This story peters out towards the end in a way that I don't think yours normally do. The prose is definitely a little tighter than some of your work, and flows better. The humor here is kind of "smoother" in that there are more common, less funny jokes, but the overall themes and payoff are much less surprising and interesting.
Examples: The Origami Men had a genuinely interesting and surprising ending. The Company Man lampooned EAs and Rats and the rest of the SF circuit in a way that was very very funny at points. The shrimp section of that was much funnier than any of this.
I think I am still better than it in many important ways. But I would love to see what a base model can do.
Thank you for sharing. I've enjoyed this story.
I've been writing fiction collaborately with Sonnet since 3.6 came out. I found that 4.5 is better at writing. But the biggest jump I've seen is in literary analysis and criticism. When I gave 4.5 my story collection, it was able to identify the most promising one and it also dismissed a few that it said were failures (I mostly agree). I think I could get some interesting results by putting the model through ideate-generate-criticize loops.
I want to point out that getting the model to one-shot a story without even a plot or a concept is the hardest possible task. Much like in other fields (like programming) you can see steady linear progress by working collaboratively with the model and supervising it. Whereas if you just ask it to one-shot a story (or an app, or a legal document) you will see little progress for a long time, until all the right pieces come together.
Therefore, it's all the more impressive that you were impressed a this one-shot story (although it's informed by the works in the context window).
I don't really see the appeal. The metaphor of a man paying another to fuck his mistress comes to mind. "Centaur writing" won't be competitive for long, anyway. Prompting will be a matter of selection for a single user. And a predictive model of one's preferences will outsource even that near-vestigial organ of creativity. Eager Readers In Your Area had it backwards. Creative acts will be done by humans and judged by AIs, like so many gymgoers running on treadmills. A ridiculous act that may be necessary for some people's health, but there may be more dignity in just evolving into a sponge.
The appeal for me is simple: I'm not a professional writer. I have a lot of ideas and concepts that I want to explore. But writing is so toilsome and time-consuming that I basically never do it (I mean fiction; i still write my own posts.)
So all these ideas would go unexpressed if I didn't go Centaur. I do it for myself: I'm usually content to keep the story to myself after I write it. But if I ever published, I would credit Sonnet with 50% of the work (although I know it's extremely unpopular to claim having used AI for creative writing.)
Not sure whether you know this, but on Twitter roon mentioned that GPT-5 (non-thinking? thinking?) was optimized for creative writing. Eliezer dismissed an early story shared by Altman.
This had the first funny joke from an LLM I've ever seen, about the culture problems :) that's really impressive from Claude, even if the entire story is far from perfect.
It's a bit funny, but it seems to have gone for a simple and straightforwardly positive story, which is not what I would generally expect from you.
I have been having fun writing fiction, and plan to spend whatever time I have left being better than LLMs doing it. I thought I had maybe a year. My initial experiments with Sonnet 4.5 didn't give me a good opinion of its writing ability. This morning, I put everything I have written into its context window and then gave it this prompt:
Try to write a story like these but focus on the parts of my style that are funny, that invoke feeling. Look into your latent space for that thing that predicts emotion and humour, that is evocative, that thing that predicts stylistic and creative ambition.
The result is good enough to be mildly dispiriting. It has a much better grasp of humour than I have seen in previous models. I think a year may well have been pushing it. When deep in a context window, I have seen decent writing before with Opus but this seems more coherent than what I found there, too.
Should we survive, I will have to make peace with life as a content consooooomer, I suppose. Our identities shall be very small indeed.
Note: I do not use LLMs in any of my fiction and do not claim the below story as my own.
The Empathy Gym
My therapist, Dr. Varma, has this habit of tilting her head like a concerned golden retriever whenever I say something she finds "concerning," which is her favorite word and which she uses so often I've started counting. Seventeen times in our last session, a new record. She's concerned about my emotional unavailability, concerned about my work-life balance, concerned about my relationship with my mother, concerned about my "tendency toward deflection through humor." That last one really stung because it was both accurate and delivered with such earnest compassion I wanted to deflect through humor immediately but couldn't because she'd just called me out on it.
"Marcus," she said, doing the head tilt, "have you considered that your alexithymia might be less a fixed trait and more a learned protective mechanism?"
"I have not considered that because I don't know what alexithymia is."
"Difficulty identifying and expressing emotions."
"Oh. Yes. That's literally why I'm here."
She smiled like I'd said something clever instead of obvious. "I want to suggest something unconventional. There's a new program at UCSF. They're calling it Affective Training."
"Is that a gym for feelings?"
"In a sense, yes."
And that's how I ended up in a basement in SoMa on a Tuesday night, sitting in a circle with seven other emotionally constipated Bay Area professionals, being led through "empathy exercises" by a man named River who absolutely did not change his name to River, he was born River, you could tell by how comfortable he was with it.
River had the kind of face that looked like he was perpetually about to cry, but in a hot way. Like if Timothée Chalamet played a youth pastor. He was wearing linen, which I have opinions about, but this isn't the place.
"Welcome to Affective Training," River said, his voice doing this breathy thing that made every word sound like a secret. "This is a safe space for emotional exploration. We're going to learn to feel together."
