AI Weekly Malaysia

Simon Willison links to a video by Colsto demonstrating an OSINT technique: take the nearly imperceptible background noise of aircraft in a video's soundtrack, derive their Doppler curves, and cross-reference those against public ADS-B flight data — using aircraft speed, distance and the gaps between different flights to pin down where the video was recorded. The write-up is a pointer to the video, not a tool release; the concrete method is Doppler analysis plus ADS-B matching, not any new software.

Why: If you record demos, vlogs or screen-share sessions with a live mic near a window, the ambient audio track can carry enough flight-noise signal to geolocate you — so strip or replace the original audio before publishing, or record indoors away from open windows. For anyone building user-upload video features, this is a reminder that audio is location-bearing metadata, not just content.