We think most of an AI model is wasted.
Modern models get better mainly by adding more numbers. We don't think that's the only way. MatterWave AI is a small research effort testing whether a sharper structure, not a bigger budget, can do the same work with far less.
Inside a normal model are large grids of numbers, and each one is free to become anything during training. That sounds powerful, but it's mostly wasted. The model relearns the same patterns over and over, and nothing stops a grid from quietly making a signal much bigger or smaller as it passes through.
Our design, the Geometric Rotation Transformer, takes that freedom away on purpose. Each grid is replaced by a set of rotations. It uses far fewer parameters to do the same work, and those properties hold before any training even starts.
We design for structure.
Architectures where every parameter earns its place: rotations instead of dense grids, phases instead of loose floats, wiring that's fixed on purpose.
We train and measure.
Real runs, real benchmarks. An architecture idea is only worth what the training curve says it is.
We build in the open.
Weights, configs, and the full walkthrough of the architecture are public — anyone can run it, take it apart, or build on it.
I'm David, the founder of MatterWave. There's no large team behind this yet — just a conviction that AI spends far more energy, memory, and compute than the work actually requires, and that better structure, not a bigger budget, is the honest way to reclaim it.
If that question matters to you too, write to me directly at [email protected].
Read the work and decide for yourself.
The full explanation, the diagrams, the size comparison, and the honest limitations are all in our first research note.
THE GRT NOTE ↘Questions, collaborations, or press — reach us at [email protected].