How a contemplative moment, a 1929 Swiss discovery, and a relentless two-week sprint produced four published papers, an endorsement from the world's leading oloid authority, and a result that rewrote its own headline mid-experiment.
I was watching a physics interview on a podcast discussing advanced propulsion systems when I encountered a profound gap between what we understand theoretically and what we can actually build. That contemplative headspace intersected with a discovery that changed everything: Paul Schatz.
In 1929, Schatz -- a Swiss engineer and mathematician -- took a cube, inverted it, and discovered the oloid. A shape whose surface contacts a plane in a way that every single point on its surface touches the ground exactly once per rolling cycle. Not approximately. Exactly.
The insight was immediate: geometry as a guarantee. A bearing failing at a contact locus? You can upgrade the material -- that delays the failure. Or you can change the shape of the roller -- that eliminates the failure mode entirely. One is an Architect solution. The other is a Gardener solution. The Substrate Geometry research program was born from that distinction.
The thesis demanded proof, not poetry. I built a computational oracle framework -- six families of metrics that evaluate how any geometry distributes physical load across its surface.
Each geometry receives a complete invariant vector. The oloid's Contact Distribution Score came back at 8.2 x 10-7, compared to a cylinder's 4.75 x 10-5. That's a 58x improvement -- not from better materials, not from tighter tolerances, but from the shape itself.
A 45-variant parametric search across the entire developable roller family confirmed Schatz's intuition computationally: none beat the oloid. The shape that emerged from physical intuition in 1929 is the local optimum in its family.
I reached out to Professor Hellmuth Stachelat the Technical University of Vienna -- the world's leading authority on oloid geometry. Together with Felix Dirnböck, Stachel formally proved the oloid's developable surface properties in 1997. His work is the mathematical foundation that Substrate Geometry builds upon computationally.
Stachel connected me with Georg Nawratil, also at TU Wien, who endorsed Paper I for arXiv submission. That endorsement -- from the academic lineage that proved the oloid's properties -- validated the computational approach as a legitimate extension of the established mathematical framework.
Stachel also has connections to Christian Langscheidt, Paul Schatz's grandson. The lineage from Schatz's 1929 discovery through Stachel's 1997 proof to this computational program is direct and unbroken.
Paper II studied thermal distribution in TPMS (triply periodic minimal surface) electrodes. The original finding was a 2.5x thermal advantage for gyroid geometry over conventional hemispheres. Respectable. Publishable.
Then the adversarial audit caught something. During a convergence study, I discovered that peak temperature had a 41% coefficient of variation across random arc placement seeds at identical mesh resolution. The original multi-arc methodology was conflating geometric properties with stochastic noise.
The pivot to single-arc evaluation revealed the true story: geometry doesn't reduce peak temperature (all geometries hit ~125K at high cooling). It distributes thermal load uniformly. The revised metric -- mean surface temperature -- showed a 25.6x advantage for diamond TPMS over hemisphere. And diamond replaced gyroid as the leader, correlating perfectly with surface area density.
A methodology that was honest enough to audit itself mid-experiment produced a result 10x stronger than the original claim. The headline rewrote itself because the framework demanded it.
The full program timeline: 14 days. April 8 to April 22, 2026. Independent researcher, Toledo, Ohio. No lab. No grant funding. No institution.
Plus a methodology paper characterizing oracle trust boundaries, pending endorsement from Max Wardetzky at the University of Göttingen -- whose 2006 work on discrete differential geometry the curvature convergence findings complement.
Substrate Geometry proposes that many current engineering solutions are material-upgrade responses to problems that geometry could solve. A bearing failing at a contact locus? Change the shape of the roller. An electrode overheating at topological bottlenecks? Add surface area. An IMU losing calibration due to gravitational torque variance? Use a mono-monostatic housing.
The long-term vision is not just that new geometric primitives can be discovered and applied. It is that mapping this field -- populating it with real working applications and new geometry that works with engineering principles -- will reveal a deeper layer: how all of these new applications of substrate geometry interact with each other to produce even more efficient methods of engineering.
The shape is the substrate. Everything else is downstream.