Hyperstructures
A trained GPT, Midjourney, and Stable Diffusion collaboration generated higher-dimensional geometries with no physical precedent, across nearly 9,000 working files.
Triply periodic minimal surfaces and Calabi–Yau manifolds gave the series its mathematical starting points; a custom-trained AI pipeline pushed those forms toward geometries that have never been built, only imagined.
Role
Concept and AI-pipeline directionGenerative geometry practice
Outcome
A multi-year body of AI-generated higher-dimensional geometry, produced through a custom-trained GPT, Midjourney, and Stable Diffusion pipeline.
The generative practice
Thousands of forms.Only coherent ones survived.
Triply periodic minimal surfaces and Calabi–Yau manifolds gave the project a real mathematical starting point with no direct physical analogue. A trained GPT expanded the language of each form while Midjourney and a separately trained Stable Diffusion model rendered variations. Nearly 9,000 working files accumulated across named series. Most were exploratory; a series existed only when a generated form proved coherent and strange enough to keep.
At a glance
~8,987 working files · GPT, Midjourney, and Stable Diffusion pipeline · 2023–24 named series
In the record
Minimal-surface and Calabi–Yau studies · Layered generative-AI source files · Named-series groupings of resolved forms · GPT language and prompt taxonomy · Midjourney / Stable Diffusion iteration boards · 8,987-file working archive
Start from real math
Triply periodic minimal surfaces and Calabi–Yau manifolds gave every series a defined, if unbuildable, mathematical starting point.
Let the pipeline explore
A trained GPT, Midjourney, and Stable Diffusion worked together to generate forms with no physical precedent to check them against.
Judge what's coherent
Named series and aliases mark the generated forms that held together as geometry, out of far more that didn't.
