A long-horizon program exploring whether frequency-ratio structure can operate as a portable information domain across heterogeneous signals.
Phideus is framed by Harmonic Information Theory (HIT) and studies whether ratio-based relational structure can serve as an information layer that transfers across modalities, not just inside one sensor type. Cross-modality deep learning is the operational experiment field, not the conceptual endpoint.
Phideus is intentionally built as a double path. Both paths are coupled: theory defines what to test, and model behavior provides the evidence loop.
- Research Path: Explore whether ratios carry reusable structure and test that claim across modalities with controlled experiments.
- AI Model Path: Design and train models that learn and operate with ratio logic, moving toward ratio-native representations and decision behavior.
- Project Thesis & Framework: An overview of Harmonic Information Theory (HIT) and what we're aiming to achieve.
- Evidence & Bias Control: Rigorous documentation of tests, empirical results, and validations of our core hypotheses. Includes our stance against fabricated metrics.
- Architectures: Deep dives into the models, mechanisms, and systems being developed to test HIT (e.g., Gate sweeps, Reverse Cross-Attention comparisons).
- Roadmap & Progress: The current state of "Escalon 1", the trajectory to "Escalon 2", and investor-relevant milestones.
- Docs Portal: Access to scientific papers, experiment setups, and data assets.
This project is built with Vite + React.
# Install dependencies
npm install
# Start the development server
npm run dev
# Build for production
npm run buildBuilt with logic, ratios, and rigorous scientific framing.