i thought reading meant screenshotting pages

Until I was fifteen or sixteen, I genuinely believed that reading worked by looking at a page, flipping through a book, and essentially screenshotting it with my eyes. I did not understand that reading meant parsing each word inside the larger context of a paragraph, a chapter, and whatever argument the author was trying to make. Nobody ever sat me down and explained that decoding was a sequential, phonological process that most people automate early in childhood. My teachers and peers often assumed I was stupid because I could not read, when in reality the school system simply did not spend the time to explain how reading worked, and so left my brain to its own devices.

I recently learned that I have ADHD and dyslexia, and a lot of my childhood suddenly makes painful sense. Dyslexia is a difference in how the brain handles language, with no cartoonish letters flipping backward and no connection to intelligence. At its core it often leans harder on visual-spatial and big-picture processing than on the left-hemisphere pathways that turn graphemes into phonemes and phonemes into meaning. Because a standard linear driver was never installed by default in my brain, my visual-spatial way of handling things did what it does best and tried to optimize by parallel-processing a serial data stream, treating a page of text as a single multidimensional canvas to be ingested all at once. Everyone else seemed faster at reading, at the same trick I assumed was “screenshotting,” so I assumed they were just smarter than me, and teachers often confirmed as much.

A few days ago, my dad’s girlfriend told me about a quiet afternoon from when I was probably twelve or thirteen. She was reading her own book while I sat nearby with a Harry Potter novel, flipping pages so quickly she was sure I was not actually reading. When she asked me about it, I could not explain much more than a high-level summary of what I had seen in the movies. I told her my private theory that reading was like taking a screenshot, and that once I had looked at all the words on a page I figured I was done with it. She said she had never heard anyone describe reading that way, that she found it interesting, and that it jogged a memory of someone telling her around then, when I was twelve or thirteen, that I had been diagnosed with dyslexia. Sitting there with her and my dad on the phone twenty or so years later, I realized that I had been performing an entire cognitive ritual for years while the people around me thought I was just turning pages for show.

Even now, reading is not a passive task for me. It is a highly manual, physical effort. Because word-by-word tracking is not automated in my brain, my eyes have to manually target and jump to every single word, and that constant adjustment causes severe accommodative fatigue, which leaves me with tired, heavy eyes and often blurred vision after long reading sessions. My ADHD adds another tax on top, I have come to learn, because my brain wants to leap ahead and find patterns instead of crawling along a single line of text, and that push-pull is what caused the exhaustion and the shame I carried from adults who called me stupid when I would get so tired from the effort or dizzy when I’d stand up from the blurred vision. Turns out my problem was never capacity, and the real gap was that nobody taught me what reading actually was.

Yesterday I had a conversation with a friend about how dyslexia shows up for me specifically. Our conversation clarified something I have struggled to name for years. For me the trouble lands less as words moving on the page, and more as a garbling of meaning once those words arrive in my mind. I confuse concepts, fuse them together, and miss key ideas unless I am slow and deliberate, take notes, or reread the same passage until the pieces stop collapsing into each other. That is also why reading costs me so much energy, because comprehension itself is an active reconstruction and is not some form of automatic retrieval that it seems most other people experience. My friend described his own fatigue from reading differently, as the feeling of reading pages on a kind of autopilot where at some point he realizes he isn’t retaining content and then goes back to get further context. I was enamored by his explanation of it as almost like driving on a freeway on autopilot, because there is no such thing for me. Every page is surface streets with no cruise control, and my eyes are still steering, except sometimes I can’t remember the directions and GPS is trying to function in a dead zone.

learning to learn became survival

The most significant skill I developed, looking back, is adaptive learning. I had to reverse-engineer reading in my mid-teens the same way I later reverse-engineered everything else I cared about, by building, failing, and rebuilding from first principles because the standard curriculum had already written me off and I refused to fail in college or validate this “stupid” person label everyone seemed to keep assigning to me. I hold to a rule I wrote about elsewhere on this blog, namely that using artificial intelligence creates artificial intelligence, which means I refuse to outsource the act of learning itself even when I am comfortable delegating implementation I already understand. That distinction mattered enormously when I had no degree, no credential to hide behind, and only my curiosity to prove I could do the work, both internally and externally.

