It’s no secret that I use AI as a part of my writing process.
Anyone who’s used ChatGPT or seen its output can spot the patterns.
And I’m not shy about it.
But what I have been shy about—at least until now—is laying out the actual process.
Not the tech. Not the prompt structure.
The rhythm.
The real, day-to-day process of what it’s like to write with a machine that thinks nothing but helps you think better.
So let’s talk about that.
On Learning to Think With a Mirror
Everyone’s heard the analogy:
LLMs are mirrors. They just reflect your words back with more fluff.
And yeah, that’s true.
But it’s exactly why they’re useful.
Because if you learn to think out loud—if you’re willing to explore a thought before you’ve fully formed it—that mirror can do something not many human conversation partners can:
It reflects your unspoken shape.
You put in fog.
It gives you contours.
What My Writing Looks Like Now
Here’s what it actually looks like for me.
It’s not “open AI, write blog post.”
It’s not carefully crafted bullet points and top-down outlines.
It’s… daily conversation.
Every so often, I open up a thread and start talking.
About parenting, software, literature, design, politics, obscure metaphors, ethical dilemmas in game narratives—whatever I’m chewing on that day.
And every once in a while—maybe every ten conversations—something catches.
Something sharp. Unresolved. Charged.
And that’s the moment that matters.
Because when the spark hits, the model doesn’t just mirror it.
It amplifies it.
It catches the edges I hadn’t seen.
It nudges me into focus.
And that’s where the real process starts.
And yes—of course you can do this with humans.
I do.
The best conversations with friends, with collaborators, with people I trust—they follow a similar shape: idea, spiral, resonance, refinement.
But those conversations take time. Presence. Vulnerability. And most importantly, another person who has that kind of time.
What I’m describing here doesn’t replace that.
It’s just something else.
A way to have that same shape of thought, even when you’re alone.
That’s what makes it powerful.
Not that it’s better—just that it’s available.
The Spark and the Spiral
Here’s the part that surprises people:
When I’m writing with AI, I don’t read most of the output.
Not right away.
I let it talk.
Literally. I use text-to-speech, and I just listen while doing something else—stretching, pacing, getting a drink, whatever.
I’m not parsing. I’m not fact-checking.
I’m waiting for a spark.
Most of it’s just scaffolding. Noise.
But every so often, it says something that snaps into place.
Something that resonates so hard it overrides everything else.
That’s the moment I’m chasing.
And when I hear it—when that one line lands—I don’t stop to praise it.
I respond. Immediately.
To just that one thought.
I zoom in.
I interrogate it.
I follow where it goes.
And suddenly… I’m not steering anymore.
I’m spiraling inward—deeper into the idea than I would’ve gone on my own.
It’s not magic. It’s not authorship in the traditional sense.
It’s more like archaeology.
I’m not building—I’m excavating my own thoughts.
And the model?
It’s just the wind brushing away the dust.
I decide what’s worth digging for.
It Doesn’t Work Without Care
There’s a misconception baked into how most people think about AI:
“It gave me that idea.”
“It wrote that paragraph.”
“It said something smart.”
But that’s not how this works.
Not when it’s working well.
Because the truth is—AI says a lot of things.
Good things. Boring things. Confidently wrong things. Occasionally brilliant things.
But only one of us is deciding what matters.
This process doesn’t work unless you’re listening with care.
You have to know when something is close to what you’re trying to say, but not quite.
You have to feel when the tone is off, even if the point is solid.
You have to develop a kind of editorial instinct that’s not just about correctness—it’s about alignment.
And that takes energy.
Not once, but constantly.
Every moment of this process is shaped by small acts of discernment.
That’s why I don’t let the model run the show.
I listen—but I don’t follow.
This isn’t a tool that builds—it’s a forge.
And I’m the one holding the tongs.
Yes, It Still Hallucinates (And Sometimes That’s the Whole Point)
People talk a lot about AI hallucinations—and yeah, it matters. Especially if you’re writing something that needs to be right.
But in creative work? Sometimes the glitch is the gift.
Sometimes the model says something that isn’t right, exactly.
It’s not something I would have said. It’s not something I’d even considered.
But it hits me in the chest.
Because it feels like something I was already trying to say—just hadn’t found the shape for yet.
Take this, for example.
That piece didn’t come from a prompt like
“Write something poetic about AI sentience.”
It came from a moment.
A glitch.
A phrase that wasn’t quite true, but felt real.
I paused, spiraled into it, and followed that thread all the way down.
The result wasn’t perfect.
But it was still fun.
And I think that’s what good hallucination actually is, in this context.
It’s a mirror saying something that might be false—
but if you’re listening carefully enough,
you can choose to make it true.
You’ll Know When You’re Doing It Right
There’s a moment—
if you’ve been doing this long enough—
when you realize your own thinking has changed.
Not just your workflow.
Not just your speed.
You.
You find yourself choosing words more carefully, even when you’re not at the keyboard.
You pause in conversations just a little longer, waiting to see if something better comes.
You start to notice your own thoughts as they form.
And you listen.
That’s the tell.
Because this process?
It’s not about productivity.
It’s not about getting “better results.”
It’s about building a new kind of cognitive rhythm.
One that’s part solitude, part dialogue.
Part silence, part synthesis.
You stop chasing answers.
You start following questions.
And when the mirror bends—when the reflection distorts—you don’t panic.
You just tilt your head.
And keep going.
