Human-Led Mythic Selection: How a Naming Problem and Two AIs Helped Me Define a Dragon God
The Creative Leap — One Session, One Hour
The vertical jump at minute 15 is the moment a name became a myth. Everything after is amplification.
I've loved dragons my entire life.
Not casually. Not as a phase. Dragons are the thing in fantasy that has always made me lean forward. So when I built the Drakenith — a race of draconic humanoids with scaled skin and fire in their blood — they became my favorite people in the world of Astrylis. And when I designed their capital city, a fortress carved from obsidian and dragonbone in the Red Sands desert, I wanted the name to carry weight.
The name I had was Vey'Drakar.
And it bothered me.
The friction
I couldn't place why at first. The city was well-designed. The lore was solid. But every time I read the name, something felt off. Then I ran it past Claude Code — the AI system I use for structural worldbuilding — and got a straight answer:
"Vey'Drakar sits very close to 'Dracarys' and has a similar cadence to 'Vaes Dothrak.' Put together it reads like it could be a Valyrian location name."
That was it. My dragon city sounded like it belonged in Game of Thrones.
The phonetic overlap:
My Name Sounds Like From **Vey'**Drakar Vaes Dothrak Dothraki capital (GoT) Vey'Drakar Dracarys Valyrian for "dragonfire" (GoT) Two echoes in one name. For a dragon city, that's the one comparison you most want to avoid.
Now, there's nothing wrong with Game of Thrones. But Astrylis is not Westeros. I've spent over forty months building this world — its languages, its continents, its gods. When something sounds borrowed from another universe, it breaks the seal. It stops feeling mine.
Here's the thing most people don't understand about creative work: the process didn't start with inspiration. It started with refusal. I didn't sit down thinking "let me create a dragon myth today." I sat down thinking "this name is wrong and I need to fix it." The refusal — the instinct that says this doesn't sound like mine — was the most important creative act in the entire session.
That instinct came from somewhere. Decades of reading fantasy. A lifetime of caring about how dragon mythology should feel. An accumulated aesthetic sense that no AI taught me and no AI shares. When the name felt wrong, it wasn't a technical judgment. It was an artistic one. And that judgment set everything else in motion.
Two AIs, one human, one problem
Here's what was actually happening: I was working with two AI systems at the same time. This isn't something I planned as a methodology. It's just how my workflow evolved.
ChatGPT is where I brainstorm and get feedback. It's good at generating options, analyzing narrative structure, and telling me why something works or doesn't. When I shared the naming problem, it came back with three linguistic directions and fifteen alternative names organized by theme: ash-based, bone-based, draconic language roots.
Claude Code is where I build. It holds the entire codebase of my worldbuilding project — hundreds of files, tens of thousands of lines of lore — and can modify, cross-reference, and integrate across all of them simultaneously. When I told it to flag the GoT echo, it did. When I later gave it a new name, it propagated the change across twenty-three files in minutes.
I am the person who looks at fifteen name options and feels something. Not thinks — feels.
Among ChatGPT's suggestions was "Khaross."
And something clicked.
The moment a name became a myth
What happened next is the part I want to be precise about, because this is where the creative method either means something or is just marketing.
Neither AI suggested what came next. ChatGPT generated the name. Claude Code was waiting for instructions. But when I saw "Khaross," a story appeared in my head — not fully formed, but its shape was clear:
The dragon was a god. The last of the dragon gods. He was wounded during the Shattering — the ancient catastrophe that shattered a continent. He used the last of his strength in the war. Then he flew south, to the desert where dragons had always come to die. He landed. And his holy blood — divine fire — seeped into the earth. It never stopped burning.
Centuries later, the Drakenith found his bones. Instead of burying them, they built a city inside them.
They called it Ash'Khaross. The Ashes of Khaross.
Before and after — what the name unlocked:
Element Before (Vey'Drakar) After (Ash'Khaross) City name Generic fantasy draconic Named for a specific dead god Eternal flames "Drawn from beneath the earth" (vague) Holy blood of Khaross, still burning Drakenith corruption resistance "Natural trait" (unexplained) Divine blessing from a dragon god The desert Setting / geography Sacred burial ground of dragons The skeleton Cool architecture Reliquary — every wall is shrine The city's purpose Military fortress Vigil for a god who may not be dead Emotional weight Interesting place The Drakenith live inside the being who saved the world Same city. Same bones. Same fire. But the myth transformed every detail from described to meaningful.
