Why can't we solve the problem of direct eye contact in video calls?

The right solution is a camera either under the screen or hanging in front of the screen. The former is costly and not widely available, the latter is a non-standard configuration that most people don't want to use (because it blocks part of the screen).

Tools that rewrite video to change the appearance of the eyes are disturbing, and also not trivially integrated into the video calls people are already making.

Embed video cam under screen, direct eye gaze to a useful portion of the screen with the image people look at: their own.

I've seen ai assisted eye gaze fakery and it's pretty average. Steps into uncanny valley fast.