GPT-4o now controls your computer. OpenAI has shipped computer use inside ChatGPT, letting the model operate browsers and desktop applications on macOS and Windows to complete multi-step tasks on your behalf.
The capability works across both an in-app browser and connected external tools like Chrome, with the model able to annotate, navigate, and manipulate interfaces in real time. The video walks through a concrete workflow: analyzing community feedback, moving between browser and desktop apps, and completing tasks inside software you already have installed. That process, not just the endpoint, is what makes this worth watching.
The integration points to a direct challenge to traditional automation tools and RPA software. The next question is where the boundaries are: what it refuses, where it fails, and what permissions it requires. The full demo answers some of that.
[WATCH ON YOUTUBE →]