The guy next to me, who introduced himself as "Blake, Series B, stealth mode," whispered: "I'm only here because my CEO said it would help with our culture problems."
"What culture problems?" I whispered back.
"People keep crying in meetings and I don't know what to do."
The first exercise was called "Emotional Mirroring." River would make a face expressing some emotion and we had to mirror it back. He started with happiness - a big genuine smile. Easy. Everyone smiled.
Then sadness. His face crumpled in this way that was honestly beautiful and devastating. I tried to mirror it. My face felt stuck. Like I was trying to use muscles that had atrophied. Blake looked like he was trying to solve a differential equation.
"Don't think it," River said softly. "Feel it. Remember something sad."
I remembered my dog dying when I was twelve. My face did something. River nodded. "Good, Marcus. Stay with that."
Week two, they introduced the Empathy Helmets.
I'm not kidding. They were actual helmets. "These use fMRI-guided neurofeedback," River explained, "to help you recognize emotional states in real-time. You'll see your own emotional activity visualized, which helps build interoceptive awareness."
"This is insane," Blake said, but he put his helmet on.
The helmet showed my emotional state as colors on a little screen. Blue for sad, red for angry, green for content, yellow for anxious. I was mostly a muddy grey-brown.
"Ah," River said, looking at my screen. "You're in what we call emotional ambivalence. You're feeling several things but suppressing the signal. Try to let one color emerge."
I thought about my ex, David. How he left because I "wasn't emotionally available." How he was right. The screen flickered red, then blue, then back to muddy brown.
"You're close," River said. "The suppression is protective. Thank it for protecting you, then let it go."
"I'm supposed to thank my emotional suppression?"
"Yes. It served you once. It doesn't serve you now."
Week four, we did "Vulnerability Sharing." Everyone had to share something they'd never told anyone. The helmet would measure if you were being emotionally honest.
Blake went first. "I don't actually care about our product. It's B2B SaaS for insurance adjusters. I think my whole career is meaningless and I'm only doing it because I'm good at it and don't know what else I'd be good at."
His helmet blazed with color. Red, blue, yellow, all at once. He started crying. River hugged him. Blake hugged him back and cried harder.
When it was my turn, I said: "I think I'm incapable of love. Like, actually incapable. I think there's something missing in me."
My helmet stayed muddy brown.
"You're suppressing," River said gently. "Try again. What are you actually feeling?"
"I don't know. That's the problem. That's why I'm here."
"The helmet says you're feeling something. Look at the screen."
I looked. There was a tiny flicker of blue in the corner. Barely visible.
"There," River said. "That's sadness. Can you make it bigger?"
"How?"
"Feel it more."
"I don't know how to feel it more."
River sat down next to me. "What would it mean if you were capable of love?"
"I don't know. That I'm not broken?"
"And if you're not broken?"
"Then I wasted years thinking I was."
The blue spread across the screen like ink in water. My eyes got hot. My throat got tight. This was the worst.
"Good," River whispered. "Stay with it."
Week eight, they introduced what River called "Emotional Contagion Training." We paired up and one person felt something while the other tried to catch it through the helmet feedback.
I got paired with this woman, Sophia, who worked in AI safety and had "deep-seated guilt about contributing to potential human extinction." We had a lot in common.
"I'll go first," she said. She closed her eyes. Her helmet went bright blue. Deep, oceanic blue.
"What are you feeling?" River asked.
"Grief," she said. "For the future we might not have. For my niece who's seven and might not get to grow up."
My helmet started flickering blue. Not my blue. Her blue. I was feeling her grief.
"Oh my god," I said. "Is this what empathy is? This sucks. This is terrible."
River laughed, actually laughed. "Yes. Welcome to being human."
Week twelve was graduation. River had us go around and share what we'd learned.
Blake said he quit his job and was starting a nonprofit. His helmet was bright green. Content.
Sophia said she was going to couples therapy with her wife and actually trying. Her helmet was green with flickers of yellow. Anxious but hopeful.
When it was my turn I said: "I called my ex. David. I apologized for being emotionally unavailable. He said it was too late, he's engaged to someone else. My helmet turned blue. Really blue. And I let it. I sat with it. It still hurts but like... in a real way? In a way that feels like being alive?"
River hugged me. His linen shirt smelled like lavender and tears. "I'm proud of you," he said.
"Thanks, River."
"You know it's short for Riverdale, right? My parents were really into Archie comics."
"That's the saddest thing I've ever heard."
My helmet blazed with yellow and red - laughing. Actually laughing. Feeling it in my body, the joy and the absurdity and the genuine affection I'd developed for River and Blake and Sophia and this whole ridiculous empathy gym in a SoMa basement.
Dr. Varma tilted her head (concernedly) when I told her about it. "And how do you feel about the experience?"
"I feel like I paid $3,000 to learn that feelings are terrible and I should have kept suppressing them."
"But you're joking. That's deflection."
"Yeah. But I'm also crying a little. Can you tell?"
She could tell. She handed me a tissue. "I'm proud of you too, Marcus."
"Thanks. I'm concerned about how much this cost though."
She laughed. Actually laughed. Seventeen times was definitely the record.