There's already ways to drive a headless UI (Selenium, puppeteer), but they require automation testers to interact with page elements via the "normal" DOM which can be quite painful in terms of creating automation tests - DSLs to drive DOM automation are necessarily huge.
I could envision browser automation via browsh using a very small DSL e.g. browsh.url() to navigate to a URL, browsh.click() to click on an element, browsh.text_exists?() to check whether text exists, browsh.link_exists?() to check whether a clickable link exists