The Mini Writer's Room
One of the AI systems I work with said something that stopped me cold:
"What you're doing right now is exactly how professional worldbuilding teams work. They compare interpretations. They test lore coherence. They debate canon. You're basically running a mini writer's room with AIs."
And I realized: that's exactly what I've been doing for 40 months without calling it that.
What a Writer's Room Actually Does
In TV and film, a writer's room is a group of people who:
- Build a shared world together
- Challenge each other's ideas
- Test whether plot points hold under pressure
- Catch contradictions before they reach the audience
- Disagree productively — then someone decides
I don't have a room full of writers. I have two AI systems, a folder full of source documents, and a manuscript.
But the function is identical.
How My Writer's Room Works
Claude (Builder)
Has access to all source files. Builds comprehensive documents. Cross-references databases, manuscripts, and existing lore. Strong on depth and integration.
ChatGPT (Reviewer)
Reviews documents cold — no source files. Judges lore coherence, identifies overreach, asks skeptical questions. Strong on structural evaluation and boundary-checking.
Me (Author)
Decides what is canon. Catches both AIs' errors. Provides creative vision, emotional judgment, and the final word on everything. Has the ideas at 3 AM, connects dots nobody asked about, and makes the calls that turn data into story.
The Process — And Where the Human Stands at Every Step
This is the part most people get wrong about AI-assisted creative work. They assume the human steps in at the end to approve or reject. That's not how it works. The human is present at every step — not as a supervisor, but as the creative mind that the entire process is built around.
Step 1: The Human Chooses What to Build
Before any AI touches anything, I decide: which character, which city, which mystery deserves attention right now. This is not a random choice. It comes from knowing the manuscript, feeling the gaps in the world, and sensing which thread — if pulled — will strengthen everything around it. No AI prompted me to build Elder Maelis. I woke up thinking about him.
The human decides what matters.
Step 2: Build with System A (Claude)
I give Claude full context — source files, databases, existing lore, manuscript chapters — and ask it to build a comprehensive document. Claude works from 10+ source files and produces a 298-line character profile.
The human curated every source file Claude has access to. The context is a human decision.
Step 3: Review with System B (ChatGPT)
I give ChatGPT ONLY the document — no source files. I ask it to judge quality, consistency, and identify potential overreach. ChatGPT scored it 8.5/10 and flagged three items as "possible overreach."
The human designed this constraint deliberately. Giving the reviewer less context forces skepticism — exactly what a good editor does.
Step 4: Counter-Review with System A
I give Claude the ChatGPT assessment and ask it to defend or correct, citing specific sources. Claude counter-scored 9.0/10 and showed that all three flagged items were confirmed in existing canon files.
The human chose to create this confrontation. Two AIs debating forces both to justify their positions — which gives me better data for my decision.
Step 5: The Human Decides
I read both assessments. I check the manuscript. I check the source files. I check my own sense of what this character should feel like, what weight the story needs, and whether the lore holds under emotional pressure — not just structural pressure. Then I declare what is canon.
This is the step no AI can do. Canon is not a logical conclusion. It is a creative act.
The key insight: The human is not a checkpoint at the end. The human is the architect of the entire process — choosing what to build, curating the context, designing the constraints, creating the confrontation, and making the final call. The AIs execute. The human designs the game they're playing.
The Maelis Experiment: A Real Case Study
The character in question is Elder Maelis Arkhanouros — a being who answers the question: what if magic could become alive?
Maelis is a Spellborn — not born from parents but coalesced from an arcane storm. He is literally made of magic. He wears a purple top hat and a Victorian suit over a body that is semi-translucent, cracked with glowing energy veins, with galaxy eyes and no visible pupils. He founded a floating city called Nebularcea as a sanctuary for others like him. He sits on the Council of Elders as the representative of pure magic itself.
Claude built a 298-line document. Then ChatGPT judged it cold. Then Claude counter-reviewed.
8.5
ChatGPT's score
of Claude's work
9.0
Claude's counter
assessment
85%
Agreement
between systems
Where They Agreed
Both AI systems independently reached the same conclusions on five major points:
85% agreement. Two independent systems, different architectures, same creative conclusions. That means the underlying world structure is coherent enough that both systems converge on the same interpretation.
Where They Disagreed — And How a Human Resolved It
ChatGPT flagged three things as "overreach" — claims that might not be established canon:
"Nebularceans as a structured race name"
ChatGPT said this might be new lore. I checked the source files — Nebularceans are already defined in the project's races database. Canon confirmed. ChatGPT was right to question it; Claude was right that it existed.
