About
Research direction with transparent boundaries
Zackary Loevseth develops mathematical and applied research programs using proof-oriented analysis, exact computation, reproducibility controls, and explicit AI-use disclosure.
Research profile
From exploratory work to scoped, reviewable artifacts
The portfolio spans discrete mathematics, finite-group counterexamples, graph-theoretic structure, semiorder geometry, epistemic-system design, computational number theory, geospatial risk communication, and research operations.
The common thread is methodological: define the target, separate proof from computation, make each check replayable, preserve failed routes, and describe every result at the level its evidence supports.
Current professional experience: Customer Support Associate I (Norwegian) | Billing, Shipping & Fulfillment at SpaceX, Sep 2025 - Present. This work includes AI-enabled billing, shipping, fulfillment, teammate training, and repeatable workflow documentation. It is separate from the independent research portfolio; SpaceX is not represented as an author, sponsor, reviewer, or endorser of the research.
Direction, verification, and final accountability
Zackary’s role
Zackary Loevseth directs problem selection, research priorities, claim boundaries, proof and computation review, correction decisions, package standards, and final publication choices. He is the proposed human author of the featured manuscripts, subject to final human confirmation.
AI and tool role
AI systems assisted source discovery, mathematical exploration, proof and code development, checker construction, workflow coordination, manuscript production, and package review. Conventional tools supported exact computation, archive creation, hashing, document rendering, and link checks.
Disclosure boundary: AI output is not mathematical evidence. AI systems are not authors or external reviewers. Final inspection, authorship, legal attestations, licensing, and public release remain human responsibilities.
Status language
How to read this portfolio
Status labels are controls, not decoration. They prevent a complete package from being confused with peer review, and an active program from being confused with a solved target.
- PUBLIC GITHUB PREPRINT RELEASE — EXTERNAL SPECIALIST REVIEW PENDING; NOT JOURNAL/ARXIV SUBMITTED
- A versioned preprint and evidence package is publicly released on GitHub. It is not peer reviewed, accepted, journal published, or arXiv submitted.
- BOUNDED COMPUTATIONAL RESULT — POINTWISE CERTIFICATE F=104
- The released certificate establishes
F(400000287233629)=104and thereforesup_n F(n) >= 104. It makes no priority, unboundedness, maximality, basin, density, inverse-tree-completeness, or global-solution claim. - WORKING TECHNICAL PROGRAM
- A mature, scoped technical architecture exists, with central empirical or theoretical frontiers still open.
- PREREGISTERED STUDY — COMPARATIVE RESULTS NOT YET AVAILABLE
- The evaluation plan exists, but comparative execution has not started and there are no outcome data.
- ACTIVE RESEARCH — TARGET OPEN
- The program is in progress; the named mathematical or empirical target remains unresolved.
- APPLIED RESEARCH — PUBLIC-SAFE SUMMARY
- A carefully limited summary is available, while source rights, validation, or reproducibility work prevents full release.
- METHODS CASE — VERIFIED WORKFLOW SUMMARY
- The operational workflow is documented and supported by package outputs; causal superiority is not claimed.
Technical review
Start with a paper, then follow its evidence
Each featured public release exposes its manuscript, supplement, reproduction guide, claim boundary, verification summary, AI-use disclosure, citation metadata, and license status. Curated first-party replay files are in the dedicated repositories; mixed-rights private archives remain excluded.