wow, what kind of stuff does that script do? I've seen non-vision models analyze images, but mostly histograms, color averages etc. This one seems to actually understand the image itself and reproduce the layout, very impressive

That's super impressive given that it doesn't have vision! Intelligence overcomes blindness.

Did the Python script call a vision API?

Either way that's pretty impressive.

Blindsight!