Quántir: A Human-LLM Co-Creation Poetic Corpus and Generative Protocol (2024-2025)

This monograph presents the complete output of the "Quántir" project, a research corpus generated through a rigorous Human-in-the-Loop (HITL) adversarial protocol involving Large Language Models (LLMs). The work documents the results of "Architectural Prompting," a methodology designed to test the semantic and structural boundaries of generative AI. Methodology: The Generative Protocol The corpus was constructed following a strict three-phase interaction loop: Input (Architectural Prompting): Imposition of strict formal constraints (meter, rhyme, hybrid syntax) by the human architect. Processing (Generative Engine): Base text generation and rendering of "controlled hallucinations" by the Transformer-based model. Post-Processing (Adversarial Curation): Human editing focused on metric polishing (for classical forms like Sonnets or Silvas) and semantic debugging (for code-based poems). Linguistic and Formal Scope The corpus is notable for its Code-Switching nature, integrating natural languages with formal logic systems. It encompasses: Natural Languages: Spanish (Primary), English, and classical tongues (Latin, Ancient Greek, Sumerian, Sanskrit). Programming Syntaxes: C++, Python, JavaScript, Fortran, Assembly (ASM), SQL, Bash. Formal Systems: Binary, Morse Code, LaTeX, and Chemical Notation. Research Significance Unlike standard literary anthologies, this corpus serves as a primary source for studies in Computational Creativity, AI Alignment, and Digital Humanities. It demonstrates the capacity of AI agents to engage in high-order aesthetic emulation and structural logic, offering a case study on the "black box" of synthetic creativity.