A geometry-first JAX engine for Light Theory: Quantum Geometric Tensor, Kaluza-Klein uplift, prime-gauge Wilson loops, and Reeb-flow dynamics. Like TensorFlow Quantum, but for information geometry.
Project description
Light Theory Realm
A geometry-first engine that measures the shape of information in particles and AI models using Clifford algebra and the Quantum Geometric Tensor (QGT).
- 🧮 Core engine: JAX-based Cl(1,3) Clifford algebra + Quantum Geometric Tensor
- 🧬 Standard Model toy: 9 fermion masses with ~2.3% average error from prime plaquettes
- 📐 Geometry everywhere: Fisher information, Berry curvature, Kaluza–Klein uplift, Reeb flow
📚 Start Here
1. Choose your depth
-
For non-physicists / engineers
👉 Foundations of Light Theory
Plain-language explanation of "geometry of information", curvature, and prime barcodes. -
For physicists / theorists
👉 Light Mechanics
Full mathematical treatment: Cl(1,3), QGT, information field equations, KK uplift, Reeb flow, geometrodynamics. -
For decision makers / hiring managers
👉 EXECUTIVE_SUMMARY.md
One-pager with results, metrics, and impact.
2. Or just run it
pip install light-theory-realm
# Standard Model mass table (Pocket_U Lite)
light-realm sm-table
# Koide relation check for leptons
light-realm koide 0.511 105.66 1776.86
More first steps: GETTING_STARTED.md
⚙️ What this library actually does
At a high level:
-
Algebra layer – CliffordEngine Implements Cl(1,3) (gamma matrices, grades, wedge product) in JAX.
-
Geometry layer – CliffordQGT Given a state and its Jacobian, computes the full Quantum Geometric Tensor:
- Fisher information (metric)
- Berry curvature (geometric twist)
-
Theory layer – geometry as dynamics Builds higher structures on top of the QGT:
- Information field equations (Einstein–Fisher-like)
- 5D Kaluza–Klein uplift of the information metric
- Contact geometry + Reeb flow (time and "dark-energy-like" terms)
-
Experiments – Pocket_U Lite & friends Uses prime plaquettes as discrete "labels" for excitations and shows that:
- 9 Standard Model fermion masses can be reproduced at ~2–3% error
- Each particle has a full geometric fingerprint (metric, curvature, KK, Reeb)
For a detailed architecture diagram and module-level docs, see: 👉 ARCHITECTURE.md and API_REFERENCE.md
🔍 Minimal code examples
Compute Fisher + Berry for your own model
import jax
import jax.numpy as jnp
from light_theory_realm import CliffordEngine, CliffordQGT
engine = CliffordEngine(seed=42)
qgt = CliffordQGT(engine)
# Example: a tiny 1-parameter family of states
def get_psi(theta):
# user-defined: returns a normalized 4-component spinor
# ...
return psi
theta0 = 0.1
eps = 1e-4
psi = get_psi(theta0)
dpsi = (get_psi(theta0 + eps) - get_psi(theta0 - eps)) / (2 * eps)
jac = dpsi.reshape(-1, 1)
fisher, berry = qgt.compute_full_qgt(psi, jac)
print("Fisher metric:", fisher)
print("Berry curvature:", berry)
Inspect a particle's geometric fingerprint
from light_theory_realm import get_particle_profile
e = get_particle_profile("e") # electron
print(f"Mass (MeV): {e['m_phys_MeV']:.3f}")
print(f"Error vs PDG: {e['error_pct']:.2f}%")
print(f"Dark energy ξ: {e['xi']:.4f}")
print(f"Fisher trace: {e['fisher_trace']:.6f}")
print(f"Berry norm: {e['berry_norm']:.6f}")
More examples in examples/.
🧭 Who is this for?
-
ML / AI researchers Use the QGT engine to probe the geometry of models: sharp directions, curvature, phase transitions, etc.
-
Theoretical physicists Treat this as a concrete playground for geometrodynamics and holographic information geometry.
-
Independent / self-taught researchers Everything is open, tested, and documented; you can run and modify any experiment locally.
🧪 Tests & quality
- 7/7 unit tests passing across algebra + geometry layers
- JAX-native, differentiable, GPU-ready
- Pocket_U Lite reproduces 9 fermion masses with ~2.3% average error
See TESTING_GUIDE.md for how to run and extend tests.
🤝 Contributing
Contributions, questions, and critiques are welcome.
- How to get started: GETTING_STARTED.md
- How to contribute: CONTRIBUTING.md
- Open implementation tasks: TODO_IMPLEMENTATION.md
📜 Citation
If you use Light Theory Realm in research:
@software{jeannoel2025lighttheory,
author = {Jean-Noel II, Dimitry},
title = {Light Theory Realm: A Geometry-First Framework for Information-Geometric Physics},
year = {2025},
organization = {Pleroma Works, LLC},
url = {https://github.com/Pleroma-Works/Light_Theory_Realm},
orcid = {0009-0009-6082-8647},
note = {Version 0.1.0}
}
Author contact:
Dimitry Jean-Noel II
Pleroma Works, LLC
djean@botwpbc.com
djeannoelii@gmail.com
ORCID: 0009-0009-6082-8647
License
This project is licensed under the Apache License 2.0 – see the LICENSE file for details.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file light_theory_realm-0.1.0.tar.gz.
File metadata
- Download URL: light_theory_realm-0.1.0.tar.gz
- Upload date:
- Size: 1.9 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
b1e61e1fa3737bed87adea31d95f5197f897722f8554061f17a10f6006c01284
|
|
| MD5 |
8649a440488e549dacd2bae7d8d369d6
|
|
| BLAKE2b-256 |
a9a1a7f70143953e1dd23ad226cbbbd44346134ecb10a179ca1f1d8fd109a1cd
|
File details
Details for the file light_theory_realm-0.1.0-py3-none-any.whl.
File metadata
- Download URL: light_theory_realm-0.1.0-py3-none-any.whl
- Upload date:
- Size: 395.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.13.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
e165a6077c3acfcef21ea548442f18afe11490af8523cb49a1f46292ecc482d2
|
|
| MD5 |
d5de6425072769933ac71b2cc7d26b45
|
|
| BLAKE2b-256 |
4a4587d9d42210003b2ca085774d5a075c3e49a303e4bac0c16598ab223f2d84
|