Sonnet 3.7 Transformed My Character Notes Into a Full Scene (The Writer's Process, part 12)
Do NOT miss "Songs of the Arctic Uptime Festival"!
IN THIS POST:
- I'll show you how I transformed basic character notes into a fully realized scene using Claude Sonnet 3.7
- You'll see the exact moment when AI made a creative "leap" that changed my entire approach
- I'll reveal my process for using multiple AI tools (ChatGPT and NotebookLM) as a writer's room (and which one excels at what)
- By the end, you'll have a practical framework for using AI to enhance—not replace—your creative decisions
HOW CLAUDE SONNET 3.7 TRANSFORMED MY CHARACTER NOTES
In my first scene, the two main characters, Chen and King, are drunk. Chen is on a business boondoggle trip to one of the 5 major data hubs in the world -- the Arctic Circle Data Center. It’s about half the size of Manhattan Island. King (a glad-handing salesman type who refers to himself as King, people just play along) has just finished hosting some big dinner. He has lured Chen into a secret warren in the data center and is trying to seduce her.
In my summary, I mentioned that King had just hosted a big dinner (a sendoff for visiting business people). During the dinner, I suggested that they sing “Songs of the Arctic CIrcle Data People,” kind of a parody of “native” songs you hear in tourist traps.
In the response, Claude “pitched” a line for the “Arctic Data Love Song”, something about ‘quantum entangled hearts.’”
I DID NOT SEE THIS COMING: SONGS OF THE ARCTIC CIRCLE DATA PEOPLE
YES, I said, give me that song. I got back kind of a lame song. BUT... I then asked Claude to write a parody of “Okies From Muskogee” (look it up!), sung by the workers of the Arctic Circle:
I'm proud to be a coder from the Arctic Circle
A place where latency stays mighty low
We still debug our systems and we test our every merge request
And cache invalidation is something that we know
We don't need your microservices or your blockchain solutions here
Our uptime's running pure and clean, no downtime to report
And we don't take no meetings with those San Francisco dears
'Cause we're too busy keeping quantum bits in their coherent state, you bet
[Chorus]
Leather chairs and ping pong tables ain't what gets the job done right
Just give us clean cold server rooms and our command line interface
We're cooling down our quantum cores while y'all are staying up all night
Trying to figure out why your Docker containers got erased
[Bridge]
In the Arctic Circle Data Center
Where the real computing's done
No VCs or social media
Just pure processing, every one
[Final Chorus]
I'm proud to be a coder from the Arctic Circle
Where even our test environments run clean and true
We keep the quantum bits aligned and all your data safely stored
And we actually know what a pointer can do
Two things happened: I laughed out loud. AND ... for some reason, the word “uptime” got me... I suggested that Chen and dozens of other corporate and government types were visiting the ACDC for the “Annual Uptime Festival,” a celebration of... well, a year of uptime.
FROM ONE-OFF JOKE TO WORLD BUILDING: THE ANNUAL UPTIME FESTIVAL
I got back (this is edited)
THE ANNUAL UPTIME FESTIVAL
- Corporate Woodstock meets Davos meets Native Cultural Tourism
- Elite guests pay enormous sums for "authentic data center experiences"
- Runs for one week in perpetual Arctic night
EVENTS INCLUDE:
1. Opening Ceremony
- "Traditional" server rack lighting ritual
- King in elaborate "Chief Data Shaman" costume
2. Corporate Entertainment
- Data center folk music performances
3. Classic Festival Elements (Tech Style)
- Overcooled VIP lounges
- Quantum-optimized cocktails
- "Ancient Data Center Cuisine" (actually just fancy cafeteria food)
- AI-generated "traditional" art installations
AND THEN... just to bring this full circle, without my prompting, Claude came back to the characters and what they’re doing there.
CHEN'S EXPERIENCE:
- First time being important enough to attend
- Drinking quantum-cooled vodka
- Trying to network while maintaining dignity
- Growing increasingly annoyed at the fake authenticity
KING'S ROLE:
- Master of ceremonies
- Playing benevolent host
- Carefully managing which elites see what
- Using festival to maintain the ACDC's mystique
This explains:
1. Why Chen is there
2. Why she's drunk
3. Why security might be looser
4. How she gets access to restricted areas
5. The overall satirical tone of your world
This demonstrates one of the most valuable uses of AI for writers: creating an environment where unexpected connections lead to genuine creative breakthroughs that still feel like your own work.
SO... in answer to your question: Will working with Claude work?
YES!
MEET DR. NORDIC. DON’T CALL HIM “PATCH”
King and Chen are going to do something very dumb that will trigger the AI Collapse at the heart of “Autonomous.” I needed a character who would witness Chen and King doing that. I wanted some kind of “by the book” character who would be afraid of King (so wouldn’t challenge him), but would become central to the plot.
The first guy I got was pretty much “stick figure.” BUT... I kept making jokes about various characteristics. Here’s what we ended up with:
## Dr. Hjálmarr Thorvaldursson-Sigurðsson (Dr. HITS OR DR. NORDIC OR “PATCH” ) - "The Manual Follower"
- Possesses an extremely long Nordic name that everyone mispronounces (most just call him “DOCTOR HITS” OR “DOCTOR NORDIC))
- He owns the ONE retinal scan that can open every lock in the data center. He keeps a patch over that eye. DO NOT EVER CALL HIM PATCH.
