I Couldn't Code Before AI. Dyslexia Kept Me Out.

I was nineteen. It was a required course, introduction to programming, and I sat in that room for three weeks before I stopped going.

Not because I did not understand what programs did. I understood what programs did. I understood the logic, the if-then, the loop, the idea that you could write a thing that would run a thing. The logic was fine. The problem was the letters. The brackets. The semicolons in places I kept forgetting. The way a single transposed character would blow the whole file and the error message would tell you nothing useful about where you had transposed it. I would write twenty lines, run the file, get an error I could not locate, spend forty-five minutes in the wrong place, and leave class feeling like someone who was not built for this work.

I was a dyslexic actor who had learned, through theater, to hold whole scenes in memory but who had never once been able to reliably proofread his own writing. Code looked like proofreading. I left.

The years I spent on the other side of the wall

What I did instead turned out to be useful. I acted. I interviewed people. I ran a photography practice for twenty years shooting theater. I learned what it meant to work in rooms full of very smart people who communicated in ways that had nothing to do with written text. I learned to listen faster than most people can talk. I learned to read intent behind language, not just language.

None of that looked like coding. It looked like a career path that ran parallel to the tech world and occasionally got close enough to smell it.

When I started building Plumwheel a few years ago, I hired engineers for the technical work and I stayed on the strategy side. That was the ceiling I had accepted. It was a reasonable ceiling. I had evidence for it. The evidence was that room I left at nineteen.

What AI actually changed

The first time I asked Claude to help me write a Python script, I expected to feel the old thing. The wall. The bracket problem. The character-by-character panic.

It did not happen the same way.

What AI does for a dyslexic coder is not magic. It does not erase the condition. What it does is remove the exact friction point that had always been in the way: you do not have to start with correct syntax. You start with intent. You describe what you want in plain language, the way you would describe it to a collaborator, and the tool gives you something to navigate from. Editing someone else’s wrong code is a completely different brain task than generating correct code from nothing. Dyslexia taxes the second task hard. The first task, the “this is almost right, here is what I want instead,” runs fine.

The dyslexic actor who could hold a thirty-page script in his head but could not reliably place a semicolon is a perfectly good person to have a conversation with an AI about what a piece of software should do. It turns out that is most of coding now.

What I have shipped since

I code all day. Not full-stack engineering; I am not pretending that part away. But the system that runs RAGnos, the automation layer, the data pipelines, the agent dispatch infrastructure, that has my hands in it. Scripts I wrote. Configurations I built. Bugs I actually debugged, in the terminal, with the AI reading the stack trace and helping me find the line.

I am a dyslexic actor who codes now. That sentence did not exist before about 2023.

The specific thing that changed is that AI gave me a collaborator who does not perform impatience when I take longer to read something. Who does not look at me differently when I describe the intent instead of the syntax. Who lets me navigate from the almost-right rather than demanding the exactly-right on entry.

The access frame, not the replacement frame

The conversation about AI and neurodiversity usually goes one of two ways. Either AI is the great equalizer that will fix all the friction for all the different brains, or AI is a threat that will remove the humans who need accommodation from the workforce first. Both of those frames are wrong and both of them are boring.

The frame that fits my experience is simpler: AI removed a specific physical barrier that had nothing to do with my intelligence or my capacity for the work. The same way a ramp does not make you a better wheelchair user, it just removes the stairs that were never load-bearing in the first place.

The stairs between me and coding were character-by-character syntax in a blank file. The ramp is a model that can read intent from description. The coding is the same coding. The dyslexia is the same dyslexia. The wall is gone.

My mother figured out that AI would talk to her patiently without anyone telling her that was what she needed. I figured out that AI would let me code without holding perfect character strings in my working memory. Neither of us was in the target audience for any of the launch announcements.

That is fine. The door was open. We walked through it.

Cheers.