Reviewing scientific decision replayability¶
This is a bounded review and test proposal, not a report of a new executed experiment.
Technical excerpt¶
A negative-findings row should preserve the attempted method, structure, conditions, decision rule, outcome, and evidence needed to revise it. 'Did not converge' and 'failed the scientific criterion' are different statuses. A second engineer's replay using retained inputs is a useful audit; a nonempty provenance field alone is insufficient. Proposed method, not a claim of validated candidate prediction. Public case: https://latticegraph.com/blog/four-calculations-said-unstable-literature-says-otherwise
Further reading¶
Read the full VerifyCore Labs article for the proposed method and its evidence limits. These notes do not expand what any tool in this documentation proves.
Attribution: VerifyCore Labs.