This explanation appears to go entirely against how OpenAI says Operator works. It takes a screenshot of the browser and performs clicks and inputs based on that. There is no processing of the underlying "semantic" HTML or CSS.
If anything, Operator demonstrates the opposite of what this article claims -- which is that semantic HTML/CSS has no bearing on how humans or machines perceive the page.
This explanation appears to go entirely against how OpenAI says Operator works. It takes a screenshot of the browser and performs clicks and inputs based on that. There is no processing of the underlying "semantic" HTML or CSS.
If anything, Operator demonstrates the opposite of what this article claims -- which is that semantic HTML/CSS has no bearing on how humans or machines perceive the page.