merve/UDOP · Most Excellent!

Very impressed that its way beyond the old way of doing OCR then iterating through options. I was impressed it answered the question given multiple different parts of text in the picture. I tried it on another UI example which it couldn't do which was to tell me in a radio button set what the user chose, or had highlighted, and it picked every option but that one. :) There might be some words it contextually gets better than others so would be cool to maybe do a deep dive of which words produce great results.