Alignment is correct, safe, reproducible behavior under explicit constraints

No comments yet.