God's Eye View: a public-data globe you run locally

What it is, and why it topped GitHub in a week

God's Eye View is a photorealistic 3D globe that runs in your browser. It markets itself as a spy-satellite simulator, but every data source is public and the data is mostly real: live aircraft, ships, satellites, earthquakes, traffic, and public cameras, all layered onto one rotatable planet. It was previously called WorldView.

The surge is not hype. In the OpenGithubs weekly GitHub ranking for 2026-09-13 it held number 1, with 29,396 total stars and a +11,455 single-week gain. A direct GitHub API check on 2026-09-14 showed 32,399 stars, 6,480 forks, and 199 open issues. It also reached number 1 on GitHub Trending daily and weekly in August 2026.

It grew out of the viral God's Eye View YouTube series (5M+ views, 25M+ social reach), hit number 8 on Product Hunt Product of the Day vetted by Chris Messina, was called "pretty cool" by JavaScript creator Brendan Eich, and was picked up by Pinokio. Popularity is not value, though; the sections below cover what actually matters.

The real thesis: public data, local-first, MIT

Its real value is pulling scattered public signals into one inspectable, extensible interface.

Flight transponders come from OpenSky and adsb.lol and show 11,000+ live aircraft with route history. Ship beacons come from AISStream for thousands of vessels. Orbital elements come from CelesTrak, 838 tracked objects color-coded by class. Earthquakes come from USGS over the last 24 hours. About 3,400 public cameras are projected into 3D across Austin, London, California, Finland, and more. These signals were already public but fragmented; God's Eye View aligns them onto one coordinate system and clock, so you zoom from a world view down to one aircraft, one ship, or one street.

The bigger point is local-first: on both paths the server binds only to localhost and Provider Settings answers only local requests, so your keys, queries, and view stay on your hardware. Add commercial-friendly MIT, confirmed by reading the LICENSE file (Copyright (c) 2026 Bilawal Sidhu): you can use it commercially, modify it, and redistribute closed source, with every line auditable. That separates it from closed-source SaaS and non-commercial projects, and from a demo video. Treat it as a modifiable open-source foundation; to add your own source, write one layer module.

What it is not: a hybrid of real and simulated, not surveillance

First, it is a hybrid of aggregation and simulation, not all real-time truth. The README is honest: most feeds are live or scheduled, but traffic is simulated along real roads from aggregate location data, and CCTV poses and rocket trajectories are coarse estimates. The cars are a simulation on real road networks; camera angles are prior estimates you calibrate by dragging, not precise measurements.

Second, it is an exploration tool, not surveillance. It models events, assets, infrastructure, and systems, aircraft, ships, satellites, fires, cameras, cities, and builds no named-person face recognition or tracking. Pull requests that cross that line are not merged; a person is not a query type.

Third, public data is not freely usable data, the privacy and compliance boundary. Cesium ion free is personal non-commercial; the direct Google Maps route is metered under quota and budget limits; public cameras come from city APIs but their poses are estimated. A derivative product must check each source's terms. The browser-side Google Maps and Cesium ion keys must be restricted at their providers, and the localhost-only bind exists to shrink that exposure. It cannot replace authoritative sources and should not be used for navigation, emergency, medical, or investment decisions.

Out of the box: what works with no keys

It is built to run with zero keys. Esri satellite imagery plus keyless terrain is the basemap, with OSM as automatic fallback. Flights, military traffic, satellites, earthquakes, public cameras, radio, and launches all work without keys.

The table contrasts no-key out-of-the-box capability with keyed upgrades (green needs no key, yellow needs a free key, red is metered):

Vessels and fires need keys but those are free, so cost is still zero. The only metered, paid items are the direct Google Maps route and OpenAI voice.

Optional upgrades: which keys to bring

Keys go in the POWER UP panel, bottom-right; they are upgrades, not prerequisites. Of six main keys, four are free and two are metered:

• Cesium ion (yellow, free): Google Photorealistic 3D cities and terrain, personal non-commercial, token at cesium.com/ion.
• Google Maps (red, metered): direct 3D plus place search via Map Tiles API, apply in Google Cloud Console with a URL restriction.
• OpenAI (red, metered): voice and AI HUD summary via Realtime API; mini is enough, standard is smarter; audio is a few cents per minute, a heavy evening is single-digit dollars, with live spend, a 2 dollar warning, and a 5 dollar hard cap.
• AISStream (yellow, free): global live ships, free signup.
• NASA FIRMS (yellow, free): live fires, free.
• TomTom (yellow, free): real flow speeds and congestion for traffic, free tier.

Two extras: OpenSky (more flight polling, works anonymous) and Launch Library 2 (higher mission allowance, works without a key). Keys land in git-ignored pinokio/ENVIRONMENT on Pinokio or repo-root .env on terminal, written owner-only before the secret and excluded from Git; env or Keychain keys show as configured externally and read-only.

Installing it: two paths

Path 1, Pinokio: version 8.2 or later (8.0.40 has an install defect and logs submitted values in its Configure panel, so do not enter credentials there). Open the project, click Install then Start; it supports Windows, macOS, Linux, installs dependencies, finds a free port, and opens the app.

