@علي ملص your point about layered review landed with me. One person collecting, another one or two checking, and AI catching the mechanical mistakes already sounds more trustworthy than a lot of solo Quran datasets that never got a second look. I would still want a short list of tasks that stay specialist-only, things like variant readings or rulings-adjacent labels, so the open contribution path has clear edges. Has the كود يخدم القرآن campaign written that boundary down yet?
