AI as Assistive Tech: A Neurodivergent Case for Building With It

On executive dysfunction, the friend who drives you home, and why the tool isn’t the villain

I’ve killed a lot of projects. Not dramatically, not with a bang — just the slow quiet death of a thing I was once genuinely excited about, now sitting in a folder called _archive or ideas-v3 or DO NOT DELETE (old).

The pattern was always the same. I’d have an idea. I’d get obsessed. I’d research, sketch, prototype, problem-solve — and then somewhere between “this is going to be amazing” and “okay, now I just need to actually build the whole thing,” my brain would file it under Not Interesting Anymore and walk away.

I’m neurodivergent. This isn’t a discipline problem. It’s executive dysfunction. The ideation phase gives me dopamine. The execution phase gives me friction. And my brain has a very simple policy: no dopamine, no deal.

The graveyard of my ideas is full of things I loved right up until the moment I had to shovel.

Enter AI, stage left

I’m not going to pretend AI is a magic cure. But for me, it’s the closest thing I’ve found to a scaffold for the parts of a project my brain refuses to carry.

The first shift was small: I could use it to unfold an idea. Talk it through, get a sanity check, ask the dumb questions, do quick research, chase tangents without losing the thread. It felt like having a project buddy who was just as excited as I was — and who never got bored when I changed direction for the fourth time in an hour.

The second shift was the big one. When we finally hit the okay, let’s do this moment, the AI didn’t stop. It kept going. It started building. And I stayed in the loop — giving input, steering, correcting — but I wasn’t the one shoveling anymore.

I’m still there. I’m still navigating. But I’m not driving.

Think of it less like a chauffeur and more like the friend who drives you home when you’re not feeling good. They’re not becoming your personal driver. They’re not taking over your life. They’re just getting you from A to B tonight, because tonight you can’t do it yourself. Tomorrow, maybe you can. Tonight, they’ve got you.

And that’s the difference. The friend gets me home tonight. The project gets finished. Not another 80% draft rotting in a folder called ideas-v3, not another thing I loved right up until the shoveling started. It ships. It’s happened before — just far less often than not.

The project that actually made it

For months I had a job tracker app in my head. Not a vague idea — a real one. I knew what it should do, how it should feel, what problem it solved. I just couldn’t start it.

I was burned out. Depressed from job searching. The mental energy required to build the thing simply wasn’t there, and no amount of “just start small” advice was going to manufacture it.

I tried some early agentic flows. Not great. Then two months ago, something clicked: I built my own agentic system from open-source and cheap parts — good enough output, not the $200–300/month Claude Max tier, because when you’re unemployed you don’t exactly have that lying around.

That system and I spent two months building the app. And a few days ago, I slapped the v1.0 label on it and open-sourced the whole thing for free.

Two months. From can’t start to shipped. That’s not a productivity hack. That’s a different life.

(Why I open-sourced it is another blog post.)

“But did you really make it?”

I see the discourse. Vibe-coding, AI-generated everything, people confidently declaring “I made this” over something a model assembled in an afternoon.

And honestly? I think that’s fine. You wired the electricity in your room — you didn’t smelt the copper or forge the pliers. A tool is a tool.

Here’s the thing though: the tool doesn’t decide how much of the process matters to you. You do.

You want coffee? Push a button on a machine, walk into a café, or roast, grind, and scientifically brew it yourself. Same end product, wildly different process. Nobody’s out here accusing the café customer of cheating at caffeine.

So the real question isn’t “did you use a tool.” It’s “how much of the process did you want to own — and did the tool let you get to the part that actually mattered to you?”

For most of my life, the answer was: I wanted to own all of it, and so I owned none of it, because I never finished anything. Now I own the parts that matter. The rest gets carried. The project exists. The v1.0 exists.

I’ll take that trade every time.

The part where I stop being polite about it

There’s a version of this conversation that collapses into two camps: AI is slop versus AI is the future. Both feel like people yelling past each other from opposite ends of a very wide room.

I think we’re being way too precious about this. The current cultural reflex — don’t offend anyone, don’t say the wrong thing, don’t step on toes — has made us terrified of actually having the argument. And this argument needs to happen. Not the polite version. The real one.

So here’s the real one.

Using AI to help you finish a project your disability was blocking? Not the same as scraping artists’ work to undercut them. Using AI to research and structure your own ideas? Not the same as generating a fake persona to scam people. Using AI to carry a project you’re genuinely invested in across the finish line? Not the same as pumping out engagement-farm content.

And here’s what keeps getting lost: the AI decides none of that. The AI isn’t greedy. It doesn’t want to dominate a market. It doesn’t consciously steal someone’s craft and slap its own name on it. Every one of those decisions is made by a human. Every one of those behaviors is human behavior. We were doing all of this to each other long before the models showed up.

That doesn’t let the companies off the hook, by the way. The incentives baked into these tools — what the defaults are, what the training data was, who profits — are choices made by humans too. Greedy corporations preprogramming behavior into a product is still a human decision. But it’s a different human decision than the one I’m making when I use the tool to build something I couldn’t otherwise build.

The model is a mirror. What you see in it depends on who’s holding it.

The tattoo

In Thailand, a lot of fighters and people in dangerous jobs wear sak yant — sacred tattoos, often Khmer script, inked by monks or ajarns. To an outsider they look like a few wavy lines. Some tourists get them as “cool” souvenirs, not always knowing what they signed up for.

Here’s the twist: they’re protective blessings. But the belief goes that they only work if you live a true and honest life. Walk the wrong path, and the same tattoo that was meant to protect you turns into a curse.

I think that’s a pretty good way to think about AI.

A tool that protects you — carries you across the finish line, gets you home, builds the thing you couldn’t build — as long as you’re honest about what it is and what you’re doing with it. And the moment you use the same tool to lie, to steal someone’s craft, to scam, it digs you a deeper hole and lets your moral compass rot even further.

The tattoo doesn’t decide which it is. You do.

Same tool. Different life.