Spend enough time around AI discussions and someone will eventually ask the question.
“What’s the best prompt?”
I understand why. Prompts are tangible. Shareable. Easy to package into a course, a cheat sheet, a Twitter thread. The question promises a shortcut — that somewhere out there is a formula separating people getting mediocre results from people getting exceptional ones.
After nearly two years of using AI almost every single day, I’ve come to believe that’s the wrong question entirely.
The biggest thing AI improved wasn’t my prompting.
It was my thinking.
I have always been a direct communicator. Sometimes uncomfortably so, depending on who you ask.
Over the years — working with clients, executives, engineers, creative teams — I learned to adapt. A CEO doesn’t need the same explanation as a junior copywriter. An engineer doesn’t think like a marketer. A customer doesn’t care about the same things a product manager loses sleep over. I learned to read a room. To say the necessary thing in the fewest words. To be clear without being clinical.
What I didn’t expect was for AI to reflect that skill back at me.
The people who consistently get strong results from AI aren’t usually the ones with the fanciest prompts. They’re the ones who know how to explain what they want. AI didn’t teach me communication. It showed me how valuable clear communication already was — and how often I was still taking it for granted.
My first real lesson arrived with Clarity Journal, the first AI-powered application I ever built.
I was using Lovable for the first time. The idea was simple: a user types how they’re feeling, and AI guides them one question at a time — conversational, not clinical. I had the concept in my head. I assumed the platform would understand what I meant.
It didn’t.
Outputs drifted. Features behaved strangely. User flows felt clumsy. I blamed the tool. Then the AI. Then the platform. Then I ran out of things to blame and had to sit with the uncomfortable possibility that the problem was me.
Not the technology.
My instructions.
I had been leaving enormous gaps, assuming the system would infer what I’d never actually said. Once I stopped assuming and started specifying — user flows, desired outcomes, edge cases, interaction logic — everything improved. The AI hadn’t changed. My thinking had.
That lesson followed me into Gamma, where I built my first AI-generated presentations and websites. Then into Claude, where I’ve now built three functional HTML sites with working CSS. Looking back, the progression had almost nothing to do with learning new tools. It had everything to do with learning to translate an idea into an instruction.
When someone tells me they’ve been using ChatGPT for months and the results are average, my response usually surprises them.
I don’t ask about their prompts.
I ask three questions.
What output are you trying to achieve? What are you giving it to work with? How many years of experience do you have in this area?
That last one always catches people off guard.
But it matters enormously.
A person with twenty years in content strategy and someone writing their first marketing brief can use the exact same model and get completely different value from it. The difference isn’t the technology. The difference is that one of them knows what good looks like.
They know when a recommendation sounds impressive but lacks substance. They know when something is generic. They know what’s missing. And they know what questions to ask next.
AI amplifies expertise. It doesn’t conjure it.
Despite spending over two decades in content and communications, some of my strongest strategic insights over the past two years came from AI.
Not because the AI was smarter than me. But because it surfaced angles I hadn’t considered — on content strategy, on legal writing and structure, on how to frame an argument. Most of those suggestions weren’t worth pursuing. Some were genuinely terrible. But every so often, one would stop me cold.
And I’d go away and actually research it. Read more. Challenge it. Stress-test it against my experience. Then fold it into my thinking.
That process made me better. I’m not ashamed to admit that. Expertise shouldn’t mean believing you already know everything. It should mean recognising value when you encounter it — and having enough experience to know what to do with it once you do.
Then there’s Artha.
I’ve been working on Artha for roughly six months. In that time I’ve made over seventy material changes — to AI workflows, N8N sequences, Telegram triggers, supporting systems. One section works the way I imagined it. Another needs refining. A third is still being worked out. There’s also a hardware interface involved that has tested my patience in ways I cannot fully describe here.
If social media told this story, this would be the part where I announce I built a revolutionary AI system in an afternoon.
The actual story is less exciting and far more real.
It looks like iteration. Like changing direction. Like realising that yesterday’s elegant solution is today’s technical debt. Like improving something in small, unglamorous increments over months.
The biggest thing Artha has taught me isn’t a workflow. It’s patience.
Not because the technology is slow. Because meaningful work takes time. AI shifts where effort is applied — less time on repetitive execution, more time on refining ideas and questioning assumptions. That’s not a shortcut. It’s a different kind of work. Slower in some ways. More honest in most.
When I look back at two years of building with AI, what strikes me most is how little it’s actually been about technology.
The tools changed. The models improved. The capabilities expanded.
But the lessons stayed remarkably human.
AI rewarded clear communication because clear communication has always mattered.
It rewarded structured thinking because structured thinking has always mattered.
It rewarded curiosity because curiosity has always mattered.
It rewarded patience because patience has always mattered.
The technology didn’t invent these things.
It just made them impossible to ignore.
So the next time someone asks me for the best prompt, I’ll probably disappoint them.
The best prompt is rarely the thing that matters most.
What matters is the clarity you bring before you write a single word.
Clarity of thought. Clarity of intent. Clarity of what you’re actually trying to build.
Get those right, and the prompt tends to take care of itself.
And that lesson will still be true long after today’s models are replaced by whatever comes next.
So I’ll ask you the same question I keep asking myself.
What has AI taught you about yourself?
Not about prompting. Not about models. Not about the technology.
About you.
