[docheinestages]

This is a helpful method for visually grounding LLMs to take actions on the screen such as clicking. For humans though, hell no.