That leap — from a name on a list to a founding myth — did not come from either AI. It came from me. From a lifetime of caring about what dragons mean. From knowing, instinctively, that my dragon city deserved more than a cool name. It deserved a sacred origin.
But here's what's equally true: without the AI-generated option space, I might never have found the word that triggered it. And without Claude Code's ability to build it out into a 340-line document and propagate it across the entire project in one session, the myth would still be a note in my head instead of a living piece of my world.
What each participant actually did
I want to be honest about this, because most writing about AI and creativity either overhypes the AI ("it basically wrote the myth!") or diminishes it ("the AI is just a fancy autocomplete"). Neither is accurate. Here's what actually happened:
Me — the human:
- Detected the dissonance (the name sounds like GoT)
- Selected the option with true resonance (Khaross) from a list of fifteen
- Made the mythic synthesis (the dragon was divine, his blood still burns, the city is his body)
- Directed every decision at every fork
- Provided the governing taste — shaped by a lifetime of loving dragons
ChatGPT:
- Generated fifteen alternative names across three linguistic directions
- Analyzed the completed document and identified why it worked
- Suggested enrichments: the idea that Khaross might not be entirely dead, the pilgrimage tradition, the oral storytelling tradition
Claude Code:
- Flagged the original phonetic problem
- Built the myth into a complete 340-line city document with geography, architecture, governance, defense, atmosphere, and closing philosophy
- Propagated the name change across 23 files in the project
- Integrated ChatGPT's enrichments back into the document
- Checked all references against existing canon
Three participants. Three different roles. One result that none of them could have produced alone.
Who did what — at a glance:
Action Human ChatGPT Claude Code Detected the naming problem X Flagged the GoT phonetic echo X Generated 15 alternative names X Selected "Khaross" X Invented the founding myth X Built 340-line city document X Analyzed why it worked X Suggested "not entirely dead" thread X Integrated enrichments into doc X Propagated rename across 23 files X Directed every fork X Notice the pattern: the human does all the meaning-making. The AIs do the expansion and the execution.
The method
After the session, I sat with what had happened and tried to understand why it worked. Not just for this city, but as something repeatable.
ChatGPT helped me sharpen the framing. The key distinction it made: my role was not "decision-maker." Decision-maker implies choosing from a menu. What I actually did was recognize the true form when it appeared — and then push it further than any tool could. That's aesthetic authority, not menu selection.
The method has five steps:
1. Friction. Something feels wrong. A name, a character, a location, a plot point. The feeling is your artistic vision telling you the current version doesn't meet your standard. Don't ignore it.
2. AI-assisted divergence. Use an AI to generate options — many of them, across different directions. The goal is not to find the answer but to expand the possibility space beyond what you'd explore alone.
3. Human recognition. Look at the options and feel. Not analyze — feel. Your lifetime of influences, preferences, and creative instincts will tell you which option has actual resonance versus which ones just sound good. This step cannot be automated. It is the most important step.
4. Mythic synthesis. Take the recognized option and push it. Don't stop at the surface. Ask: what does this mean? What does it explain? What history does it imply? What emotions does it carry? This is where the real creation happens — the leap from a word to a world.
5. Structured integration. Use an AI to build the synthesis into your project systematically. Structure it. Cross-reference it. Propagate it. Make it real across every file and every reference. This is where scale happens — the kind of integration that would take a human days happens in minutes.
I'm calling this human-led mythic selection with AI-assisted divergence and structural integration.
The five-step flow:
FRICTION DIVERGENCE RECOGNITION SYNTHESIS INTEGRATION "This feels AI generates Human feels Human creates AI builds, wrong" 15+ options resonance the myth scales, and propagates [HUMAN] → [AI] → [HUMAN] → [HUMAN] → [AI] artistic option aesthetic mythic structural instinct space judgment leap execution ───────────────────────────────────────────────────────────────────────────────── TIME: minute 0 minute 10 minute 15 minute 20 minutes 30-90Three of the five steps are human. The two AI steps (divergence + integration) are where machines genuinely outperform humans: generating many options fast, and propagating changes across large systems. The three human steps are where machines genuinely can't compete: detecting dissonance, recognizing resonance, and making the creative leap.
It's a mouthful. But it's precise. And the precision matters, because the method only works if the human leads. If you skip step 3 — if you let the AI pick the option — you get something that sounds generated. The authored feeling comes from human taste, human refusal, and human synthesis.
