Google SWE used 48 AI agents to find an ancient Mayan pyramid

A Google SWE team used a system of 48 AI agents to analyze remote sensing data provided by NASA. As a result, the multi-agent complex discovered a previously unknown and unmapped Mayan pyramid in Mexico in just 72 hours.

The study utilized massive lidar datasets (NASA Lidar) covering hard-to-reach forest areas. The agents automatically distributed tasks for noise filtering, identifying geometric anomalies in the terrain, and verifying findings.

Researchers have officially notified the relevant Mexican government agencies about the discovered archaeological site for further field study and inclusion in official registries.

Such cases demonstrate the practical applicability of autonomous agent swarms for large-scale geospatial data analysis, reducing manual processing time from months to mere days.