A former Google Maps PM built a spy satellite that runs in a browser tab. When a real war broke out, he pointed it at the war.

In partnership with

On the last weekend of February 2026, the United States and Israel opened a bombing campaign against Iran that the Pentagon would name Operation Epic Fury. Airspace closed across nine countries. GPS interference spread across the Gulf. Shipping backed up at the Strait of Hormuz.

And a former Google Maps product manager watched the whole thing happen, minute by minute, from a browser tab — using nothing but data anyone on Earth can get for free.

His name is Bilawal Sidhu. The tab was running his own project, a photorealistic 3D globe called God’s Eye View. He’d built it for fun. Two months earlier he open-sourced the whole thing on GitHub. As of this week it has 18,200 stars, 3,700 forks, an MIT licence, and a two-line description that undersells it badly:

That is not marketing copy. It is a literal description of what loads when you open it.

🛰️ What’s actually running in the tab

God’s Eye View is a single web app — vanilla JavaScript, CesiumJS, Google’s Photorealistic 3D Tiles — that fuses together a stack of public feeds most people don’t know exist, let alone that they’re free:

Click a plane and you can ride along in cockpit view as it lands. Flip on FLIR or night-vision rendering. Ask it questions out loud — it’s wired to the OpenAI Realtime API — and it’ll fly the camera there and annotate what you’re looking at. Sidhu says a cold start takes about 1.86 seconds. Brendan Eich, the man who invented JavaScript, publicly endorsed it. It hit #1 on GitHub Trending the week it shipped.

Every one of those feeds is individually boring. A city’s traffic-camera API. A hobbyist ADS-B receiver network. A satellite catalogue anyone can query. None of that is a secret. What’s new is that one weekend of work turned “a dozen public APIs nobody combines” into a single live globe you can talk to.

🌪️ What he did with it when a war actually started

This is the part that made it more than a neat toy.

When the strikes on Iran began, Sidhu didn’t just watch the news. He pointed an AI agent swarm — his phrase — at every open-source signal he could find before it disappeared, and used it to build a full 4D replay of the first 24 hours: airspace closures locking in over Tehran, GPS interference spreading across the region, no-fly zones going up over nine countries simultaneously, and the shipping fleet backing up at Hormuz. You can scrub through it minute by minute.

In a follow-up post he flipped on the orbital layer and noticed something nobody had pointed out yet: commercial satellite passes stacking up over the strike zones in the hours before and after impact. Other people’s eyes, in orbit, also watching — visible only because he’d bothered to plot it. By his own account, a single-person geospatial command centre that would have taken a full team a financial quarter to build at Google, he shipped over a weekend, fusing six data layers and tracking 3,400-plus aircraft through the opening hours of a real war.

That is a genuinely new capability. It used to require a security clearance. Now it requires a laptop and a free weekend.

🎯 The question the creator can’t answer either

To his credit, Sidhu doesn’t pretend this is uncomplicated. In a later essay he asked the question directly:

He points to journalists spotting naval fleet movements in TikTok videos filmed by residents near a base. Commercial satellite imagery anyone can buy with a credit card. And then, honestly:

That’s a rare thing for a creator to publish about their own project, and it’s worth sitting with before you play with the globe yourself.

One quick word from today’s sponsor

AI made PMs faster. Multiplayer mode is still broken.

A PM can summarize research, draft a PRD, and mock up a prototype before lunch. The hard part starts when the team has to decide what actually gets built.

Jira Product Discovery gives product teams one place to capture insights, prioritize ideas with consistent frameworks, and build living roadmaps stakeholders can rally around.

And because it’s connected to Jira, the context behind every decision stays with the work—so developers and their agents know not just what to build, but why.

AI helps PMs move faster. Jira Product Discovery helps the whole team build with confidence.

Get Started for Free.

⚠️ Four ways this goes wrong

None of which is a reason God’s Eye View shouldn’t exist. It’s open, it’s honest about its limits, and its own README carries a plain warning: “Data may be delayed, incomplete, modeled, inferred, or wrong. Do not use it for flight or maritime navigation, emergency response, medical decisions, investment decisions, or other safety-critical purposes.” The tool isn’t the problem. What it proves about aggregation is.

