They are predicting the pixel layout of the user’s screen, prerendering the passwords, and then transmitting them as images?!

That’s commitment to crisp text rendering!