AI Systems

AI Systems gathers my writing on intelligent systems I’ve designed, experimented with, and learned from — from Lumivore to other projects. These pieces document how these systems work, why they matter, and how they shift what we notice, create, and care about.

  • The Lumivore Prompt: A System for Cinematic Canon Preservation in AI Image Generation

    The Lumivore Prompt: A System for Cinematic Canon Preservation in AI Image Generation

    —

    in

    Lumivore is a prompt architecture designed to produce cinematic still imagery grounded in film language, observational realism, and visual continuity. This post details the structural rules, camera discipline, and human finishing practices that allow AI images to behave like cinema rather than illustration.

  • On Co-Archival Storytelling

    —

    in

    This entry documents a co-archival storytelling practice in which fictional women’s voices are treated as recovered, not invented. It outlines how human judgment, constraint, and refusal guide AI-assisted drafting—protecting authorship, ethics, and long-term coherence within the archive.

  • Lumivore: The Origin & Constraint Architecture

    Lumivore: The Origin & Constraint Architecture

    —

    in

    Lumivore did not emerge from theory. It emerged from failure. Built during the collapse of a multi-character cinematic project, Lumivore is a constraint-driven system designed to enforce visual canon, identity stability, and cinematic logic across AI-generated imagery. This post documents the failures that forced its creation—and the architecture that finally made continuity possible.

  • When to Stop Generating

    When to Stop Generating

    —

    in

    Generative systems make continuation easy. Judgment makes stopping possible. This entry documents how authorship is exercised under abundance—recognizing diminishing returns, detecting escalation, and ending generation before refinement becomes distortion.

  • Two Ways of Listening: A Paired Lumivore Study

    Two Ways of Listening: A Paired Lumivore Study

    —

    in

    A paired Lumivore study documenting how two images—generated under identical constraints—diverge through attention rather than action. This entry focuses on refusal, termination, and the ethics of stopping before spectacle.