Request one look
The app takes one photo when requested. If start-on-launch is enabled, opening navigation requests the first photo automatically.
Experimental assistive technology
WayFinder takes one photo when asked, describes visible objects, and speaks a cautious response. It is designed around audio, large controls, and fail-closed guidance.
A deliberate three-step loop
The app takes one photo when requested. If start-on-launch is enabled, opening navigation requests the first photo automatically.
The server asks a hosted vision model for structured scene facts, then applies deterministic safety rules.
Hazards and STOP responses take priority. Old, raced, uncertain, or failed results are not presented as clear paths.
Current status
WayFinder is not a validated mobility aid and is not available for unsupervised use. Real-device accessibility and supervised safety evaluation remain required before a public release.