Why a Single Word Changes Everything
How Large Language Models Actually Read Your Prompt
Research and practical case studies on AI in writing, publishing, and video production. Many come in series drawn from a single Future Fiction Academy session.
A four-part white paper series distilled from the Future Fiction Academy session of the same name — opening the black box of how language models read your prompt, why a single word reorganizes a scene, what names smuggle into your fiction, and why the em dash became the great tell of AI prose.
How Large Language Models Actually Read Your Prompt
Why Negative Prompting Works Again
Demographic and Stylistic Bias in AI-Assisted Fiction
A Case Study in Tokens
A transparent cost and capacity model for an author building a short-drama studio: hardware investment, GPU rental, API generation, retries, and a five-to-ten-episode daily goal.
Eighteen matched before-and-after tests compare an RTX 5070 with an RTX PRO 5000 72GB. A 96-cell follow-up explores H3 and LTX at 480p, 720p, 1080p and 2K, with clear graphs and downloadable results.
The October 25 local-video review snapshot: 43 completed clips, five- and ten-second timings, 480p through 1080p, and the practical limits of 12 GB VRAM and 32 GB RAM.
Six local video tests on an existing RTX 5070: render times, actual clips, INT8 explained, and the compromises of working with 12 GB VRAM and 32 GB RAM.
An agentic system for long-form fiction organized like a Formula 1 race weekend. Fifteen specialized roles protect author judgment at the moments where it actually matters.