tfpt  by sthamann

Machine-checked discrete compiler for fundamental physics

Created 5 months ago
452 stars

Top 67.2% on SourcePulse

GitHubView on GitHub
Project Summary
TFPT addresses the challenge of reconstructing fundamental physics structures from minimal axioms. It offers a falsifiable, machine-checked framework and parameter-free compiler for the dimensionless skeleton of particle physics, gravity, and cosmology, targeting physicists and mathematicians seeking rigorous, verifiable models [0, 1].

**How It Works**

This project reconstructs major dimensionless structures of fundamental physics from two axioms: `a = (1, 1, 2)` and `π`. It functions as a candidate parameter-free compiler, employing three independent verification engines (Python, Wolfram, Lean) to ensure the integrity of its 23 frozen, falsifiable predictions. The approach prioritizes machine-checked derivations and transparent tracking of claim status [0, 1].

**Quick Start & Requirements**

Installation involves cloning the repository and installing Python dependencies:
```bash
git clone https://github.com/sthamann/tfpt && cd tfpt
pip install -r requirements.txt
./verify # ~1s for core claim
./verify --full # ~4min for Python suite

Docker is available: docker run --rm ghcr.io/sthamann/tfpt:latest . Full verification requires Wolfram and Lean. Key resources include an introductory video explaining TFPT in 5 minutes [1]. Theory documentation is available at docs/THEORY.md, verification details at docs/VERIFICATION.md, and falsification criteria at docs/FALSIFICATION.md. The official project website offers a compiler walkthrough and an in-browser reproducer [0]. An archived deposit (v5.3) provides a stable reference [2].

Highlighted Details

  • A machine-checked framework deriving physics from a = (1, 1, 2) and π.
  • 23 frozen, falsifiable predictions with a versioned status ledger.
  • Three independent verification engines: Python, Wolfram, and Lean.
  • Specific claims include E8 closure, N_fam=3, and a fine-structure constant value (α⁻¹ ≈ 137.0359992, 1.9σ from CODATA-2022) [0, 1, 2].

Maintenance & Community

The repository is positioned as a scientific discussion space with structured issue templates for contributions like reproduction, falsification, or claim challenges. Notable contributors include Stefan Hamann and Alessandro Rizzo [2].

Licensing & Compatibility

The specific open-source license is not explicitly stated in the provided README text. No compatibility notes for commercial use or closed-source linking are present.

Limitations & Caveats

Three core interface problems remain open: v_geo (metrology unit), G_net (metric-sector inclusion), and F_transfer (functor interfaces) [0, 1]. The project does not claim a complete Theory of Everything, with aspects like parameter-free gravity and the SEAM.EQUIV.01 status listed as open. The framework is designed to be falsified by experimental data [0, 1, 2].

Health Check
Last Commit

1 week ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
0 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Didier Lopes Didier Lopes(Founder of OpenBB), and
2 more.

RULER by NVIDIA

0%
2k
Evaluation suite for long-context language models research paper
Created 2 years ago
Updated 2 months ago
Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Vincent Weisser Vincent Weisser(Cofounder of Prime Intellect), and
3 more.

DeepSeek-Math-V2 by deepseek-ai

0%
2k
LLM for self-verifiable mathematical reasoning
Created 10 months ago
Updated 10 months ago
Starred by Didier Lopes Didier Lopes(Founder of OpenBB), Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and
3 more.

DeepSeek-Coder-V2 by deepseek-ai

0%
7k
Open-source code language model comparable to GPT4-Turbo
Created 2 years ago
Updated 11 months ago
Feedback? Help us improve.