It’s literally just an LLM that’s given access to your browser through screenshots. If runs them through OCR to read text, then interacts with elements in the DOM using mouse and keyboard.
Maybe moderately okay for automating page unit tests, but opening it up to the general public is a terrible fucking idea. Since they wouldn’t be running it in a development sandbox and by design it will have access to whatever they have access to in their browser.
If they want this to be available it definitely needs to be aggressively sandboxed.
AI… Browser? Do I even wanna know?
It’s literally just an LLM that’s given access to your browser through screenshots. If runs them through OCR to read text, then interacts with elements in the DOM using mouse and keyboard.
Maybe moderately okay for automating page unit tests, but opening it up to the general public is a terrible fucking idea. Since they wouldn’t be running it in a development sandbox and by design it will have access to whatever they have access to in their browser.
If they want this to be available it definitely needs to be aggressively sandboxed.