It looks like a spy thriller cockpit: a dark HUD, a photorealistic globe, aircraft tracked in real time, a FLIR-style heat filter you can flip on. It started as a weekend project and grew into one of the biggest open-source releases of the month, and the data behind it is public, not classified.
God's Eye View is a browser app that fuses open-source intelligence feeds onto a 3D Earth. The code was released under the MIT license by Bilawal Sidhu, and it hit #1 on GitHub Trending this past week.12 As of this writing the repository sits at 16.4k stars, with about 10.7k of those arriving in the last week.1 Sidhu spent six years at Google, four of them on Maps, before this project.3
The project started life in late February as WorldView, which Sidhu described as a "geospatial command center" that fuses open-source intelligence feeds onto a 3D globe. It was a weekend project, built by himself.3 The demo went viral online. Days later, when the Iran strikes happened and airspace shut down across the Gulf, Sidhu reconstructed the sequence of events, from satellite passes to widening no-fly zones, using nothing but public feeds and AI coding agents.3 The "God's Eye View" name belongs to the episode that rebuilt the strikes from public data.3
The open-source release followed in August after Sidhu polled his audience about making it public.4
Thirteen live layers ship in the app, and eleven of them need no account or API key at all.1 The keyless stack includes:
- Live flights from OpenSky Network and adsb.lol, including military traffic
- An 838-object satellite catalog from CelesTrak, with an option to drop in the whole Starlink shell
- Earthquakes from USGS, covering the last 24 hours
- About 800 public traffic cameras projected into the 3D scene, from Austin, California (Caltrans), and London (TfL)
- Geolocated radio stations via Radio Browser, with an analog-style tuner
- Bikeshare station availability from GBFS feeds
- A rolling 30-day list of space launches from Launch Library 2
- Bundled static datasets: 4,351 datacenters, 704 dams, and 712 submarine cables1
The optional paid tier is modest. A free Cesium ion token enables Google Photorealistic 3D cities, and a metered Google Maps key is another route to the same 3D plus place search. A free AISStream key adds live ship positions, NASA FIRMS adds active fires, and TomTom swaps simulated traffic for real congestion.1 OpenAI is the one cost that can grow: voice control and the AI HUD summary run over the Realtime API, with an in-app session-spend readout, a $2 warning, and a $5 hard cap per session that ends the session.1
The machine that makes it feel real: satellite positions run through SGP4 propagation, aircraft and ships track their true heading on screen at any camera angle, and the live feeds arrive every 15 to 30 seconds, with the globe rendering a beat behind and interpolating between fixes.1 The stack is deliberately plain: no framework, vanilla JavaScript, CesiumJS, and Vite.1
The current version of the voice agent exposes 28 tools across four jobs, and Sidhu has shared example prompts in the README.1 Queries like "how many flights are over Texas right now?" or "outline the state of Texas" work because the agent pulls live scene context, including coordinates, street names, active layers, and view scale, before answering.1
The honesty is part of the pitch. The README tells you which layers are simulated or estimated: traffic is a simulation until you add a TomTom key, camera poses are priors you calibrate by dragging a gizmo, rocket launch replays are labeled "RECONSTRUCTED ESTIMATE," and the military installations layer is described as "incomplete by nature."1
The MIT license covers only the code. Several bundled datasets keep their own licenses, and the README is unusually careful about this: the TeleGeography submarine cable data is CC BY-NC-SA 3.0, which means it is NonCommercial, and commercial users are told to delete that one folder or license it from TeleGeography.51
There is also a deliberate line about what the project will not do. It models events, assets, infrastructure, and systems, and it does not build features for named-person search, face recognition, or tracking individuals; the README says pull requests that cross that line will not be merged.1 The page's end-user warning is blunt: the data may be delayed, incomplete, modeled, inferred, or wrong, and it should not be used for navigation, emergency response, or other safety-critical purposes.1 The project is maintained by Sidhu and Sameh Khamis under the Halfpixel label.1
What is next is a hosted version. After the launch, the loudest request was simply "just give me a link," so the team is building an official hosted God's Eye View at Halfpixel, with no installation needed.1
- https://github.com/bilawalsidhu/gods-eye-view
- https://github.com/trending?since=weekly
- https://www.spatialintelligence.ai/p/the-intelligence-monopoly-is-over
- https://www.spatialintelligence.ai/p/i-open-sourced-gods-eye-view
- https://raw.githubusercontent.com/bilawalsidhu/gods-eye-view/main/DATA_SOURCES.md