Fractals make non-closure visible because under magnification, the non-attainment of a terminal witness is more than mere absence. Smooth objects close under zoom by yielding a tangent element. Exactly self-similar objects close by return. The target fractal case does neither: the instance under magnification does not settle to one form or one finite orbit. What may stabilize instead is a law of instancing: a scenery distribution, invariant measure, or statistical pattern one level up. The asymmetric commuting square is the minimal grammar for this situation. It certifies lawful reading across loss, but it does not say which distinctions a reading has retained; the one-point reading commutes with everything while retaining no distinction. Fractal non-closure therefore names three tasks: identify the closure not attained under magnification, identify the law-level closure that may replace it, and state the domain’s collapse condition so that the resulting law remains informative for that domain.