An ex google dev vibe coded a Palantir-like surveillance clone in 2 hours

WHY THIS MATTERS IN BRIEF

If everyone can create software then everyone can repeat this experiment, and do much more …

In a demonstration of how rapidly Generative Artificial Intelligence (AI) is collapsing the barrier between hobbyist coding and enterprise-grade software development, a former Google Maps product manager has built a sophisticated geospatial intelligence dashboard in just three days.

Bilawal Sidhu, utilizing a technique he describes as “vibe coding,” created a browser-based visualization tool that mirrors the capabilities of high-end intelligence platforms like those produced by Palantir Technologies. The project merges Google Earth’s 3D tiles with a suite of real-time global data streams, effectively turning a standard web browser into a command-and-control center.

See it in action

As you can see from the video the dashboard provides a high-fidelity, interactive map of the globe augmented by several layers of live data. The system tracks over 6,700 commercial flights via Open Sky and utilizes crowdsourced ADSB data to monitor military aircraft – vessels that are often filtered out of consumer-facing flight trackers. Real-time orbital paths of satellites, including specific IDs and classifications such as geostationary orbits, are then neatly overlaid on the 3D map alongside live global seismic activity.

In a feat of data synthesis Sidhu integrated live, public CCTV feeds from cities like Austin, Texas, projecting these frames directly onto 3D map geometry. For urban environments lacking full 3D scans, such as parts of Dubai, he utilized OpenStreetMap road network data to generate particle systems that convincingly emulate street traffic. To lean into the “spy-thriller” aesthetic, the interface features toggleable visual modes including night vision, thermal (FLIR), and a retro CRT filter.

The project’s speed highlights a shift in software engineering. Rather than manually writing thousands of lines of code or utilizing traditional visual effects software like After Effects, Sidhu acted as a conductor for an “army” of AI agents. By commanding multiple models -including Google’s Gemini and Anthropic’s Claude – through terminal interfaces, he delegated specific technical hurdles to different agents, such as handling complex shaders for visual effects or managing the integration of disparate data APIs.

However, Sidhu noted that the process still requires significant human oversight. The developer must engage in “creative problem-solving,” such as instructing the AI to sequence data loading – prioritizing main arterial roads before smaller streets – to prevent browser crashes.

While his project proves that aggregated, public data can be synthesized into powerful tools, it also underscores the widening gap between public information and private “walled gardens.” Much of the data used in the dashboard – satellite orbits, flight paths, and seismic events – is accessible for free. However, individual-level behavioral data, such as that held by Meta, remains a proprietary frontier. While such data is currently leveraged for hyper-targeted advertising, the emergence of sophisticated, DIY visualization tools raises questions about the future of data democratization versus data privacy.

In the world of high-stakes technology and finance, the most valuable assets remain the ones not for sale. Just as a “Sharpe 5+” trading strategy – a metric indicating exceptionally high risk-adjusted returns – would likely be guarded as a proprietary secret rather than sold as a retail product, the most potent data remains behind closed doors. For now though Sidhu has shown that with enough “vibe” and a few AI agents, the public can get closer to the “eye in the sky” than ever before.