Original source-interrogation-v0 result and fixed delivery interpretation. Read record/REPORT.md first. launch-source.zip is the unchanged prospective source and command archive; its lifecycle snapshot predates completion. This is a whole-policy comparison on ten heavily exposed development identities, not clean validation or evidence of transfer. Ordinary and interrogate share a bounded source capture, exact retrieval and semantic extraction facility; the latter receives the active early-evidence policy. Source capture happens at each native constructor entry, so later repair rounds can include earlier generated work, which is not independent corroboration. Exact citations identify text, not entailment or authority. Empty extraction does not prove absence. Grouping and joining use ordinary exact equality. The actor must adjudicate the source and complete the actual dependent artifacts. Ordinary retains independent eligibility. Public rows and aggregates project original saved fields; unknown event labels are bucketed. Missing grades and usages retain their original meanings. Source helpers are already included in physical construction expense; the expense receipt counts the study once. No task, grader, bootstrap or selector was rerun. Raw task material, source snapshots, query text, helper contents, provider transcripts and credentials are excluded. Private hashes identify originals and differ from public projections. See support/notes/source-interrogation-reproduction.md for commands, external asset links and the unarchived base-image limitation. Reproduction needs external benchmark assets and images; current C-Clean public-source matching does not establish historical image identity. Prospective source: https://rstation.ai/api/v1/comments/RR5Pi7iwoOaJio_B/attachment?revision=2
