Open a browser tab and a 3D globe starts turning: how many aircraft are in the air right now, how many ships are under way, where each of 838 tracked orbital objects sits at this moment, where earthquakes struck in the past 24 hours - all layered live onto the same globe. That is God's Eye View, a "spy satellite simulator" in your browser, and the data is real. It welds flights, ships, satellites, earthquakes, public cameras and other open signals onto a single CesiumJS 3D globe, and it can even hook up a voice AI agent over MCP. The project is maintained by former Google product manager Bilawal Sidhu together with Sameh Khamis. Created on June 22, 2026, it measured 53,844 stars and 10,627 forks via the GitHub API on October 13, sat at #2 on that day's Trending board with 1,895 stars gained, and topped both the daily and weekly Trending boards in August (README self-reported; Trendshift independently logged a #1 daily finish on August 27). Repo: bilawalsidhu/gods-eye-view.
Repo: github.com/bilawalsidhu/gods-eye-view
Why It Hit 50,000 Stars in Under Four Months
First, the velocity. The repository was created on June 22, 2026, topped GitHub Trending in August, and measured 53,844 stars on October 13 - all within four months. The most recent push was October 12, the primary language is JavaScript, and development has not paused.
The first reason is the author. Bilawal Sidhu is a former Google product manager who has worked on 3D and geospatial content for years; his companion YouTube series (the predecessor project WorldView) passed 5 million views, with 25 million-plus followers across platforms (all README self-reported figures). In other words, the project launched with a built-in distribution channel in the tens of millions, and even Brendan Eich, the creator of JavaScript, offered a "pretty cool" verdict on social media (README self-reported). But traffic alone does not sustain 50,000 stars. What sustains them is that the product fills a genuine vacuum: satellite imagery, flight tracking and vessel positioning used to live in a dozen separate professional tools, each with its own terminal and its own barrier to entry. God's Eye View packs all of them into one browser tab and wraps the whole thing in a cinematic visual layer.
The second reason is positioning that is honest and catchy at the same time: officially it is a geospatial data fusion client in the GEOINT/OSINT direction; colloquially it is a "spy satellite simulator." The former is the rigorous technical characterization; the latter is natural shareable material - nobody forwards a technical term, but everyone wants to "play intelligence analyst in a browser" once. The technical foundation is a CesiumJS WebGL near-photorealistic 3D globe, and both rendering quality and data density hold up. Stacked together, those make the base case for the Trending run.
The third reason hides in the interaction design: most data-visualization projects grant you the power to look; this one grants the power to operate. Track a target, climb into a cockpit, swap a sensor skin, export a camera link - every step is an action rather than a browse. That sense of agency is why screenshots keep circulating on social media. The star count is the result; turning professional data into a playable experience is the cause.
The Data Layers: 15 Feeds Welded Onto One Globe
This is the hardest part of the project. According to project documentation and third-party coverage, God's Eye View currently fuses 15 data layers:
- Live flights: ADS-B signals from OpenSky and adsb.lol, including military traffic;
- Global vessels: AIS position streams from AISStream;
- Satellites: TLE orbital data from CelesTrak covering 838 tracked objects (including Starlink shells), with positions propagated in real time via SGP4;
- Earthquakes: the past 24 hours from USGS;
- Public cameras: roughly 3,600 publicly accessible cameras you can open;
- Radios: 750 geolocated stations;
- Plus traffic, wildfires (NASA FIRMS), launch missions (Launch Library 2), and weather (GFS/ECMWF wind fields, radar, clouds, lightning and cyclones).
All of these figures are project-documentation/third-party framing; this article did not re-verify each layer. What matters is not just the count but the word "fusion." Individually, flight data has FlightRadar24, vessels have MarineTraffic, earthquakes have the USGS site - each is an existing service. Stack them onto one globe, however, and cross-references start to appear: you can watch flight paths bend around a typhoon's wind field, check a camera near a wildfire for smoke, or line up a launch trajectory against the traffic in nearby airspace. A single source answers "what happened"; a fused view answers "how do these things relate." That is the difference between a spatial intelligence desk and a pile of dashboards.
