How to Independently Reproduce the Record
Anyone can fetch the breeder deposit, verify the files, rebuild the environment, and rerun the 81 analyses; this page is the entry point to doing so.
Reproduction in five steps
Independent reproduction is the point of the whole record. A third party does not need Kronos to check the breeder (Hyperion) numbers; the deposit contains what is required. The path is deliberately short.
- Fetch the deposit from Zenodo (DOI 10.5281/zenodo.21795620 for the current version).
- Verify every file against the published checksums.
- Rebuild the pinned computational environment.
- Rerun the analyses of interest from their frozen inputs.
- Compare outputs to the published expected values within tolerance.
Two acceptable outcomes
A reproduction that matches confirms the arithmetic. A reproduction that fails to match is equally valuable: it points at either a mistake in the record or a difference in environment, both of which the record wants surfaced. Discrepancies should be reported so they can be entered in the corrections process.
Scope of reproduction
Reproduction checks that the computation is faithful — that the published numbers follow from the inputs and code. It does not, and cannot, validate against a machine that has not been built. Confirming the model against hardware is a separate step that begins at FOAK. See verification vs validation.
Reproduction is also the cheapest external audit a program can offer. It requires no site visit, no privileged access, and no cooperation from Kronos beyond the deposit already published. A skeptic with a laptop and the DOI can check the arithmetic that underlies the headline claims, which is a deliberately low bar to clear on purpose: the harder the record is to check, the less its openness is worth.
This page describes a design-and-simulation study, not a built machine. Construction of the breeder (Hyperion) begins Q2 2027; first-of-a-kind first tritium is targeted near 2030. No hardware net-gain claim is made before FOAK.