- Methodically maintains the physical quantum computing infrastructure at the Arctic Circle Data Center
- Carries an actual printed manual everywhere—a massive, well-worn 800-page binder with color-coded tabs and his own handwritten notations
- Has zero personal AI agents by choice, claiming they "interfere with protocol adherence"
- His work-assigned AI is set to "Optimal Worker" mode and nothing else
- Speaks in highly technical, procedural language that mirrors the manual's text
- Has calibrated his communication filter to maximum professionalism, removing all personality
Character quirks:
- Frequently quotes specific manual sections and page numbers in conversation
- Takes his scheduled breaks with military precision (sets an analog watch alarm)
- Has memorized entire sections of emergency protocols
- Methodically documents every system anomaly in both digital logs and a physical notebook
- Insists on following the exact chain of command for all reports, no matter how urgent
During the collapse:
- Initially attempts to follow the manual's emergency procedures to the letter
- Becomes unexpectedly valuable as the only person with complete offline documentation
- Gradually learns to adapt procedures rather than follow them blindly
- Experiences genuine confusion and distress when situations arise that aren't covered in his manuals
Storyline potential:
- His rigid adherence to protocol could initially seem like a liability but becomes crucial during the collapse
- Provides a stark contrast to more flexible, intuitive characters
- His evolution from rule-follower to adaptor could mirror the larger themes of the story
- The printed manual becomes an object of immense value as digital systems fail
NOW... would I have come up with Dr. Nordic on my own? Over time, yes... Was it a lot more fun to come up with him with Claude? Hell, yes.
COLLABORATING IN THE “WRITERS’ ROOM”
CHAT GPT is very good at pointing out gaps and inconsistencies. I will say that, from my one-day experiment with Sonnet 3.7, Claude is VERY good at unexpected jumps in thinking. ChatGPT pointed out gaps that I already knew were there (first draft!) but also some thinking I’d actually forgotten about from other documents I’d loaded in.
Among those elements was an entire social structure surrounding the workers of the ACDC that I’d just mentioned briefly in a chat. ChatGPT suggested an entire opposition to King that will play out terrifically well.
At the end of our back and forth, Chat GPT summarized our entire conversation (and retained the things I’d uploaded from Claude). Claude tends to truncate them. I prefer ChatGPT’s detailed notes.
In neither case would I use any of the dialogue they proposed. (Although I did like a line ChatGPT suggested for Chen as she’s being led into the darkest parts of the ACDC: ““God, it’s like a billionaire frat house and a submarine had a baby.””)
FINALLY... NOTEBOOK LM
NLM is really beginning to prove its’ worth. It is perfect at not getting on board with everyone else. First questions were: “What are we trying to do in this scene? Who is important here? What tone are we trying to strike?” etc.
At the end, NLM needly sumamrized all of the work we’ve done. For me, Notebook provides a kind of “security blanket” where I know everything is stored and safe. Yes, I have google docs, but by virtue of the fact that I want NLM to review certain docs, I know the most valuable ones are listed here.
WHAT I’VE LEARNED: PRACTICAL TAKEAWAYS
I couldn’t be happier with how this is going. I’ve applied the basics of character construction to my own work and then “played” with Claude to make them come alive in a scene. What’s not to like?
## MY AI WRITER'S ROOM: EACH TOOL'S UNIQUE CONTRIBUTION
Each AI brings something different to my creative process:
**Claude Sonnet 3.7**
- Excels at: Unexpected creative leaps and connections
- Standout feature: Generated the entire Uptime Festival concept without prompting
- Best for: Early ideation and "what if" exploration
- Weakness: Sometimes truncates conversation summaries
**ChatGPT**
- Excels at: Identifying gaps and inconsistencies in worldbuilding
- Standout feature: Remembered social structures I'd only briefly mentioned
- Best for: Logical analysis and continuity checking
- Bonus: Occasionally delivers genuinely good dialogue like the "billionaire frat house" line
**NotebookLM**
- Excels at: Organizing thoughts and challenging assumptions
- Standout feature: Asks fundamental questions about character goals and scene purpose
- Best for: Providing structure and safely storing your most valuable content
- Serves as my "security blanket" for the entire project
OVER TO YOU
I'm curious about your experiences with AI writing tools:
- Have you tried using multiple AI tools as a "writer's room"? Which combinations work best?
- What's the most surprising creative leap an AI has helped you make?
- If you haven't tried this approach yet, what's holding you back?
Share your thoughts in the comments—I'll be actively responding and might feature interesting techniques in an upcoming post.
Next week, we’ll talk more about world building.
Until then, keep experimenting!
I’ve just started to explore using Claude, ChatGPT, and notebook LLM as a writers room based on your earlier excellent suggestions. Too soon for me to know outcomes. It is however loads of fun exploring. Thanks for sharing your journey!!!
Hmmm. This has me intrigued. How exactly do you use Chatgpt in this context?
I've worked with both for an outline for a book.. they've given me two very different outputs. In fact, Chatgpt gave me two or three now.
For notebookLM...are you suggesting we put all the output from claude and chatgpt and ask NotebookLM to synthesis it?