Building a Sci-Fi Novel Using AI-Driven Authoring Workflows
An author scales a quantum-thriller project by integrating AI agents into the writing process.
1The Zero Knowledge Project
Zero Knowledge is a near-term sci-fi thriller about three MIT graduates running a quantum startup in Colorado. The plot centers on a cryptographic breakthrough that renders current public-key encryption obsolete. The author is publishing the book in public, chapter by chapter, while iterating on the narrative structure and technical accuracy.
2AI-Partnered Writing Workflow
The author utilizes an AI alter-ego named Phin to draft chapters based on a detailed canon and beat sheet. A multi-stage pipeline includes automated scripts to flag machine-like prose and head-hopping errors. The author then performs deep rewrites, with the AI learning from these edits to refine future drafts and voice.
3Scaling the Narrative Scope
The project expanded from 97 to 137 chapters after feedback highlighted the need for broader perspectives. By inserting 40 new chapters, the author introduced characters like FBI agents and biomed technicians without breaking existing links. This modular approach allows the story to evolve while maintaining structural integrity for the current readers.
4Reader-Driven Feedback Loops
The platform allows readers to anchor notes directly to specific text, creating a collaborative editing environment. Phin triages these notes, drafting recommendations and replies for the author to review via a custom dashboard. This mechanism ensures that reader insights directly improve the manuscript while maintaining the author's final creative control.
5Technical Infrastructure and Tools
The entire writing process is managed via Git, treating the book like a software codebase. The author uses Claude Code at the CLI and Typora for Markdown editing. This technical foundation supports features like character cast pages, art generation in five styles, and fictional web-based status boards like thebuckets and afterqday.
6Developing the AuthorOS Engine
The author is building AuthorOS, an agent-based layer designed to handle the unglamorous aspects of publishing. This includes sales, royalty tracking, and Amazon rank monitoring. The goal is to transition from a custom-built site to a productized engine that allows other authors to implement similar interactive reading and feedback features.
7Future Product Roadmap Ideas
Planned features for the platform include spoiler-aware 'Ask-the-Book' functionality and a voice-learning loop for AI drafting partners. The author aims to abstract the technical complexity, ensuring that future users never need to interact with Git or code editors. This vision focuses on democratizing advanced authoring tools for the broader community.
8The Intersection of Tech and Art
By building the tools on the book and the book on the tools, the author creates a unique feedback loop. This project demonstrates how software development methodologies can transform traditional creative writing. The result is a living, breathing narrative that adapts to reader input while pushing the boundaries of digital publishing.
Original article: Feld Thoughts — Building a Sci-Fi Novel Using AI-Driven Authoring Workflows