The Technology of Generative Fiction
Underlying technical stack: language models, multimodal models, agent architectures, evaluation, infrastructure.
Clusters
Language models for fiction
Model comparisons, temperature and creative output, fine-tuning vs. prompting, output length, context economics.
The fiction generation stack
Layer-by-layer reference: RAG, embedding-based memory, prompt-chaining, JSON-structured output, banned-word lists.
Multimodal: image generation for fiction
Character consistency, image model comparisons, reference-image workflows, LoRAs, style transfer.
Multimodal: audio for fiction
TTS comparisons, voice-cloning ethics, AI dubbing for translated fiction, the audio narration frontier.
Agent architectures for storytelling
Plan-then-execute vs. single-pass, multi-agent character simulation, why interactive fiction is harder than generative.
Evaluation & quality measurement
Frameworks for measuring fiction quality, banned-phrase metrics, AI tells, inter-model evaluation, the Nuvvel Quality Index methodology.
Infrastructure & cost
Cost-per-chapter, caching strategies, batch inference, streaming, and the cost trajectory through 2030.
Hallucination & coherence in long-form fiction
Hallucination as feature or bug, multi-chapter consistency, named-entity drift, world-state representation.
No published pieces in this pillar yet.