48 AI Agents Found an Unmapped Maya Pyramid in 72h
Original: Google SWE points 48 AI agents at NASA Lidar, finds unmapped Maya pyramid in 72h
Why This Matters
Demonstrates that commodity AI agents can meaningfully accelerate archaeological survey at scale.
A Google SWE deployed 48 AI agents to scan 1,200 km² of NASA lidar data and flagged an apparent Maya site — leveled plaza, 8–12m pyramid, possible reservoir — in under 72 hours, prompting an official INAH response.
Pietro Miotti, a Google software engineer, was traveling in Peru in October when he started thinking about what modern tools could do for archaeology. Inspired by Hiram Bingham's 1911 'discovery' of Machu Picchu, he asked several AI agents (Gemini, ChatGPT, Claude, Grok) what approach they'd recommend. The unanimous answer: lidar. Peru's coverage was sparse and surveying it would require permits and costs, so he pivoted to the Maya region of Mexico, which has extensive public lidar data. He set 48 agents loose on roughly 1,200 km² of jungle imagery overnight. No specialized classifier was involved — the agents simply examined lidar-rendered images, flagging geometric shapes inconsistent with natural terrain. By morning they had identified multiple candidate sites. After cross-referencing archaeological literature and official archives, one hilltop stood out: unregistered, with straight edges, a leveled plaza, an oval pyramid estimated at 8–12 meters tall, an attached raised courtyard, and what may be a water reservoir. Miotti contacted INAH (Mexico's National Institute of Anthropology and History) and received an official reply within 72 hours. There's a catch, though: a November 11 update reveals the site had already been documented in Estrada-Belli et al. 2025 — the agents missed it in the literature sweep. It was a rediscovery, not a first discovery, and it wasn't yet in INAH's official register. Miotti acknowledges the work isn't done.