How we test
What each test observes, and where it stops
Every page starts with a specific job. We compare current tools and user questions, then check the job against browser specifications and real API support. A feature ships only when it produces a useful observation without making a stronger claim than the browser supports.
Search and product consensus
Current US English results are reviewed as two samples: the top three and the full first page. The top three receive explicit priority because they shape immediate user expectations. Only tools doing the same job count toward interface consensus. Articles, forums, videos, stores, and off-intent results remain useful for questions and troubleshooting, but do not vote as tools.
Observable results
The result language says what reached or left the page: an event appeared, a stream produced frames, a sound was scheduled, or a visual pattern was rendered. It does not say the hardware is certified, repaired, calibrated, or healthy.
Testing
Deterministic calculations receive unit tests. Automated browser flows exercise representative start, stop, reset, denied-permission, unsupported, disconnection, cleanup, narrow-screen, focus, and keyboard-navigation cases across the engines that expose each API. Hardware-dependent behavior still needs real devices because a browser mock cannot prove physical compatibility.