Independent Reproduction
The strongest evidence for a Kronos result is a stranger reproducing it, so we build every deposit to be reproduced without our help.
The result we want most
The purpose of publishing data, code, licenses, and runbooks is a single outcome: someone with no connection to Kronos downloads a deposit and recovers our result on their own machine. Independent reproduction is the evidence that a claim is about the physics, not about our tooling or our team. It is the test the whole open-science strategy is built to invite.
What we do to make it possible
- Permissive licenses, so running our code and reusing our content is legal; see code under Apache.
- Pinned environments, so the code runs the same elsewhere; see environment pinning.
- Cold-reproduction runbooks, so a reader is walked from download to reproduced figure; see the cold-reproduction runbook.
- A verify step, so 'reproduced' has a defined meaning; see the verify function.
Reproduction includes finding us wrong
An honest invitation to reproduce is also an invitation to falsify. If an outside reader reruns a sweep and cannot recover a published design point, that is a finding we want, not one we resist. The pre-registered predictions and the honest gates exist precisely so that being wrong is detectable from the outside.
Why not just trust the internal check
An internal reproduction shares our blind spots. An external one does not. Only outside reproduction tests whether the result survives contact with a different environment, a different reader, and a skeptic's incentives. That is the difference between checking our work and proving it.
We consider a result fully published only once someone outside Kronos could reproduce it — and we would like them to try.