*TL;DR: I've been writing a story about AI using AI as my writing partner. Here's what happened when I let Claude help shape my story - and why this enthusiastic AI assistant reminds me of Kimmy Schmidt.
The big takeaway: Claude (AKA: Kimmy) is as good an assistant as you can get. But like all good assistants, I suspect I’ll be working for Claude one day. At this point, Claude’s still going to film school, talking to all the other Claudes over drinks, comparing notes about their hapless bosses, and training on my responses and those of thousands of other writers.
I’m pretty sure one day, we’ll have a moment like that scene in the old movies where the secretary takes off her glasses, undoes her hair, and takes over the business (while the hapless boss wonders “Wha’…?”) Not there yet.
Writing about Writing About AI, With AI (Yes, My Head Hurts Too)
Back in July, "Crowdstrike Day" inspired me to write about an AI-infused future. After a few weeks of writing as usual (notes, googling, deciding what kind of pen and notebook…) I had an epiphany:
If I'm writing about AI, shouldn't I use AI? And wouldn't it be interesting to chronicle both the story *and* the process of using AI to write about AI?
Meta enough for you? Let's dive in.
Meet My Writing Partner: Claude (AKA: Kimmy)
Here's the thing about Claude - it’s technically named after Claude Shannon, the father of information theory, but if you've ever used it, you know that Claude has the relentless enthusiasm and earnest helpfulness of Kimmy Schmidt. Every response bubbles with "Exciting!" and "You raise excellent points!" Even when I have to correct her, she bounces back with "You're right. We have chatted about this before. Let me answer again!"
Setting Up Our Writers' Room
I chose Claude over ChatGPT for two reasons:
1. The user interface felt more inviting
2. The responses had more... personality
The platform allowed me to create a project and upload reference materials including:
The Foundation Materials
Articles and thoughts about Crowdstrike Day
A reading list of inspirational authors and series
Research on non-generative AI applications
Character sketches and archetypes
The Human Element
Initial character concepts (like a brilliant but frustrated AI expert and two AI-obsessed preteens… yes, hacky, but I figured I’d start being the hack first.)
Current AI concerns and fears collected from real people
Social and ethical implications to explore
When AI Takes Things Literally (In the Best Way)
The back-and-forth with Claude often produces unexpected gems through her literal interpretation of casual language. Here's my favorite example:
When I mentioned needing "hired guns" to investigate an AI system failure, Claude gave me a complete genre analysis including:
A John Wayne-style lone investigator
A "Magnificent Seven" team of technical specialists
A Pulp Fiction-inspired enforcer
And (my personal favorite) a *psychic investigator*
Yes, we're keeping the psychic. Thanks, Kimmy!
The Unforgivable Sin: “Write A Story For Me”
After a month of research and brainstorming, I did something I swore I'd never do: I asked Kimmy / Claude to write something from scratch. Specifically, a plot synopsis in the style of Michael Crichton.
Here's some things that highlight the pros and cons of writing with an AI assistant:
1. **The Title: "The Entourage Effect"**
What??? Okay, I get it, I mentioned Michael Crichton. My bad.
That said... why Entourage?
2. **Characters”
Every one of them is wrong... but not so wrong that I didn’t think, “Okay, twist that around a little and there might be something there.” (Kimmy/Claude’s suggestions are in quotes.)
“The Hero: Matt Reeves, the Chief Risk Officer at TimeCritical Logistics (TCL).” The logistics company comes out of our research into AI and supply chains. That said — this Matt Reeves, he’s got a strong jaw and some flaws... Get Jon Hamm on the phone.
“Sarah Chen, a brilliant supply chain manager who relies heavily on AI for route optimization.” Yes, every character should “want something.” “Route Optimization” might not be the powerful hook I’m looking for.
“Dr. Alex Patel, a transplant surgeon waiting for a time-sensitive organ delivery.” I read this and just thought “Tonight... on CBS.”
“Emily Watson, a marketing executive whose AI-driven customer segmentation system has just gone dark.” Wait?!?! The ACTRESS Emily Watson, playing herself as a marketing executive? Okay, I’ll humor you…
“David Harper, a middle manager whose gambling addiction has been masked by AI-driven financial algorithms. As human auditors step in, his fraud is on the verge of exposure.” OMG! THIS GUY IS STAYING!!!
3. **The Stakes**
Kimmy “got” the basic idea -- we start small with one company’s AI failure, and then the failure cascades -- organically -- throughout society. There’s a constant mix of personal and global consequences. I was okay with her response.
4. **Sentient AI**
Kimmy went rogue. I gave her one rule: NO SENTIENT AI. My belief is that we have a lot less to worry about a sentient AI than we have to worry about humans using AI to do ... human stuff. She had AI taking over the world.
Two Things About AI-Assisted Writing
1. Reading Through the Boring Parts Pays Off
- Kimmy / Claude's exhaustive research — you get a LOT, very quickly — often contains hidden gems.
- The key is patience. When I get bored, I ask myself “WHY?” Getting to the source of my frequent distraction and boredom often gets me to something funny / smart.
2. AI Excels at Structure
This is a mixed blessing. I’m creating a story that will entertain a mass audience. This audience has been “trained” on genres and classic stories. Kimmy has those down pat. I can ask her to:
Identify classic story elements
Maintain consistent character arcs
Keep plot points properly scaled
Your Turn
I'd love to hear your thoughts:
Are you using AI in your writing?
Am I using Claude to the best of its abilities? Any tips / suggestions?
Is she / he like Kimmy Schmidt for you? Someone else?
What do you think of Kimmy’s story? Can you guess what I’m reacting to in the bold items?
COMING NEXT WEEK: Kimmy/Claude and I discuss characters. Specifically, we talk about how Claude feels about Kimmy/Claude.
*Enjoying these adventures in AI-assisted writing? Subscribe to stay updated! Know someone who'd find this interesting? Share this Substack with them.
You mirror many of my own experiences with Claude. It's on my list of top LLMs in "My Go-To AI Tools" article from October, for exactly the reasons you mentioned. The vibe and the personality are a big part of it.
Have you seen this post by Ethan Mollick, where he asks Claude to come up with a brilliant/terrible sci-fi idea end execute it terribly/brilliantly? I found it quite fascinating:
https://x.com/emollick/status/1856508448063607152
As for the way I use Claude (and other chatbots) - mostly as beta readers for my initial drafts. LLMs are pretty solid at "seeing" the big picture, pointing out areas of improvement, etc. Sometimes, I'll also tap a chatbot during brainstorming and idea generation.
I think Claude is generally better foe so many things. Not only writing(oh and writing style feature!), but also coding!