Independent computational research · Separate track
Formalize the claim. Compute. Try to falsify it.
A separate laboratory for exploratory work in mathematics and computation. The discipline is the same one I bring to engineering systems: define the object, expose the assumptions, measure what can be measured and keep uncertainty visible.
Featured computed object
E8 root system · Coxeter plane projection
One mathematical object viewed as structure, projection, neighborhood and symmetry. Select a root to inspect its local relationship to the rest of the system.
Loading computed geometry…
Computed geometry is unavailable. The E8 model contains 240 roots arranged in eight shells of thirty points.
All 240 roots and the edges between roots at 60°. Eight shells, thirty points each.
240 roots · 8 shells · Coxeter number 30
What this figure establishes. The geometry is computed from all 240 E8 roots and projected onto the Coxeter plane. It makes the research loop concrete: define the structure first, compute the representation, inspect the result and keep the claim status explicit.
CLAIM → OBJECT → COMPUTATION → CHECK → STATUS → ARTIFACT
Projects, checks and published artifacts
Independent research gives me practice in defining objects, exposing assumptions and designing checks that can disprove a claim. Those habits return to engineering software; the mathematical results do not validate the commercial work.
A reproducible framework representing polynomial roots through a companion operator, isospectral transport and numerical resolvent projectors.
github.com/voidzeit/seion-poly-roots →Typed finite-dimensional laws, associator and symmetry defects, structure-preserving projectors and finite cohomological checks.
github.com/voidzeit/seion-math-core →Every image of a fixed size is one integer. The navigator moves through that coordinate exactly in base-256 arithmetic, checked exhaustively on universes small enough to enumerate and differentially against arbitrary-precision integers.
Read the write-up →With Alexa Pérez Cornejo. A constrained late-time dark-energy transition analysis using supernova and baryon acoustic oscillation data.
doi.org/10.5281/zenodo.18091273 →Layer-wise matrix structure checked against functional behavior rather than assumed from compression alone.
Private — not yet publishedRepairing and stress-testing protein structure files before downstream physics-based simulation.
Private — not yet publishedSandbox · separate from the research results
Synthetic illustration — not a paper result
Rank reduction vs. reconstruction error
A small interactive model of the trade-off between discarded structure and reconstruction error.