🌌 Finite Knowledge Theory (FKT) - Scientific Laboratory

Canonical Scientific Research Laboratory for the Finite Knowledge Theory

A scalable research repository designed to support years of independent, parallel investigation by human scientists and autonomous AI agents.

The long-term objective of this laboratory is to discover whether it is possible to represent every natural language, programming language, and structured document using a finite hierarchy of foundational primitives.

This repository serves as the permanent memory and environment where FKT research is produced, challenged, documented, approved, and preserved.

We maintain five immutable research principles:

1. Research before implementation: Never write production code before understanding the problem. Research always comes first.
2. Every decision must be documented: Every architectural or theoretical decision becomes permanent documentation. No undocumented decisions are allowed.
3. Everything must be challenged: Do not assume existing AI architectures, Transformers, or tokenizers are optimal. Question every assumption.
4. First Principles: Always reason from mathematics and information theory before software engineering. The implementation must emerge from the theory.
5. The repository itself is part of the research: Documentation, architecture, experiments, results, failures, and ideas all belong to the repository.

To support parallel, independent investigations:

• Isolated Workspace: AI models and researchers operate inside dedicated slots (e.g., research/gpt/, research/claude/, research/gemini/, research/deepseek/).
• Scientific Consensus: Validated conclusions are promoted to the shared database (research/shared/) via formal promotion processes linked to Architectural Decision Records (ADRs).
• Trazabilidad Histórica: Failed hypothesis and rejected ideas are preserved under status 🔴 Rejected to serve as valuable learning resources.

finite-knowledge-theory/
├── README.md               # Human-facing introduction and principles
├── AGENTS.md               # Agent entry point and operational constraints
├── GEMINI.md               # Instructions for Gemini agents
├── LICENSE                 # License file
├── prompts/                # Canonical prompts of the repository
├── docs/                   # Guides, changelogs, and agent meta-docs
│   └── RESEARCH_LAB_ARCHITECTURE.md # Master layout and workflows standard
├── research/               # Investigation workspaces (isolated & shared)
├── adr/                    # Architectural Decision Records (ADRs)
├── specification/          # Language-agnostic specifications
├── reference-implementation/ # Canon and optimized implementation source code
├── validation/             # Compliance and validation test vectors
├── benchmarks/             # Portability, throughput, and memory benchmarks
├── experiments/            # Reproducible simulations and math checks
├── examples/               # Practical integration examples
└── scripts/                # Repository maintenance scripts

For a comprehensive guide on directory responsibilities, workflows, and promotion rules, see: 👉 FKT Research Laboratory Architecture Standard (docs/RESEARCH_LAB_ARCHITECTURE.md)

If you are an AI assistant or coding agent working in this repository, you must read and follow the instructions in: 👉 AGENTS.md

Licensed under the MIT License. See LICENSE for details.