Path 2, terminal: Node.js 24.x (24.14.0 or later) or 26.x; Node 25 is end-of-life and the doctor warns.

git clone https://github.com/bilawalsidhu/gods-eye-view.git
cd gods-eye-view
npm ci
npm run doctor
npm run dev

Open http://localhost:4173; the first panel offers Live Contacts, Space Missions, Environmental, or manual explore. On macOS, ./scripts/dev-fresh.sh clears the Vite cache, pulls Keychain keys, and can start keyless. The README records a median 1.86-second cold start from one M5/Chrome capture, explicitly a comparison baseline, not a guarantee.

Data sources and the honesty caveat

The globe also bundles static infrastructure: 4,351 datacenters, 704 dams, 712 submarine cables, each with provenance. Features: cockpit view keeps terrain below as you ride a plane down; Contacts lists everything within 250 km and drops into any cockpit; click-to-track locks the camera, draws a trail, shows metadata, and hands a fire or vessel to the nearest live camera; voice whiteboard speaks real boundary polygons, marks, and routes onto the world; the 3D hangar models 787, ATR-72, Citation, Bell 206, MQ-9, swapping glyph to model as you close in; GLSL sensor looks give CRT, NVG, FLIR, Noir, Snow; plus detection boxes, a military HUD, one-switch global picture, scene director tours, share links, and reset.

The honesty caveat repeats: most sources are live or periodic, traffic is simulated on real roads, CCTV poses and rocket paths are coarse. Launch replays are labeled RECONSTRUCTED ESTIMATE, a reconstructed estimate, not a precise path.

Under the hood and performance

No framework: vanilla JavaScript with CesiumJS and Vite, plus Google 3D Tiles and the OpenAI Realtime API, so the code is easy to read and hack. Honest details: aircraft and ships project along true heading and never spin; feeds arrive every 15 to 30 seconds and the globe renders one interval behind with interpolation and dead reckoning; satellites use SGP4 with GMST-locked orbit rings that do not drift; entity heights use a real geoid-aware vertical datum sampled against rendered terrain, so aircraft park and cameras stand instead of floating. Caching and request budgets (OpenSky governor, TomTom daily budget, disk-cached TLEs) cut repeated requests but do not replace provider quotas. Security: every private-key call, OpenAI, AISStream, OpenSky OAuth, camera frames, goes through a hardened server-side proxy with SSRF protection, response caps, and sanitized errors; the browser sees only the Google Maps and Cesium ion keys, which you must restrict. SECURITY.md covers LAN rules: localhost-only by default, explicit opt-in to go on LAN, and Provider Settings disabled once shared so remote users cannot reach the key panel.

License: how permissive MIT really is

MIT is about as permissive as it gets: commercial use, modification, and closed-source redistribution, with the single requirement of keeping the notice. For an internal or customer-facing product, the compliance cost is minimal.

Read this against sibling coverage: another hot-take piece this week uses CC BY-NC-SA 4.0, non-commercial and share-alike. You may remix it but not for commercial purposes, and derivatives must use the same license. CC BY-NC-SA 4.0 closes both the commercial and closed-source doors; MIT opens both. For a technical reader this decides what you can build: MIT lets you tinker and also drop it as a foundation into a commercial system, provided you handle each data source's terms, because the underlying data is not necessarily MIT. Code freedom is not data freedom.

Who it is for, and related reading

It fits three groups: geospatial and visualization engineers using it as a reference for aligning a dozen live feeds onto one globe; security and OSINT enthusiasts who want public signals in one interface for situational-awareness teaching; and product people scoping the experience floor of spatial-intelligence tools. It is not for those wanting production-grade, high-availability service with someone behind it, nor for anyone tracking specific individuals, a line the project will not cross.

Related pieces in the same batch are worth reading together: the hot-take explainer at /zh/posts/lingbot-world-2-0-hotspot, the multi-tool comparison at /zh/posts/open-world-model-comparison-review, and the local deployment SOP at /zh/posts/lingbot-world-small-local-deploy-sop. Side by side they show the full shape of the public-data, local-first, commercial-friendly line of work.

FAQ

Q1: Does it really work with no keys?

A1: Yes. Esri satellite imagery plus keyless terrain is the basemap, and flights, military traffic, satellites, earthquakes, public cameras, radio, and launches all work without keys. Only photorealistic 3D, place search, and voice cost money or need a free key.

Q2: How is this different from a surveillance tool?

A1: It models events, assets, infrastructure, and systems, aircraft, ships, satellites, fires, cameras, cities. It builds no named-person face recognition or tracking, and such pull requests are not merged. People are not a query type, and public data still carries each source's terms.

Q3: Are the keys safe?

A1: The server binds to localhost only and Provider Settings answers only local requests; any private-key call goes through a hardened server-side proxy. The browser sees only the Google Maps and Cesium ion keys, which you must restrict at their providers. Keys are written owner-only and excluded from Git.

Q4: What does the MIT license mean?

A1: You can use it commercially, modify it, and redistribute it closed-source, keeping the copyright and permission notice. That contrasts with the CC BY-NC-SA 4.0 license in some sibling coverage, which forbids commercial use and closed derivatives.

Q5: Are the performance requirements high?

A1: Not really. The README records a median 1.86-second cold start from one M5/Chrome capture, explicitly a comparison baseline. You need Node.js 24.14.0 or later, or 26.x; Node 25 is end-of-life. With no keys it opens at localhost:4173.