Jarvis Arena

Watch a voice assistant try real tasks. A synthetic voice talks to Jarvis, Jarvis drives a real browser, and every run is recorded with a pass or miss for each step and what it cost.

Everything runs on Cloudflare, and the code is open: deploy your own (opens github.com) and run the same workflows.

Runs

The line under each date fills as it ages: empty when new, full at a week.

Workflows