FuaFlow

Deterministic algebraic semantics for neural networks. From mathematical foundation to end-to-end language models to training analysis tools.

Read the Docs GitHub

3 Projects

3 PyPI Packages

3 Research Papers

2 HuggingFace Models

Research Pipeline

From Theory to Practice

Foundation

Engine

LSH-to-prime projection for algebraic verification

→

Application

TriadicGPT

40M GPT learns primes end-to-end, zero cost

→

Analysis

Reptimeline

Track how representations evolve during training

Tools & Research

Projects

Foundation

Triadic Neurosymbolic Engine

28.4x faster than cosine

Deterministic algebraic framework that projects neural embeddings into composite prime integers. Exact subsumption, composition, gap analysis, and analogy operations.

PyPI: triadic-engine Paper BUSL-1.1

pip install triadic-engine

Docs GitHub PyPI

Application

TriadicGPT

98% analogy accuracy

40M-parameter GPT that learns prime-factor signatures end-to-end alongside language modeling. Drop-in triadic head for any HuggingFace transformer at zero language cost.

PyPI: triadic-head Paper HuggingFace BUSL-1.1

pip install triadic-head

Docs GitHub PyPI

Analysis

Reptimeline

3 phase transitions detected

Track how discrete representations evolve during training. Lifecycle events, phase transitions, bottom-up ontology discovery, and 3-way AND-gate interactions.

PyPI: reptimeline Paper Software DOI BUSL-1.1

pip install reptimeline

Docs GitHub PyPI

J. Arturo Ornelas Brand

Independent Researcher

GitHub PyPI Engine Paper TriadicGPT Paper Reptimeline Paper HuggingFace