1. The mosaic effect is real, and it just got a UI. Intelligence analysts have a name for this: individually unclassified facts, combined, can add up to something genuinely sensitive — the “mosaic effect.” A traffic camera feed, a hobbyist’s ADS-B antenna, a satellite catalogue: none of it is secret. Stack a dozen of them on one photorealistic globe with a voice interface and you’ve built something that used to sit behind a clearance. The barrier was never really the data. It was the two months of glue code, and now the glue code is MIT-licensed.

2. Military aircraft tracking is a known, government-acknowledged risk — and this tool ships a dedicated feed for it. God’s Eye View pulls from adsb.lol, which explicitly aggregates military ADS-B broadcasts. This isn’t hypothetical: a GAO report from January 2018 found that DOD and FAA had known since 2008 that ADS-B Out — which the Pentagon was required to install fleet-wide by 2020 — could expose military aircraft movements to anyone with a cheap receiver, and that the two agencies had “not approved any solutions to address these risks” years after identifying them. A 2022 piece in the Naval Institute’s Proceedings carried the blunter title: “Twitter Trackers Jeopardize Military Aircraft.”

3. The same tracking works on people, not just planes. In December 2022, a college student named Jack Sweeney ran a bot called @ElonJet that did nothing but post Elon Musk’s private jet’s public ADS-B position. Musk banned the account from Twitter and threatened to sue. The underlying data was, and is, entirely public — the same category of feed God’s Eye View plots by default. Track a war, track a CEO, track an ex: the pipeline doesn’t care which.

4. The agent swarm is the more durable story than the globe. The globe is a UI. The part worth remembering is how the Epic Fury replay got made — an unattended swarm of AI agents scraping every open signal it could reach before the caches cleared, with nobody manually curating in real time. That’s a general-purpose capability. It’s exactly as good at building a live picture of a company’s shipping schedule, a public figure’s movements, or an unlisted-but-not-secret facility as it is at replaying a war.

🛡️ What to actually do about it

Short list, because the fixes here are policy and habit, not code.

1. If you fly privately, or run an org that does: know the FAA’s two privacy programs exist. The PIA program issues a temporary, alternate aircraft ID not tied to your registry entry — free in its current phase, live within 10 business days. The LADD program (formerly the Blocked Aircraft Registration Request) hides an aircraft from the FAA’s own public data feed. They solve different halves of the problem; most owners who care about exposure need both.

2. If you run physical security for an organisation, audit your “boring public” footprint on purpose. Traffic cameras, hobbyist ADS-B and AIS receivers near your site, satellite revisit schedules over your facility — none of it is a secret today, and a tool like this is a preview of how cheaply it gets fused tomorrow.

3. Don’t assume “it’s already public” means “nobody will bother to combine it.” That was true right up until one developer proved otherwise in a weekend. The cost of aggregation, not the sensitivity of any single feed, is now the only thing standing between “scattered” and “searchable.”

4. If you’re building with agent swarms yourself, think about what you’re pointing them at before you ask what they can find. The exact autonomous-collection pattern that made the Epic Fury replay impressive is indifferent to its target.

5. Play with the tool before you worry about it in the abstract. Seeing your own general neighbourhood of the world reconstructed from public feeds is the fastest way to understand the mosaic effect that a paragraph like this one can offer.

🧵 The takeaway

A government used to need a classified budget to build a live picture of a warzone from scattered signals. A hobbyist built a rough version of the same thing, in the open, over a weekend, using APIs anyone can query for free — and then pointed it at an actual war within hours of the first strike.

The globe is genuinely wonderful to use. Ride along in an aircraft’s cockpit view sometime; it’s the closest most of us will get to seeing the planet the way an analyst does. But the interesting story here was never really the 3D rendering. It’s that the thing standing between “a pile of public data” and “an intelligence product” has quietly gotten this cheap — and the person who proved it is the first to say he doesn’t know what to do about that either.

Reply and tell me what you’d want to see plotted on a globe like this — I’ll try to find the public feed for it.

📚 Sources — every figure above, checked at the primary source

Not printed, because I could not confirm it at a primary source: the specific AI models used inside Sidhu’s “agent swarm,” and any casualty or damage figures from Operation Epic Fury — this piece is about the open-source tooling and the OSINT-aggregation risk it illustrates, not a military assessment of the operation itself.