Independent Research.
No GPU. No Institution.

The research and the motivation

Geometric Context Modeling was developed by Tiffney Bare as an independent research project. Twenty-seven experiments were conducted on a consumer CPU, with no GPU, no institutional affiliation, and no training budget. Five papers have been published on Zenodo. Two more are in active development.

The central question that motivated the research was not how to make AI faster, but whether the transformer approach is the right foundation at all. The GPU hardware barrier does not sort researchers by aptitude. It sorts them by economic circumstance. A researcher with a five-year-old laptop and no cloud budget cannot meaningfully experiment with current AI systems. This research was built from the ground up to change that constraint.

If geometric AI matures to the point where semantic understanding, mathematical reasoning, and agent intelligence run on any CPU, the people and communities currently excluded from shaping AI will finally have the tools to participate. The decisions being made now about what AI can do and who it serves should not be made exclusively by those who can afford the hardware.

The research program

The language track established the geometric architecture across 27 experiments and five papers. The mathematical track is active, building a parallel axis space for mathematical reasoning. The geometry investigation is testing hyperbolic and directed geometry for containment hierarchies and causal relationships. Experiments 28 and 29 will add episodic memory and dual state tracking for both sides of a conversation.

Every experiment result, every paper, and the full codebase will be released as open source. The goal is independent replication of every result by any researcher with a standard laptop, no GPU, no cloud account, and no institutional access required.