If you prompt Gemini about the characteristics of a "Where's Waldo" image it will give a good answer, so it "knows" but (in this case) didn't demonstrate understanding. An average human would do a worse job with the art but would never make this kind of error

Anyway I think this is an interesting test case: a type of image where the most important detail is something that is, by necessity, visually insignificant

I asked it to try again but with the figure of Waldo less prominent and more difficult to find. Its second attempt also put Waldo front and center. I tried a 3rd time, specifying a particular location for Waldo. He was added to the background crowd, but his twin was still present in the foreground.

It was not even especially sharp about gender

Anyway I guess that experience is scalable now

RBC Bank has a cross-border account that might be useful for you if you need to move money between countries a lot.

Not bad