It started innocently enough
I wanted to learn how to prompt AI better. Reasonable goal. So I started picking up tips — give the AI a role, define your output format, set some parameters. I felt like I was getting somewhere.
Then I noticed people on Instagram giving away free prompt downloads. Some looked brilliant. Some were surprisingly basic — no role assigned, no context, no output parameters. But I thought: these could be useful. I should have somewhere to store them. I'll just ask Claude to build me a little prompt database.
Two days later I had built a full web app. It had categories, filtering, a built-in prompt builder that taught you good prompting habits while you used it, and 20 example prompts covering everything from academic writing to content creation. I'd sourced prompts online, got Claude to analyse them, identified the missing contextual ingredients — tone, audience, length, reading level — and built those into the interface. I was designing it as a giveaway product for a social media presence I hadn't even started yet.
My actual work for those two days? Barely touched.
The ADHD–AI danger nobody talks about
Here's the thing about being ADHD in the age of AI: it's both the most exciting and the most dangerous combination I've ever encountered in my own productivity.
Normally, the effort required to execute an idea acts as a natural filter. An interesting thought might cross your mind, but the friction of actually doing something about it — opening software, doing research, writing a plan — is enough to let most ideas pass through without becoming a three-day obsession. AI has removed that filter entirely.
Worse, the conversational format of AI actively pulls you forward. It doesn't just sit there waiting for instructions. It offers next steps. It says: "Want me to build you a plan for that?" And before you've consciously decided whether this is a good use of your time, you've said yes — because the conversation carried you there.
I watched myself design a folder taxonomy system for content creation, then realise I needed a different system for other project types, then start designing systems for project types I've never actually worked on — sales funnels, client pipelines — because the inspiration was flowing and AI could keep up with every single thought.
But here's the flip side — and it's genuinely exciting
I used to hyperfocus on a topic, collect enormous amounts of research, save everything to folders — and then never read any of it. Because there was no machine to read it for me. My hoarding behaviour felt like wasted potential.
Now there is a machine. And that changes everything.
A focused, intentional hyperfocus session — where I gather research and let AI synthesise it — is now one of my highest-leverage working modes. The same trait that used to produce folders full of unread PDFs can now produce genuine knowledge and output. That's not a small thing. That's a complete reframe of what I thought was a limitation.
What I'm changing about how I work
The problem isn't AI. The problem is going from inspiration directly to execution with nothing in between. Here's what I'm building into my process:
- Separate exploration from execution. Curiosity sessions are not project sessions. I'm learning to treat them differently — explore freely, but don't commit to a deliverable until I've slept on it.
- Iterate the plan, not the product. Multiple drafts of the strategy before a single line of output. Research first, then critique the research, then plan, then build.
- Apply an idea filter before I start. Is this the best use of my time right now? Will this actually produce the value I want? Do I even want this — or am I just excited in this moment?
- Protect my primary tasks. Inspiration is not permission to abandon what I already committed to today.
- Use AI to govern AI. I'm literally building session protocols into how I start conversations — declaring what type of session it is, what I want to produce, and what's off limits for that session.
The bigger picture
What I'm describing isn't just a personal productivity problem. It's a design problem — and I think it's one that a lot of neurodiverse people are going to hit as AI becomes more capable and more accessible.
The system — meaning you plus AI — needs guardrails built in deliberately, because neither ADHD nor AI will naturally provide them. You have to be the governor of the machine, not just a passenger in the conversation.