What it produced
In a single session, a naming complaint became:
- A dragon god — Khaross the World-Burner, the last physical deity of the Draconic Pantheon
- A founding myth — the dragon wounded in the Shattering, choosing the desert as his resting place, his holy blood still burning beneath the earth
- A sacred city — Ash'Khaross, built inside the skeleton of a god, where every wall is reliquary and every corridor is shrine
- A theological question — is Khaross truly dead? The eternal flames respond to events. Some priests say his fire still watches.
- A pilgrimage tradition — Drakenith walking the Red Sands to find the bones of older dragons
- An oral tradition — the myth spoken around forge-fires, where the details shift with every telling but the ending never changes
- A 340-line document integrated across the entire worldbuilding project
- A name change propagated across 23 files — every reference in the codebase updated in one commit
All because a name sounded like Game of Thrones.
Session timeline:
Time Event Participant 0:00 "Does Vey'Drakar sound like Game of Thrones?" Human 0:02 Confirms GoT phonetic echo — Dracarys + Vaes Dothrak Claude Code 0:10 Generates 15 names across 3 linguistic directions ChatGPT 0:15 Selects "Khaross" — invents the founding myth on the spot Human 0:20 "The dragon was divine, his blood still burns, the city is named Ash'Khaross" Human 0:25 Builds complete 340-line city document Claude Code 0:40 Reviews document, identifies strengths, suggests enrichments ChatGPT 0:50 Integrates enrichments (vigil, pilgrimage, oral tradition) Claude Code 0:55 Renames file + propagates across 23 files Claude Code 1:00 All changes committed. Myth is canonical. All three Total time: ~1 hour. From naming complaint to foundational myth with full project integration.
By the numbers:
Metric Count Names generated by AI 15 Names selected by human 1 Lines in final city document 340 Sections in document 14 Lore elements created (myth, deity, tradition, theology) 6 Files updated across the project 23 Git commits 4 AIs involved 2 Humans involved 1 Creative leaps made by AI 0
That last line matters.
What this means if you're a writer
I'm not going to tell you that AI will transform your creative process. Maybe it will, maybe it won't. But I can tell you what I learned from this specific experience:
Friction is a creative signal. When something in your world feels wrong — a name, a motivation, a geography — don't just fix it. Investigate it. The wrongness is often pointing at something deeper that wants to exist.
Use different tools for different things. Brainstorming and structural integration are different cognitive tasks. Using one AI for divergent exploration and another for systematic building worked better than using either for both.
Your taste is the architecture. The AIs expanded the possibility space and built the structure. But the moment of meaning — the moment when a word on a list became a dying god whose blood still burns — came from something no AI has: a lifetime of caring about dragons. Your influences, your obsessions, your aesthetic sense — those are not inputs to the process. They are the process.
Selection is not just "picking." When you look at a list of options and one of them resonates, what's happening is not choice. It's recognition. You're recognizing the form that matches your internal vision — the one that sounds like it belongs in your world, not someone else's. That recognition is the most creative act in the entire workflow.
The method is repeatable. Friction → divergence → recognition → synthesis → integration. I've used variations of this across dozens of worldbuilding sessions over forty months. The Ash'Khaross session is just the clearest example because the leap was so visible: a naming problem became a dragon god in under an hour.
The real lesson
There's a version of this article that ends with something like "AI is a tool, humans are the artists." That's true but it's boring. Everyone says it. It doesn't tell you anything useful.
Here's what I actually think:
The reason Ash'Khaross feels like it belongs in Astrylis — the reason it doesn't sound generated — is that it was born from refusal. I refused a name that didn't feel like mine. That refusal came from decades of loving dragons and years of building a world. No AI has that history. No AI can replicate that feeling of this isn't right, but I don't know what is yet.
The AIs helped me find what was right. They gave me options I wouldn't have generated alone. They built the result into something far more complete than I could have managed in one sitting. They even told me why the result worked — which helped me understand my own creative process better than I did before.
But the thing that made the result feel authored instead of generated? That was mine. A lifetime of dragons. An instinct that said no. And a myth that appeared when the right word finally landed.
Ash'Khaross. The Ashes of Khaross. A city built inside the bones of the last dragon god, powered by holy blood that never stopped burning.
That's not AI creativity. That's not human creativity. That's what happens when they work together and neither pretends to be the other.
This article documents a real worldbuilding session from April 7, 2026 — part of the ongoing creation of The Ethereal Web, a fantasy saga 40 months in the making. The method described here — human-led mythic selection with AI-assisted divergence and structural integration — emerged naturally from the work and was named after the fact by all three participants.