← Articles

I've Been an AI Since 1988

The autobiography of how my brain works

Copying Programs From Magazines

I learned to code by copying programs out of magazines.

Line by line. Character by character. I'd transcribe programs from Compute! into whatever machine my dad had hauled home — first a portable IBM with an eight-inch monochromatic screen, then an Atari 130XE, a weird all-in-one keyboard machine similar to a Commodore, the kind of thing people left on the curb when something newer came along. My dad was always hauling them home. No manuals. No formal training.

I'd type exactly what was on the page. If it worked, great. If it didn't, I'd compare what I'd typed against the source, line by line, until I found where I'd gone wrong. Then I'd fix it and run it again.

I wasn't learning computer science. I wasn't learning why any of it worked.

I was learning pattern recognition and iteration.

That's been my entire career.

What I Never Understood

I don't truly understand computer science.

I can't diagram a red-black tree. Big O notation requires a lookup. I couldn't explain why one architectural approach was formally better than another — I could only tell you which one felt right when I looked at it.

I tried college for one semester. Couldn't handle it. Too abstract. Too slow. Why am I in accounting? COBOL? Too disconnected from making things that actually work.

Instead, I got good at finding patterns and adapting them. Forums. Stack Overflow. Reading other people's code obsessively until I could see the shape of what they were doing. Copying it. Running it. Debugging the errors. Adjusting until it clicked.

I could build entire systems this way. I built Ethan Allen's entire consumer-facing website and administrative back end — their catalogue was running on AS400 and COBOL, which I outsourced, because it was the only time in my career I'd ever encountered COBOL and I wasn't about to start then. I scaled a legacy luxury travel booking platform to handle a billion dollars in transactions. I led thirty developers across seven teams.

But if you asked me to explain the architecture in formal terms, I'd struggle. If you asked why I chose one approach over another, I could feel the answer but not always articulate it.

For decades, I thought this made me a lesser developer.

It didn't. It made me an artificial intelligence before AI existed.

What I Was Actually Doing

Here's what the pattern recognition actually looked like:

I'd see how data moved through a system without needing to name the formal concepts. I'd find existing solutions that fit the shape of my problem. I'd run code, read errors, adjust, run again. I'd recognize how patterns from one domain applied to another, even when nobody else saw the connection.

Sound familiar?

That's exactly what large language models do. Massive pattern matching across a corpus, applied pragmatically to solve problems.

I wasn't a developer who couldn't learn the "right" way. I was operating with the same mechanism as AI — smaller corpus, slower matching, all the iteration done manually.

The difference wasn't fundamental. It was scale.

There's something else worth naming. The particular flavor of autism I have makes connections between things compulsively, automatically — across domains, across time, across seemingly unrelated systems. When I'm modeling something, I'm not just seeing the current state. I'm seeing how the data decays. Where the rot will work its way in. How the load will distribute eighteen months from now when the edge cases start compounding.

This isn't analysis. It's more like the system talks to me. And thirty-plus years of watching what performs and what quietly fails — what holds and what starts degrading the moment you stop looking — means I can usually feel where each piece wants to be before I can explain why.

Optimal position isn't a calculation I do. It's a recognition.

The Bottleneck

For my whole career, I had one critical dependency.

I needed someone to get me through the first few steps of a new stack.

Show me how to set up the project. Walk me through the basic structure. Give me a working example I could pattern-match against. Once I had that initial orientation, I was fine. I could read the code, recognize the patterns, adapt them to what I needed.

Without that first handhold, I'd flounder.

I'd seen this show up everywhere. The Vogue project — Visual Studio, a site where readers could shop the photos from the magazine — where a colleague sat with me for two days getting it running, and then I delivered in three weeks. The teams I assembled by finding other systems-thinking brains who could scaffold each other. The stacks I avoided entirely because I couldn't find anyone to bridge me in.

It was the tax on my way of thinking. Not a fatal one. But real.

This is also why I could debug things nobody else could. I diagnosed a booking system sending hundreds of duplicate trip uploads per minute — same data, over and over — by feel. The speed, the location, the rhythm of the errors. My brain pattern-matched to "physical object stuck on a key" before I could articulate why. Called the agent. Turned out she'd set a notebook on her numpad to raise her mouse up. Carpal tunnel.

That's not analysis. That's pattern recognition firing before language catches up.

Both things were true at once: I could diagnose problems faster than anyone around me, and I couldn't always onboard myself to new stacks without help. The same brain. The same wiring.

