Separating Chromium as a (Javascript-enabled) renderer eventually enables natural-language input from terminal to drive browser interactions. Assertions could be recorded as tests.

From VNC in-browser, potentially could automate any UI with coordinates, clicks, and snapshots. Assuming AI provides visual recognition capability.