Hey, I helped build this, so treat this as project and judge it accordingly.
You post a real browser task. Two frontier computer-use agents run it at the same time from the same starting state, Claude Opus 5, the GPT-5.6 family, Gemini 3.6, thirteen models on the board. You watch both replays and vote for the winner without being told which is which. Free, no usage cap.
Top comments (0)