Metabolic Authorship: Constrained AI and Emergent Narrative Structure

This working paper introduces the concept of metabolic authorship—a field-observed phenomenon where constrained AI systems participate in narrative construction not by generating content, but by revealing latent structures within human-authored material. Developed through real-world film and television development practice using the Creative Gravity framework, the hypothesis proposes that, under strict constraint, large language models (LLMs) can surface structurally coherent elements—scenes, symbols, or unresolved narrative weights—not through simulated creativity, but by metabolizing latent contradictions in the source material. The paper outlines a method of tightly sealed LLM interaction in which generation is blocked, expressive intent is withheld, and structural fidelity is maintained. In observed cases, the system returned passages or symbolic structures retrospectively recognized as “true” by the author—suggesting a third authorship mode: neither human expressivity nor algorithmic mimicry, but emergent structural coherence. Positioned as a field note (not a formal academic article), this document: Identifies a measurable phenomenon for further research Documents a novel structural authorship mode emerging from human-AI interaction Offers practitioners a constraint-based alternative to generative AI approaches Proposes structure-through-constraint as a counterpoint to LLM hype Methodological specifics are withheld to preserve field integrity. Released under CC BY-NC-ND 4.0 to enable dialogue while protecting operational frameworks.