Mathematical research · exact computation · evidence design
Researcher in mathematics, exact computation, and verifiable human–AI workflows.
Permanent public portfolio URL
https://zackaryloevseth.github.io/research-portfolio/
Research profile
My work combines structural mathematics, exact computation, and adversarial checking to turn promising ideas into bounded, reproducible claims. I keep finite evidence, conditional results, open questions, and publication status visibly separate so each artifact says exactly what has—and has not—been established.
Featured publications
Each live public preprint release includes a manuscript, supplementary appendix, reproduction guide, scope statement, verification summary, and AI-use disclosure. Curated first-party replay files are in the dedicated repositories; mixed-rights private archives remain excluded.
PUBLIC GITHUB PREPRINT RELEASE — EXTERNAL SPECIALIST REVIEW PENDING; NOT JOURNAL/ARXIV SUBMITTED
Constructs exact infinite families refuting five specific finite-group inequalities, with explicit finite witnesses and a database-scoped computational audit.
The dedicated public repository and preprint release are live. The curated first-party replay package grants no reuse license.
PUBLIC GITHUB PREPRINT RELEASE — EXTERNAL SPECIALIST REVIEW PENDING; NOT JOURNAL/ARXIV SUBMITTED
For represented semiorders, partitions nonempty ideals by largest selected index and derives translated-zonotope bounds for planar projections of their ideal hulls, supported by exact finite verification.
The dedicated public repository and preprint release are live. The curated first-party replay package grants no reuse license.
PUBLIC GITHUB PREPRINT RELEASE — EXTERNAL SPECIALIST REVIEW PENDING; NOT JOURNAL/ARXIV SUBMITTED
Proves a lower bound on excess degree for a hypothetical lexicographically minimal counterexample; the global conjecture remains open.
The dedicated public repository and preprint release are live. Curated first-party replay files are public while the mixed-rights private archive remains excluded.
Research programs
Open programs are presented as open. Public artifacts appear only where a stable, commit-pinned source is available.
BOUNDED COMPUTATIONAL RESULT — POINTWISE CERTIFICATE F=104
The released certificate establishes F(400000287233629)=104, with the consequence sup_n F(n) >= 104. It does not claim world-record priority, unboundedness, global maximality, basin infinitude, density, inverse-tree completeness, or a solution of the global problem.
WORKING TECHNICAL PROGRAM
A formal architecture for premise-governed reasoning under revision. Exact finite additive results are stated only with their assumptions; the closure-constrained continuous-cost compression frontier and comparative superiority remain open. The 37-page v2 document is a private working manuscript, not a peer-reviewed publication.
PREREGISTERED STUDY — COMPARATIVE RESULTS NOT YET AVAILABLE
The public-safe parity scaffold passed 40/40 checks; the broader authored-gold/adversarial development suite passed 208/208 checks. Neither count is efficacy evidence. Confirmatory execution has not started: zero model runs and no pilot or held-out outcome data.
UNSOLVED / IN PROGRESS — CURRENT ROUTES STOPPED
RIEMANN HYPOTHESIS STATUS: UNSOLVED / IN PROGRESS. Version 2 was rejected at N06, and version 3 stopped at a precise frozen dependency-schema obstruction. No accepted EvalW certificate, negative Weil entry, off-line zero, or global positivity theorem exists.
ACTIVE RESEARCH — TARGET OPEN
Exploratory theorem-building programs remain in progress. Finite witnesses and internal constructions are not presented as unrestricted theorems.
Methods and verification
Claims move through a reproducible workflow: scope definition, proof reconstruction, exact computation where appropriate, independent replay, hostile audit, correction or supersession, and human-final release controls.
Finite, conditional, internal, and process evidence are labeled before public claims are drafted.
Packages pair exact checks with manifests, hashes, source instructions, and clean reproduction steps.
AI assists research and production, but AI output is not evidence, authorship, or external review.
Reproducibility methods Current Scientific Director status brief Current Scientific Director methods brief (PDF)
Applied research
APPLIED RESEARCH — PUBLIC-SAFE SUMMARY
A public-safe applied case summary examines how dynamic hazard layers could support operational risk communication. The current evidence is descriptive; it does not establish predictive validity or causal effectiveness.
Role and AI-use disclosure
Zackary Loevseth directs problem selection, claim boundaries, verification standards, and publication decisions. AI systems assisted source discovery, proof and code development, checking, workflow coordination, and manuscript production; they are not authors, external reviewers, or substitutes for mathematical evidence. Legal, authorship, affiliation, and submission attestations remain human-only actions.
Read the full role and status statement AI-use and research-role disclosure
Contact
Email zloevseth@gmail.com or use the verified GitHub profile below.