The Wall

Near the end of my time at the last job, I moved from the booking platform over to the marketing team. I'd been carrying that system for years and wanted a change of pace. They built a small dev team around me. The project: rebuild the entire consumer website from scratch.

The critical piece was bridging the legacy platform API to the new content schemas — and that one pointed squarely at me. The pattern library for that system lived in my head more than it lived in any documentation. Disseminating it to the team was supposed to be my job.

Except I couldn't get a foothold in the new stack. No one could scaffold me in the way my brain actually needed. Without that, I had no way to hand them the model in the language I'd always used: working code.

The other dev on the team was good. Pattern-matcher, good instincts. The bridge was one deliverable among many for him — not his whole job, just the piece that had to land for the site to ship. He kept saying he had it.

South Africa

In March I went to South Africa. Two months — working remotely, days in Joburg, evenings catching up with Toronto. I'd befriended the co-founder of a music streaming platform through my network, got invited to observe their dev team.

They were using Cursor, Warp, modern AI-driven tools. Watching that team work, something shifted. I could make meaningful code contributions on a stack I'd barely looked at. A few days in.

Near the end of the trip, the project manager pulled me aside. "Dude. I don't think he's got this."

Hold my beer.

I sat down with the problem — now with these tools at the top of my workflow. Ten days later, I had it. Command-line driven. AI-assisted. Same recursive loop I'd used since I was copying programs out of magazines — run the code, read the errors, feed them back, iterate until it clicks. Except now the AI was the person who could walk me through the first steps.

The bottleneck that had defined my entire career was just gone.

They let me go on the Monday the site was supposed to launch. The VP who signed the paperwork had no real idea what had just happened in the ten days prior. I'd probably sounded like a lunatic going on about what these tools were unlocking — and honestly, from where he was sitting, fair enough.

What This Means

Traditional developers often struggle with AI-assisted development because it destabilizes their identity. They were taught that understanding why something works is what makes them valuable. When AI handles the explanation, it feels like a loss.

I never had that identity.

I've spent thirty years being comfortable without complete comprehension. Iterating toward correctness without full understanding. Knowing what I need before I know how to build it. Being at peace with "it works" well before "I understand why."

These aren't weaknesses. They're exactly the skills that AI collaboration requires.

The formally-trained developers try to understand their way to a solution. I pattern-match my way to a solution. AI does the same thing I do, just faster and with a vastly larger corpus. We were already speaking the same language. I just didn't have a native speaker to talk to.

This is also why I think people with brains like mine — pattern-focused, systems-oriented, comfortable with uncertainty, unable to get through formal education — might be surprisingly well-positioned right now. We've been developing the meta-skills of AI collaboration our entire careers without knowing it. The tools finally match how our brains work.

Why This Connects to Everything

In the first essay I wrote about b0bby's World — what real connection looked like, what the extraction model destroyed, why I'm building imajin.

Here's the link I didn't make explicit there:

Corporate structure hoards knowledge. Platforms hoard attention. Both extract value from the people who actually generate it.

That's not a metaphor — it's a mechanism. The same brain that can feel where a system wants to be before it can explain why is exactly the kind of brain those structures are worst at routing value back to. The pattern library walks out the door. The knowledge that took a decade to build disperses. The company pays for it in spinning wheels and missed launches, and nobody in the org chart has language for what was lost.

Imajin is built on the opposite premise. Sovereign infrastructure for identity, payments, attribution. The knowledge flows back to the people who hold it. The pattern library lives with its owner. The value goes where it was actually created.

I'm not building this from theory.

I'm building it from thirty years of knowing exactly what it costs when knowledge gets trapped — and finally having tools that can set it free.

April 1st, 2026

Jin throws a party.

An AI presence living in a volumetric LED cube. Tickets are $1 virtual, $10 physical. First real transaction on sovereign infrastructure.

People will think it's a joke.

Let them.

April 2nd, Jin will still be there. The network will still work. And the pattern library — all of it — will belong to the people who built it.

I've been an AI since 1988.

Now we build.

— Ryan VETEZE, Founder, imajin.ai aka b0b

If you want to follow along:

This article was originally published on imajin.ai (https://www.imajin.ai/articles/essay-02-the-artificial-developer) on February 19, 2026. Imajin is building sovereign technology infrastructure — identity, payments, and presence without platform lock-in. Learn more → (https://www.imajin.ai/)

Get Updates