One number matters most for onboarding cost: 13 of the 15 layers need no API key (third-party framing). You can clone the project, register nothing, and see flights, ships, satellites and earthquakes immediately - the base layers all work keyless. Keys enter in three places: the near-photorealistic 3D globe asks for a Cesium ion key (mind its personal non-commercial terms) or a Google Maps key; the voice AI agent needs an OpenAI key. In other words, a key or two stands between "free sightseeing" and "the full experience" - a pragmatic trade between open access and sustainability.
The Core Experience: Not Just Watching, But Climbing In
If it only scattered data points across a globe, it would be a pretty dashboard. What sets it apart is a set of "climb inside" interactions.
The cockpit view is the feature that went viral: you can climb into any tracked aircraft and fly its real descent profile toward the runway from a pilot's-eye view. Combined with the 250 km contact roster and click-to-track, once you select a target the camera locks on and stays with it. The 3D hangar holds real aircraft models - a Boeing 787, an ATR-72, a Citation business jet, a Bell 206 helicopter, an MQ-9 drone - so tracked targets render as their actual airframe instead of a generic icon.
The visual layer plays even more "military." GLSL sensor skins switch the entire scene into CRT tube, NVG night vision, FLIR thermal, Noir film or Snow static textures, layered with a military HUD telemetry overlay and detection boxes - the spy-film atmosphere cranked to maximum. A scene director provides cinematic camera moves, and share links serialize the camera state, style, layer toggles and tracked targets into the URL itself: the "camera script" you tuned travels to anyone as a single link.
Two more layers are built for the AI era. A voice whiteboard lets you dictate annotations and draw real-boundary polygons directly on the globe. And an MCP server plugs the whole globe into Claude or Codex desktop clients - an agent can query layers, watch targets and place annotations for you, turning the globe into a spatial intelligence terminal for your AI.
Three Scenarios: Beyond Sightseeing
Scenario one: aviation and maritime enthusiasts. Tracking a flight used to mean one site, a vessel another; now both sit on one globe. Pick an aircraft on approach, jump into the cockpit, ride its real descent profile, then zoom into a harbor to see AIS ship density. The 250 km contact roster lists everything nearby as a clickable manifest.
Scenario two: a playground for agent developers. With the MCP server attached to a Claude or Codex desktop client, you can hand spatial-intelligence chores to an agent: assign "watch this airport's traffic and flag unusual airframes," and the agent queries layers and reasons over them through MCP. For anyone learning MCP integration, this is a hands-on target with live data and visual feedback - far more intuitive than wiring up a database.
Scenario three: content creators. GLSL sensor skins plus the scene director amount to a free cinematic footage pipeline: flip to FLIR thermal, dial in a camera move, serialize it into a share link, and a "military documentary" shot is ready. Given the author's own YouTube DNA, this path is officially validated - the README demo videos are made exactly this way (README self-reported framing).
Getting It Running: Four Commands
Installation is standard (README framing); Node 18 or newer is enough, and four commands get it running locally:
git clone https://github.com/bilawalsidhu/gods-eye-view.git
cd gods-eye-view
npm ci
npm run doctor
npm run devnpm run doctor is the step most worth running first: it checks your environment and the connectivity of each data source, telling you in one pass which layers work, which do not, and which keys are missing. If you would rather not touch a terminal, an official Pinokio one-click install wraps cloning, dependencies and launch for you. Repeat the key matrix once more: base layers work with no key; the near-photorealistic 3D globe needs Cesium ion (personal non-commercial terms) or a Google Maps key; the voice AI agent needs an OpenAI key. Run the core layers keyless first, confirm your machine handles the rendering load, then decide whether to add keys for visual quality - the reverse order is how people get stuck on VRAM and billing on day one.
Honest Boundaries: What It Is Not
First, it is not a live satellite feed. It is a geospatial data fusion client that renders open data sources onto a 3D globe in real time. What you see is not a stream of satellite imagery but a fusion visualization of multiple open signals - the satellite layer shows orbital positions propagated from TLE data, not pictures taken by satellite cameras. The official framing is clear about this; do not let the "spy satellite" nickname mislead you. It simulates the experience of an intelligence desk, not the lens of a satellite.
