I built a self-hosted 3D ADS-B viewer to see how obstacles affect receiver coverage. It builds a 3D coverage dome from 30-day history.
My receiver has a structure blocking about 30° of the sky, which reduces its range by about 100 km in that direction.
The dome is just a mesh to hide a receiver's exact coordinates. On my simple test, GPT-5.6 Sol couldn't narrow down the location more than about 2.5 km in 9 different methods.
Also, each aircraft rolls and pitches using real ADS-B data. The protocol does not transmit pitch angle, so it is estimated from the vertical rate.
My tiny single receiver is in South Korea, so the local traffic could be very small when people from the other side of the Earth visits. There is an ADSB.lol network feed on the demo for that case.
It is mostly AI (Codex and Claude) generated.
Source: [https://github.com/luftaquila/skytrace](https://github.com/luftaquila/skytrace)
Comments URL: [https://news.ycombinator.com/item?id=49109712](https://news.ycombinator.com/item?id=49109712)
Points: 1