There’s a debugging trick programmers use called rubber duck debugging. You get stuck on a problem, so you explain it out loud to a rubber duck sitting on your desk. The duck says nothing. But saying the problem out loud forces your brain to line it up in order, and somewhere in the explaining, you spot the answer yourself. The duck is just a piece of rubber.
I’ve done a version of this for years. The difference now is that my duck is an AI, and it talks back.
How This Usually Went
I think in systems, and I do most of that thinking at inconvenient hours. In the shower. On the drive. At 11pm when I should be asleep. There’s always a lot in motion for me: several projects, several brands, several lanes open at the same time. Ideas were never the shortage. The shortage was the distance between having an idea and doing anything with it.
That distance is where the hard part lives. Between the spark and the finished thing, there’s a brutal middle stretch where you have to pressure-test the concept, sketch the architecture, map the flows, research the tools, write the brief, and build the plan. Ideally you do all of that with another sharp person in the room, someone who can push back, catch the gaps, and warn you when you’re about to walk into a wall.
At 11pm, that person is rarely around.
So ideas would sit. They’d marinate. A few lasted long enough to turn into something. Most got quietly buried by whatever idea showed up next.
Enter the Duck
I started taking AI tools seriously a few years ago. Not for the obvious tasks, not “write this email” or “summarize this doc.” I started treating it like a genuinely smart colleague who happened to be awake at every hour and never got tired of my tangents.
I’d think out loud. It would answer. I’d push. It would sharpen. I’d wander off course, and it would follow, then help me find the road again.
The surprise wasn’t that my ideas got better. It was how much faster they moved.
I want to be careful here, because it’s easy to say the wrong thing about this. The AI isn’t doing my thinking. The ideas are still mine. So are the instincts, the direction, the “yes, that’s the one” and “no, that’s off.” What the AI took away was the tax on thinking out loud, the friction of not having the right person around. It shrank the gap between “I have an idea” and “I understand what this idea actually is.”
This Morning, For Example
I sat down with a modest agenda. Some news to catch up on, a few emails, a calendar to glance at. Nothing unusual.
Partway through, I mentioned that I’d been designing a couple of websites with AI tools, and I found myself wondering out loud whether WordPress, which has been my go-to for a very long time, was still the right answer for every situation.
For context: I’ve been a WordPress person for years, and for good reason. Plugins, flexibility, clients who can update their own pages without calling me in a panic. I was loyal to it.
What happened next was not on the agenda.
We started poking at headless CMS options. I watched a demo of one and decided almost immediately it was far more than I needed, a real “that’s a lot” moment for anyone who just wants something simpler. So I asked, more or less: what if we just built the simpler thing ourselves?
That’s when the duck started talking.
The Idea That Ate My Morning, In the Best Way
“I want a simpler CMS” turned into a full product concept in about forty minutes.
Not because the AI dropped an idea in my lap. Because every time I said something, it answered in a way that either backed the direction, pushed it further than I’d taken it, or filled a gap I hadn’t spotted.
I’d say, “I want something simple a small business owner can run from their phone.”
It would say, in effect, “Here’s what that stack looks like, here’s the user flow, and here’s where it falls apart if you’re not careful.”
I’d say, “What if it also posted to social?”
It would say, “Sure, and here are six other things you’ll probably want while we’re in there.”
I’d say, “Wait, is this actually a product?”
It would say, “Yes. Here’s the business model.”
By the time I looked up, I had a named product, a three-layer architecture, a feature roadmap across three versions, a reference client, a tech stack, a business model, and a written product brief ready to hand to a design tool.
Total elapsed time: one conversation.
None of that was planned.
What’s Actually Happening Here
I want to be straight about what AI is and isn’t doing in this, because the conversation around it tends to overshoot in both directions.
It is not doing my thinking. The instincts behind that morning, knowing a full CMS was overkill, knowing small business owners live on their phones, knowing the methodology I’d built up over the years could become a product, none of that came from the tool. Those came from years of doing the work. AI can’t manufacture that. It can only work with what you already carry in.
What it is doing is removing the activation energy an idea needs to get moving.
The old path from “shower thought” to “coherent brief” cost real time alone at a whiteboard, or the right person in the room, or enough caffeine to force it by hand. Now it costs a conversation. The duck answers in real time, and by the end you have bones, a real structure to build on.
There’s a second thing it does, and this one caught me off guard. It makes me a better thinker in the moment. When you know the tool will chase a tangent with you and help you find your way back out, you’re far more willing to chase the tangent in the first place. You take an idea further than you would alone, because the cost of hitting a dead end drops close to zero. That willingness to explore is where the good stuff hides.
The Speed Is Real, But It’s Not the Point
Everyone talks about how fast AI is, and it is fast. I went from “what are the alternatives to WordPress” to “here’s your product brief” in a single morning. That’s genuinely wild.
But speed isn’t the part that actually frees you.
The part that frees you is that the idea doesn’t have to wait anymore.
It doesn’t wait for the right meeting. It doesn’t wait for the right collaborator to be free. It doesn’t wait for a quiet afternoon that never comes while the idea slowly goes stale over six months. You have the thought, you open a chat, and you go. The idea either survives the conversation or it doesn’t, and you find out fast either way. If it dies, you’re out twenty minutes. If it lives, you walk away with a brief, a plan, and momentum.
For someone running a lot of lanes at once, with a habit of thinking at odd hours, that’s the real unlock.
The Rubber Duck, Evolved
Here’s the thing I keep circling back to.
The original rubber duck works because saying something out loud forces it into order. You hear yourself explain a broken idea, you catch the flaw, you fix it. The duck is a mirror and nothing more.
What AI adds is that the duck pushes back.
It doesn’t just reflect what you said. It responds. It builds. It raises something you hadn’t considered and asks the question you were circling without naming. It takes the half-formed thing you handed over and hands you back something closer to whole.
That back-and-forth is where the building actually happens. The plain rubber duck could never do that part.
And now that I’ve worked this way, I can’t picture going back to thinking alone.