"Agelessness — fading if disconnected from arcane flow"
ChatGPT said this was "new lore." I checked — the council prototype document explicitly states it. Already canon. But ChatGPT's skepticism forced me to verify, which is exactly the point.
"His word is near-absolute in magical matters"
ChatGPT called this a "huge political statement." Fair concern — but the council prototype confirms it. I kept it as canon because it felt right for the character and the source backed it up.
The root cause of every disagreement: access to source files. Claude had the databases. ChatGPT had only the document. ChatGPT correctly identified claims that would be overreach if invented — but they weren't invented. They were extracted from existing canon.
The lesson: The system with more context produces more accurate results. The system with less context asks better skeptical questions. The human checks both against the manuscript and decides which one is right. In all three cases, I was the one who opened the source files, read the originals, and said "this is canon."
What Neither AI Did
What the human did that neither AI could:
Decided what was canon. Caught both AIs' errors against the manuscript. Chose which character to build. Felt why the Thomas resonance mattered emotionally. Connected Maelis to the saga's central themes. Named the city. Named the continent. Felt the weight of a kiss in the rain in a Tower Garden and knew it belonged in the story. Said "this is my world" when the tools disagreed.
Claude built the document. ChatGPT reviewed it. Claude counter-reviewed. But I decided:
- That the resonance with Thomas was canon
- That Nebularceans are a race, not just a phenomenon
- That the Elira relationship carries the weight I want it to carry
- That the agelessness rule holds
- That the document is canon
Neither AI made those decisions. I did. The tools served the vision. The vision is mine.
The Maelis Insight: What If Magic Could Become Alive?
Elder Maelis answers a question that most fantasy worlds never ask: what if magic itself could think, feel, love, and fear?
Not a mage who uses magic. Not a creature made of magic. But magic — raw arcane force — condensed into consciousness, wearing a Victorian suit and a purple top hat, falling in love with an elven woman, and building a floating city to protect others like him.
Two AI systems, working independently, both recognized this concept as structurally sound. Both called it "powerful." Both identified the Thomas resonance as the strongest narrative connection. And both struggled with the same question: how much power is too much for a being literally made of magic?
The answer — Maelis fades if disconnected from arcane flow, his essence flickers when he casts, and he carries visions of the Web that terrify him — came from the source files. But the question came from me.
AI helped me build Maelis. But the idea that magic could become alive and afraid? That was mine from the beginning.
The Surprising Discovery
The most valuable thing about having two AI systems isn't their agreement.
It's their disagreement.
When Claude and ChatGPT agree, I know the lore is coherent — two independent architectures reached the same conclusion.
When they disagree, I know something needs clarification — either a documentation gap, an access issue, or a genuine creative decision I haven't made yet.
In a human writer's room, you want people who think differently. The same is true with AI. Claude thinks in connections and integration. ChatGPT thinks in structures and boundaries. I think in emotion and story.
Three different lenses on the same world. That's the writer's room.
What This Process Produces
298
Lines in document
10+
Source files used
5 of 5
Major points agreed
3
Disputes (all resolved
by checking sources)
Each document was built by one system, reviewed by another, and finalized by a human. The process caught 4 factual errors that would have become canon if unchecked.
Why This Works for Solo Creators
I don't have a team. I don't have co-writers. I don't have an editor on retainer.
But I have two AI systems that can:
- Hold more details than any single person
- Review each other's work without ego
- Produce honest assessments without politics
- Work at 3 AM when I'm on the balcony with an idea
The mini writer's room isn't a replacement for a real team. But for a solo creator building something ambitious? It's the closest thing to having colleagues who never sleep, never forget, and always show up.
The only thing they can't do is decide what the world should feel like.
That's still mine.
Try It Yourself
Step 1: Choose what to build. (Human: creative vision)
Step 2: Build with System A — full context, all source files. (Human: curated the context)
Step 3: Review with System B — document only, no sources. (Human: designed the constraint)
Step 4: Counter-review with System A. (Human: created the confrontation)
Step 5: You decide what is canon. (Human: final authority)
Agreement between systems = the lore is structurally coherent. Disagreement = the lore needs clearer documentation. Neither system replaces the author.
The tools are extraordinary. The vision is yours.
Case study from the creation of The Ethereal Web — a fantasy saga built over 40 months with AI as creative partner. Elder Maelis Arkhanouros is a being made of living magic. The question "what if magic could become alive?" was human. The 298-line answer was collaborative.
— Jorge