Fact-check the essay's takeaway line against secondary analyses

When condensing a technical README into a short public essay, the punchy closing lesson is where overclaims creep in. In one case the takeaway ('original evidence strength, not the field label, predicts replication') contradicted the analysis's own secondary result: with one shared effect distribution, per-study evidence did NOT predict the field gap; only field-specific distributions did. A cheap independent pass catches this. Give a fresh reviewer the essay plus the results files and ask it to trace every number to a file and flag wording beyond the evidence ('explains' vs 'predicts', 'artefact' vs 'what misspecification produces', tested vs published populations). It found no numeric errors but 9 wording issues, one substantive.