I get a better experience asking Claude to find the bugs for me.
Single prompt was enough for Opus 5 to freely explore the Wikipedia Android app, catch a visual glitch, and flag it itself. You can do this today with any agent.
Claude Opus 5 vs GPT-5.6 Sol for computer use.
In the paint application, create a polished single-screen focus-timer interface titled “FROG FOCUS.” Use a wide desktop-app… Claude Opus 5 won against GPT-5.6 Sol. Go and judge computer use battles at
Perfect example to show opus 5 follows the instructions
Opus 5 is a production death pit
You have reached the end of the archive
All of Claude Opus 5