> As a photographer, I care about robustness of the visual output. Which means, as a designer, designing for the worst possible image and taking numerical metrics with a grain of salt.

I think it's kind of silly how the author pooh-poohs averages and demands that whoever is working compression algorithms should focus on the worst possible image. If you know anything about information theory, you know that is literally mathematically impossible to make a compression algorithm that always performs well in the worst possible case.