We ran one of our hardest computer-use benchmarks on Anthropic Sonnet 4.5, side-by-side with Sonnet 4. Ask: "Install LibreOffice and make a sales table". Sonnet 4.5: 214 turns, clean trajectory Sonnet 4: 316 turns, major detours The dif…

App-Use lets you scope agents to just the apps they need. Instead of full desktop access, say "only work with Safari and Notes" or "just control iPhone Mirroring" – visual isolation without new processes for perfectly focused automa…

We are bringing Computer Use to the web, you can now control cloud desktops from JavaScript right in the browser. Until today computer use was Python only shutting out web devs. Now you can automate real UIs without servers, VMs, or any weird wor…

Same tasks, same grounding model we just swapped GPT 4o with GPT 5 as the thinking model. Left = 4o, right = 5. Watch GPT 5 pull through. Grounding model: Salesforce GTA1-7B Action space: CUA Cloud Instances (macOS/Linux/Windows) The task is: &…

Photoshop using c/ua. No code. Just a user prompt, picking models and a Docker, and the right agent loop. A glimpse at the more managed experience c/ua is building to lower the barrier for casual vibe-coders. Github : https://github.com/trycua/cu…