Observe, compare, then interpret
Tools read browser APIs and render the values they receive. We do not use a universal “healthy” score. Start from a reproducible action, change one condition at a time and keep a second device or application as a control. Results can depend on the entire input path.
Keyboard and mouse
The keyboard tool separates keydown events marked repeat from other keydowns, and clears held keys when focus is lost. Mouse tests count pointer-down events in the target area. Short left-click intervals are a user-selected filter, not an electrical switch diagnosis. Wheel units remain visible in the log.
Controller resting sample
The five-second sample computes sqrt(x²+y²) for axes 0/1 and 2/3 on each animation frame. Reports include sample count, duration, mean and maximum. These frame samples are not USB packet measurements. Driver calibration and deadzones can already have transformed the values.
Visual display checks
Solid backgrounds are generated in the browser. A person inspects the physical display. Reports explicitly label the assessment as user-entered; there is no automated pixel detection, luminance measurement or repair routine.
Data and reports
The tests use no backend, device-data upload, advertising code or analytics script in this release. Reports are generated locally only when requested. Unsaved observations clear on reload. Choosing Save snapshot explicitly stores up to 20 reports in this browser and origin, until you delete them or the browser clears them. JSON imports are read locally and never uploaded. Hosting still processes ordinary web requests; see Privacy for that distinction.