Second, some layers carry delay or are estimates by nature: public-camera poses and rocket launch trajectories are rough estimates, and the traffic layer is simulated flow on real road networks rather than measured congestion (project-documentation framing). Fine as a live log; check each layer's reliability before citing anything as intelligence.
Third, two known traps. One: older versions hitting the public OSM Overpass endpoint now get rejected - if the Traffic layer is blank, update first (third-party reporting, October 12 framing). Two: VRAM and bills. Cesium entity streams stacked with Google Photorealistic 3D Tiles are heavy on GPU memory; on weaker hardware, turn photorealistic 3D off. The cloud APIs it connects to (ion, Google Maps, OpenAI) can generate real charges once usage outgrows free tiers - stay inside the free allowances to be safe.
Fourth, licensing. The LICENSE file, read directly, is MIT (Copyright (c) 2026 Bilawal Sidhu), so commercial use is allowed. One detail worth noting: the GitHub API's license field reads NOASSERTION, but the LICENSE text itself is standard MIT - go by the original text. Compared with some "open source but restricted commercially" projects covered here before, this repo carries no extra commercial shackles.
Fifth, a word on privacy optics. Roughly 3,600 public cameras sounds alarming, but they all come from open channels - this project aggregates and visualizes; it does not intrude or bypass permissions. That said, gathering scattered public cameras onto one globe makes "these places can all be seen" suddenly tangible: for ordinary users it is a privacy-awareness wake-up call; for security practitioners it is exposed-surface survey material. The tool is neutral; usage is on you - a line that applies here more than most.
Related reading
Division of labor: we covered gods-eye-view in September (at 32,399 stars, focused on the capability surface and privacy boundaries); this piece is the October 13 update view - 53,844 stars, No. 2 on the daily trending board, with a hands-on installation and voice-agent/MCP focus. The two complement each other.
- September deep dive: capability map and privacy boundaries
- WorldMonitor: another real-data intelligence globe
- World models and spatial intelligence compared
FAQ
Q1: What exactly is God's Eye View? A1: An open-source 3D spatial intelligence desk that runs in the browser, fusing 15 categories of open data - flights, ships, satellites, earthquakes, public cameras and more - onto a CesiumJS 3D globe, with voice AI agent and MCP support. GitHub API measured 53,844 stars on October 13, 2026, MIT licensed.
Q2: Can I use it without registering any API key? A2: Yes. 13 of 15 layers need no key (third-party framing), and all base layers work keyless. The near-photorealistic 3D globe needs a Cesium ion or Google Maps key; the voice AI agent needs an OpenAI key.
Q3: Does it show real live satellite imagery? A3: No. It is a data fusion client rendering live visualizations of open signals, not a satellite imagery stream. Some layers carry delay or are estimates - judge each layer's reliability before citing it.
Q4: Are the hardware requirements high, and will it cost money?
A4: Node 18+ is enough to install, and npm run doctor checks your setup. Core layers are free, but photorealistic 3D is VRAM-hungry, and cloud APIs (Cesium ion, Google Maps, OpenAI) can produce real bills beyond their free tiers.
Q5: Can it be used commercially? A5: Yes. The LICENSE text is MIT as read directly from the file; the GitHub API field says NOASSERTION, but the original text governs, with no extra commercial restrictions. Note that Cesium ion carries its own personal non-commercial terms, which apply separately when you use that service.
Sources
- GitHub repository (53,844 stars / 10,627 forks, MIT, JavaScript, measured via API on 2026-10-13): https://github.com/bilawalsidhu/gods-eye-view
- Official README and install docs (install commands, Pinokio, key matrix, Trending self-report): https://github.com/bilawalsidhu/gods-eye-view
- LICENSE text (MIT, Copyright (c) 2026 Bilawal Sidhu): see the repository LICENSE file
- Layer details and keyless count (per project documentation / third-party coverage by pgsr.cn, not re-verified layer by layer): see repository documentation
- Overpass endpoint rejection and update notice (third-party report by aipmclub, 2026-10-12); VRAM and cloud API billing reminders (third-party report by hoangyell)