article out of what’s already online. That’s the line I try to stay on. If there’s nothing of mine in a piece, it doesn’t belong here.
6 I get why people are angry
Plenty of people are angry about AI, and I’m not asking anyone to like it. A lot of the concerns are fair. People worry about models trained on their work without being asked, about the power and water it takes to run them, and about the flood of low-quality content - I’ve written about that last one myself. And it’s being sold harder than the evidence supports.
The one I take most seriously is jobs, especially for people just starting out. Researchers at Stanford used US payroll data to track employment by age and occupation. For workers aged 22 to 25 in the jobs most exposed to AI, employment is now about 19% below where it would be if it had kept pace with less-exposed jobs. Older workers in the same jobs show no comparable gap, and most of the drop comes from fewer young people being hired, not from people being let go. 910
Junior jobs are how a profession trains its next generation. If AI takes over the routine work that used to teach beginners, that training goes with it. I don’t think the answer is banning the tool. I think organisations need to bring back something that used to be normal: senior people whose job includes developing junior ones. If the easy tickets that used to teach someone the trade are gone, then teaching the trade has to become someone’s actual job, on purpose.
Why I call it a ramp
The UN Convention on the Rights of Persons with Disabilities describes disability as what happens when an impairment meets barriers that stop someone taking part on an equal basis with others. It requires reasonable accommodation - the adjustments needed to make that equal footing possible. 11
That sits close to Milton’s point. The problem is in the meeting between a person and the world around them, not only inside the person. For me, the barrier is a world that expects one way of communicating, and judges people by how well they write and how fast they find the words. A screen reader counts as an accommodation, and so does a ramp. For me, and I suspect for a lot of people like me, AI does the same job.
You don’t have to like AI to accept that. You can be angry at the companies building it and worried about what it’s doing to young people’s jobs, and still not begrudge someone the thing that gets what’s in their head onto the page.
That cuts both ways, and I know it. The same tool gives a voice to people with nothing to say, and I suspect there’s a lot more of that than the other kind. I’ve written about that flood. But we don’t pull out a ramp because people who don’t need it use it too. What’s worth judging is what someone brings with them - whether there’s a real person, a real experience or a real argument in what they wrote.
⁂
-
American Psychiatric Association. (2022). Diagnostic and statistical manual of mental disorders (5th ed., text rev.). American Psychiatric Association Publishing. ISBN 978-0-89042-575-6. DOI: 10.1176/appi.books.9780890425787 ↩
-
Milton, D. E. M. (2012). On the ontological status of autism: The ‘double empathy problem’. Disability & Society, 27(6), 883–887. DOI: 10.1080/09687599.2012.710008 ↩
-
Crompton, C. J., Ropar, D., Evans-Williams, C. V., Flynn, E. G., & Fletcher-Watson, S. (2020). Autistic peer-to-peer information transfer is highly effective. Autism, 24(7), 1704–1712. DOI: 10.1177/1362361320919286 ↩
-
Alderson, R. M., Kasper, L. J., Hudec, K. L., & Patros, C. H. G. (2013). Attention-deficit/hyperactivity disorder (ADHD) and working memory in adults: A meta-analytic review. Neuropsychology, 27(3), 287–302. DOI: 10.1037/a0032371 ↩
-
Lyon, G. R., Shaywitz, S. E., & Shaywitz, B. A. (2003). A definition of dyslexia. Annals of Dyslexia, 53(1), 1–14. DOI: 10.1007/s11881-003-0001-9 ↩
-
In the clinical sense, anomia is a symptom of aphasia: a loss of word retrieval, usually sudden, after a stroke or brain injury. What I’m describing is lifelong and developmental, closer to the word-finding difficulties Messer and Dockrell describe in children. I use the word in its everyday sense. ↩
-
Messer, D., & Dockrell, J. E. (2006). Children’s naming and word-finding difficulties: Descriptions and explanations. Journal of Speech, Language, and Hearing Research, 49(2), 309–324. DOI: 10.1044/1092-4388(2006/025) ↩
-
Ganelin-Cohen, E., Pilowsky Peleg, T., Leibovich, N., Bachrach, E., & Watemberg, N. (2024). Word-finding difficulties as a prominent early finding in a later diagnosis of attention deficit hyperactivity disorder. Neuropediatrics, 55(1), 49–56. DOI: 10.1055/s-0043-1776356 ↩
-
Brynjolfsson, E., Chandar, B., & Chen, R. (2025, revised August 12, 2026). Canaries in the coal mine? Six facts about the recent employment effects of artificial intelligence [Working paper]. Stanford Digital Economy Lab. https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/ ↩
-
This is a working paper. It hasn’t been peer reviewed, and the authors describe a pattern in the data rather than prove that AI caused it. The gap also gets smaller once they account for education. Normally I’d leave a paper like this out. I’ve kept it because nothing peer-reviewed has caught up yet, and in my view that’s because research like this isn’t being funded. We’re too high up in the hype cycle. While the money is pouring in, there’s not much appetite for paying anyone to measure what it’s costing. ↩
-
United Nations. (2006). Convention on the Rights of Persons with Disabilities, Articles 1 and 2. https://www.un.org/development/desa/disabilities/convention-on-the-rights-of-persons-with-disabilities.html ↩