The Research

AI companies forgot
to use your brain.

You've absorbed ~300 billion tokens of narrative by age 18. The most powerful reasoning systems ever built are communicating with you through instruction manuals. Two papers on why that's insane — and what the correction looks like.

Read the papers Get early access
"You already know everything you need.
Nobody taught you how to give it to AI."
The one sentence that produced two papers, one product, and this research program.
Paper One
The Human Layer: Narrative Oversight Protocol for Human-AI Accountability
FAccT Submission · 2026 · Tyson Ray Cueva Caves

A narrative oversight protocol for improving human oversight of complex AI systems. Core claim: narrative can help not only by explaining a system but by uncomplicating it enough for a human to use correctly. The uniform activates the role. Story saves time. Archetypes function as behavioral contracts.

FAccTNarrative CognitionHuman OversightAI AccountabilityExplainability
Paper Two — In Progress
The Ambient Guardian: How Continuous Signal Creates the Conditions for Human Genius
2026 · Tyson Ray Cueva Caves · Built on Paper One

Paper One solved how humans understand AI. Paper Two solves the inverse: how AI understands the human well enough to serve them. Proposes the Donna Standard — a design philosophy for AI support systems built on continuous ambient signal, local-first data ownership, and narrative-based context compression.

Ambient SignalFlow StateHuman-AI CollaborationLocal-FirstEthics
Six original contributions across both papers
1
The Pre-Trained Human: Childhood narrative exposure (~300B tokens) is analogous to LLM pre-training. AI interface design has been ignoring this entirely.
2
Narrative as highest-bandwidth AI interface: Activating pre-existing human schema transfers more context per word than any instruction-based approach.
3
The Donna asymmetry: AI has no competing life. That is not a limitation — it is the one capability that makes ambient guardianship possible at the level human genius requires.
4
Local-first as ethical prerequisite: The same ambient signal architecture enables a guardian or a weapon. Data ownership is the only determinant.
5
The activation philosophy: The most valuable resource is always already present, unseen. Intelligence is activation, not creation.
6
Archetype as AI context-loading mechanism: The costume/uniform applied specifically to human-AI interface design. The knighting ceremony before every session.
The evidence spans ten thousand years
The barbarian proof
Story changed behavior where law failed. The oldest known human behavior-change technology.
The Viking conversion
Christianity civilized through narrative, not rule. Civilizational-scale behavior change via story.
The grandma saying
Seven words that activate ten thousand tokens of pre-existing understanding. Zero tokens to run. Forever.
Elf on the shelf
Parents figured out priming, trusted intermediary feedback, and ambient signal before AI was invented.
Agent BTZ
The adversarial Donna. Same ambient collection architecture. Different master. The proof by negative example.
The knighting ceremony
Warrior, armor, chef's prep, flower watered. Every domain that produces excellence has a preparation ritual.
Ready to experience the thesis?
Rafiki is the paper as a product. He sees who you are before you speak, sets the scene, and hands you the armor that fits. Your grandma saying at the end.
Ask for access