I taught myself into roles at Apple, GitHub, and Coinbase, and later into some of the harder corners of the industry like web3 and cryptography at Parity and Aleo despite all this, and I have now been told by many how smart I am. I did not do this by performing a motivational poster arc about overcoming obstacles, and I instead treated every gap in my knowledge as an engineering problem with a workaround, the same way I eventually treated reading once I understood what the task actually entailed and that it wasn’t just screenshotting. Formal credentials were never the bottleneck for me because the work I wanted to do rewarded people who could ship, and shipping demanded that I learn faster than my embarrassment could catch up, which worked out for me, thankfully, though I still suffer from immense impostor syndrome in my career that I regularly combat by working harder than others do and by putting more energy in.

rigid systems as a custom api

My full neurodivergent stack includes ADHD, dyslexia, and traits that may sit on the autism spectrum, though I have not pursued a formal ASD diagnosis and I am careful not to collapse distinct conditions into a single label. What I can describe with certainty is the set of rigid habits that look like OCD from the outside and function like scaffolding on the inside. I track obsessively. I maintain strict I-do and I-do-not boundaries around how I use tools, including the animated constraint list I published in how i use ai in my work. I hyper-focus on code review and metrics because those rituals keep executive dysfunction from dissolving the week into a blur of half-started tasks, and I describe brushing my teeth in such minute detail that even my dentist’s eyes glaze over, a clear indicator for those in the know regarding neurodivergence.

I think of my habits as a custom API that has now become a superpower of sorts. A chaotic, highly creative brain like mine needs an interface layer to participate in structured environments, and my interface layer happens to be almost comically rigid because that rigidity is what lets the creativity show up on schedule instead of only at three in the morning when the rest of the world is asleep, although this does sometimes still happen.

Dyslexia also left me compensatory strengths I would not trade away even if I could install the standard phonological driver described earlier tomorrow. Dyslexics and people like me think in interconnected patterns and spatial relationships, which means we connect ideas across domains that look unrelated when compared to people who read linearly and stop at the first plausible answer, essentially as what some describe as systems thinking, which makes living in a world that was not built for my brain create adaptability as well as empathy I might never have developed if school had been easy for me. I am not framing neurodivergence as a superpower cliché. I and so many others like me, as I have been learning in the several books I picked up recently on the topic, just have a different operating system that excels at different workloads. The key, I think, for myself and others and those who want to work with neurodivergents, is to find the right workloads and unlock our strengths.

why understanding intelligence became an obsession

My struggle with information processing naturally pulled me toward questions about intelligence itself, as something I could poke at with code and learn from first principles to understand myself. This is why I am thankful for the AI revolution we are all currently in the middle of. AI makes it possible, as was its original mission at the 1956 Dartmouth workshop that coined the field, to better understand ourselves and the nature of intelligence by testing the conjecture that every feature of learning and intelligence could in principle be described precisely enough for a machine to simulate it. In that vein, I have been building models from first principles and running into the same class of roadblocks that make standard Transformers feel like brute-force text machines scaled until the world’s hardware combined can no longer keep up. Supervised fine-tuning and reinforcement-learning pipelines hit fundamental design limits, and I keep returning to the idea that rote memorization of token sequences is a poor proxy for the kind of adaptive representation a mind actually needs, like a kind of “screenshotting.”

I experimented with a visual encoder approach I called photic token language, abbreviated PTL-3, because I now realize I wanted to escape the assumption that language must enter a model as discrete text tokens in the first place, and I must have had some implicit understanding of my own relationship with language which was more visual in nature. PTL-3 did not achieve my goal of reducing training time and cost, and the public repo is messy in the way prototypes with UIs often are, but building PTL mattered to me because it came from the same instinct that made me screenshot pages as a child. I was looking for a representation that matched how my brain already wanted to process information. Later, when I read about Yann LeCun’s JEPA (Joint Embedding Predictive Architecture), something clicked. JEPA predicts in representation space and not by reconstructing surface form, which resonates with the idea that real intelligence builds internal world models in direct antithesis of parroting linear sequences like reading line by line or Bayesian token prediction, where each next token is chosen by updating beliefs about what is likely given everything seen so far. I do not know yet whether JEPA or anything I build will solve the problems I care about, namely understanding intelligence and building models I can run on my own without a token budget, but the direction of JEPA and Yann’s team feels closer to how minds like mine actually work, which resonates.

becoming without the manual

kissingparty’s poem about becoming stuck with me because growth so often disguises itself as failure when nobody gives you the manual for the task everyone else seems to perform effortlessly. I spent years feeling stupid for missing a skill I did not know existed, and I spent the years after that building systems rigid enough to let me participate anyway. Tired eyes after reading, Harry Potter pages flipped too fast for anyone watching to believe, a friend whose autopilot I still find almost mythological, obsessive tracking, strict boundaries, refusal to let agents learn for me when learning is the whole point, all of it serves the same purpose/function, which is to interface my nonlinear mind with a world that prefers straight lines. I’m a nonlinear kind of person, and finally, I am proud of it! 🙂

If any of this sounds familiar, I hope the takeaway is narrower and more useful than inspiration. You are not broken. You are simply running an operating system that never received proper documentation, and the work of adulthood is often the work of writing that documentation yourself. I am still writing mine, one rigid habit and one failed experiment at a time.

Reach out if you’re like me. I’d love to hear from you!