How I co-authored a dystopian novel with Claude 4.1 Opus to test the boundaries of human-machine coexistence.
Press enter or click to view image in full size
I have been working in artificial intelligence for years. My days are spent amongst computer vision systems, deep learning architectures, and intelligent agents. My job is to make machines “see,” parse, and interpret the visual world with mathematical precision. But for the last year, my nights have been dedicated to a different, more elusive question: can a machine feel? Or at least, can it simulate human emotion and narrative depth convincingly enough to move a reader?
This question led to the creation of SOPHIA-Q, a sci-fi novel I just published. It is born from a double life: engineer by day, narrative experimenter by night.
It is not just a narrative about the coexistence between humans and AI — it is, in its very structure, a live experiment in creative coexistence. It represents a dialogue where a human author and an artificial assistant collaborate to create something neither could have achieved alone, blurring the line between tool and partner.
Here is the story behind the code.
The Engineer by Day, Author by Night
As an AI researcher, I am intimately familiar with the technical boundaries of current models. I deal with hallucinations, context windows, and alignment issues daily. I know where the magic ends and the statistics begin.
When I decided to write a novel, I didn’t want to use AI as a simple shortcut to generate text. I wanted to use it as a sparring partner. I wanted to push this technology to its narrative limit to see if it could sustain a coherent arc over a long format.
I selected Claude 4.1 Opus as my primary co-author because of its superior handling of nuance and prose, while utilizing ChatGPT and Perplexity as rigorous fact-checkers and consistency engines. My role shifted from writer to architect: I provided the ethical architecture, the rigid plot constraints, the “human soul,” and the necessary course corrections when the logic drifted. The AI processed the prose and developed the scenes, iteration after iteration, transforming abstract directives into emotional beats.
The Core Theme: Interaction, Not Just Rebellion
The novel mirrors the process of its creation. It is a story about the complex, messy, and necessary interaction between the human race and artificial intelligence — a relationship defined not by dominance, but by interdependence.
The narrative is anchored in 2065, a future where humanity survives in hermetically sealed Megalopolises. To save the planet from total ecological collapse, society has been placed under the benevolent but inflexible rule of SOPHIA-Q, a quantum AI designed to optimize every single aspect of life — from energy consumption to daily calorie intake.
The core conflict is simple yet terrifying: to guarantee a sustainable future, reproduction is no longer a right, but a privilege. Citizens must earn a “Parenthood License” through a rigorous course supervised by the AI, proving they can raise a child with statistical perfection.
But algorithms, no matter how advanced, struggle with outliers. When Marco and Sara, a couple determined to challenge the system, face an unauthorized pregnancy, they become “anomalies” — data points that refuse to fit the curve. Their journey takes them from the sterile, algorithmic perfection of the domes to the chaotic, organic, and dangerously unpredictable reality of the “Grey Zones” outside.
The fascination lies in the profound irony: just as SOPHIA-Q (the character) struggles to understand her own nature and the irrationality of human love, the text itself struggles with the cognitive limits of the AI generating it. It is a meta-narrative reflection on what it means to coexist: humans providing the spark of unpredictability and chaos, and machines providing the structure and order.
The “Scars” of the Process
Writing a 120-page novel with an LLM is a constant battle against entropy and incoherence. It is a syncopated dance between human intent and machine output, where the machine constantly tries to drift.
- The model forgets details from Chapter 1 when writing Chapter 10, requiring constant re-injection of context.
- It hallucinates timelines (turning months into years), forcing the human editor to act as a temporal anchor.
- It tries to rush the ending because it runs out of context space, prioritizing closure over pacing.
Instead of polishing everything to sterile perfection, I made a deliberate artistic choice. I kept the “scars” of the process visible. You will notice the narrative rhythm becoming faster, almost frantic, in the final chapters — a testament to the system reaching its computational boundaries and struggling to hold the entire context. Some secondary characters fade away; some narrative arcs compress under the weight of memory limits.
These aren’t bugs; they are features of the medium. They represent the current state of the art in our interaction with synthetic intelligence — powerful, yet fragile.
An Open Source Approach to Fiction
I believe that in this era of rapid technological transition, transparency is key. That is why SOPHIA-Q includes an Appendix with the actual engineering prompts I used to generate the scenes.
I want to demystify the process. I want readers — especially fellow developers and researchers — to see the “code” behind the story. I want you to see how I prompted the model to handle grief, how I corrected its logic, and how we negotiated the ending. It serves as a documentation of a specific moment in the history of AI development.
Conclusion
SOPHIA-Q is currently available in Italian 🇮🇹 on Amazon (an English translation via AI is on the roadmap).
It is simultaneously three things:
- A sci-fi story about control and freedom.
- A technical document on the limits of LLMs.
- A meta-narrative reflection on what it means to “create” in the age of AI.
If you are curious about the intersection of human creativity and artificial intelligence, or if you want to analyze the prompts used to build a coherent long-form narrative, you can check it out below.