Lee Sun-woo

A screenshot cannot verify a scrolling effect

A gradient on this site looked correctly positioned in a screenshot and disappeared while scrolling in mobile Safari. In the October 2026 implementation, the browser could report the expected position while failing to paint the effect where it needed to appear.

The gap between those two observations matters when reviewing an AI-generated interface. A still image can establish where something appears at rest. A scrolling effect needs to be judged during the gesture that makes it useful.

Define what the fade is supposed to communicate

The fade’s purpose was to indicate content beyond a visible edge. At the top of the page, there is no earlier content to suggest. At the bottom, the last paragraph needs to remain unobscured. A page short enough to fit the viewport needs neither treatment.

Those conditions give the effect a behavior to satisfy. “Add a subtle gradient” would describe its appearance while leaving the important decisions open.

Illustrative scroll states. Fade only the edges that have more content beyond them.

Reproduce the browser’s movement

The top fade had been implemented as moving page content. In the recorded Safari issue, that layer could lag behind scrolling, allowing dark artwork to flash below the status bar as the toolbar changed. Returning the top layer to a fixed position addressed that behavior.

The bottom edge had a different problem. An artificially long movement track eventually caused the fade to stop appearing, even when its calculated position looked correct. Changing the track to follow the page’s actual scroll range addressed the observed painting issue.

The two gradients looked symmetrical in a mockup, but the browser imposed different constraints at each edge. Their implementations had to follow the observed behavior.

A useful reproduction records the page length, scroll direction and toolbar state. Start at the top, scroll until the toolbar collapses, reverse direction and reach the last paragraph. Repeat on a short page. That sequence tests the effect’s purpose as well as the failure.

Resizing a desktop window helps inspect a narrow layout; it does not reproduce the phone browser’s toolbar or rendering behavior. These changes addressed the observed Safari problems, without establishing compatibility with every device or browser version. The next review should repeat the gesture when the surrounding layout or browser changes, rather than rely on another screenshot of the resting page.