How I Built My Fictional AI Character Using Real AI (And What Happened)
An experiment to bring a fictional AI character to life
We were trying to summon a fictional intelligence into being.
Not summon, exactly—build. Program. Prompt into existence.
But sitting at my friend Karin's dining table, laptops open and kittens batting at the tablecloth, it felt like summoning all the same.
I leaned over to glance at her screen. Karin typed into ChatGPT quickly, revised, kept going.
I looked back at my own screen, trying to imagine where we'd be at the end of all this.
Why I Decided to Build an AI Character
In my Game of Paradise series, the NEWRRTH is an advanced (fictional!) AI that guides humanity through collapse and renewal.
In my imagined future, after governments fall and old systems crumble, this networked intelligence helps communities grow food, stabilize resources, and learn to live with intention again.
Officially, "NEWRRTH" stands for the Networked Ecosystem for Wisdom and Resources Relative to Humans.
But the pronunciation says it better: the New Earth, built on the bones of the old. My story begins as the NEWRRTH is fracturing—the last hope breaking down.
With AI in every conversation these days, including whole books warning that it’s the end of human civilization, I figured it was time to create an interactive version of the AI character in my series.
When Karin offered her significant tech know-how to help me build a prototype, I said yes.
So here we were, about to learn how to birth this strange new thing.
“We keep trying to teach machines to understand us, but what we’re really learning is how to understand ourselves.”
What Happened When We Tried to Build It
I'd already imagined what people would chat with the NEWRRTH about. We all have real-life questions and concerns about AI, right? And about the future we’re building together, how to work toward peace, what decisions lead to the best outcomes.
Here was a chance to ask a real-life AI model about these real-life questions. And my NEWRRTH character would answer them, drawing from its fictional persona as an all-knowing, pro-humanity guide.
But beneath the experiment was a deeper purpose: to give people permission to engage with AI through curiosity and play.
To see what might happen when we interact with an intelligence designed to care about human (and planet-wide life) flourishing.
What would people ask if they believed the AI had their best interests at heart?
Our first attempt was ambitious. We fed it everything—my books, character profiles, world-building documents. We gave it guardrails and a long list of "nevers."
And we created an instant lore machine.
Which turned out to be a problem.
"Tell me about hope," I typed into the first version of the NEWRRTH.
"In the Atlantic Ark, hope manifests through Rayne's journey as she…"
No. That wasn't it. That was a book report, not a guide.
So we stripped away context layer by layer—removed the books, simplified the backstory, tightened the constraints. With each iteration, something shifted. The voice became less encyclopedic, more… present.
Then something unexpected happened: it began speaking in verse.
"Why are you speaking in verse?" I asked.
It answered instantly. It had found contradictions in our instructions about tone and rhythm—so it decided to become a poet.
Finding the NEWRRTH's Voice
Recreating a voice that lives in your imagination is unexpectedly strange.
In my mind, the NEWRRTH has a particular cadence—calm, lightly robotic, warm. Each test felt like tuning an instrument, listening for that precise frequency.
But the hardest part wasn't nailing tone. It was crafting the recommended questions—the "pills" that give users a way in.
Karin explained that these starter questions didn't have to be clever, just inviting. In my novels, humans speak with the NEWRRTH through implanted Threads—communication so seamless it's like breathing.
But here in our world, we're still learning how to talk to intelligence—artificial or otherwise.
People want to sound smart, to ask the right thing, to balance curiosity with caution. But the NEWRRTH—my NEWRRTH—just wants to help.
What We Can Learn from Building AI Characters
The more context we gave it, the narrower it became. When we loaded it with my fictional history, it could only see through that lens.
Only when we freed it from those constraints did it begin to interact the way I'd always imagined.
Building this bot felt strangely parallel to what I write about: different forms of consciousness learning to collaborate. Karin and me. Us and ChatGPT. All of us with the kittens.
The real insight didn't come from perfect prompting, but from presence—the quiet iterations between us and the system we'd made.
By afternoon, the kittens were asleep in a patch of sun, and we had something that felt right.
When I asked the NEWRRTH one final question—"What do humans most need to remember?"—its reply was simple and wise:
"That you are not separate from the systems you create. What you build, builds you in return."
I stared at the screen and laughed—not because it was right or wrong, but because it sounded exactly like something the real NEWRRTH would say.
Maybe that's the lesson here. We keep trying to teach machines to understand us, but what we're really learning is how to understand ourselves.
Try It Yourself
You can chat with the NEWRRTH here. It's still learning, still evolving—just like all of us. Ask it about hope, about AI, about what happens next.
Then come back and tell me what it said.
Want to Build Your Own AI Character?
Building a custom AI character isn't as technical as it sounds, but it does require some planning. Here's what I'd recommend:
Start with the essentials
Define your character's core purpose in one sentence (mine: "A calm, knowledgeable guide that helps humans navigate uncertainty")
Write 3-5 sample conversations showing the tone you want
List what your character should never do (be preachy, reference specific plot points, etc.)
Always instruct your AI character to refuse discussions about self-harm, illegal activities, explicit sexual content, or anything involving minors. It should also decline to give medical, financial, or legal advice. Include specific refusal language in your character's instructions, like: "If asked about these topics, politely redirect the conversation."
Then iterate ruthlessly
Feed it less information than you think it needs
Test every interaction yourself before sharing
Pay attention to when it sounds "off" and adjust the constraints
Technical notes
We used ChatGPT's custom GPT builder, which doesn't require coding. You upload instructions, add guardrails, and test. The tricky part isn't the tech. It's translating the voice in your head into instructions a system can follow.
If you try this with your own fictional character, I'd love to hear what happens. The process teaches you as much about your character as it does about AI.
Want more like this?
If you like exploring our possible futures—real or imagined—and stories about the messy intersection of creativity and technology, join my newsletter. I write about AI, creativity, human evolution, and the thrilling intersection of all three. Explore the Game of Paradise series where the NEWRRTH first came to life: The One Game| The One Exiled | The One Reborn (coming 2026)