A Google Engineer’s SI Found a Lost Maya Pyramid in 72 Hours. Experts Got There First.

Hiram Bingham needed a local farmer to lead him up a mountain before the world learned about Machu Picchu. Pietro Miotti needed a laptop, a free NASA dataset and 48 SI helpers that worked through the night while he slept.

Miotti is a software engineer at Google. He is not an archaeologist, and he says so up front. But in less than 72 hours last month, his swarm of SI agents went from a stray idea he had on vacation to a pyramid hidden under the jungle in southern Mexico, a written report in Spanish, and a reply from Mexico’s national archaeology agency saying it had no site on record at that spot.

Then came the twist. It turns out archaeologists had already found it.

His post took off. As of Wednesday it had passed 1.2 million views and more than 20,000 likes, and the story has spread across tech sites. The honest version, which Miotti himself added on Wednesday, is more interesting than the hype.

A vacation thought and a night shift

According to the diary Miotti published, it started on Oct. 18 while he was traveling in Peru, reading up on Bingham. Bingham had actually been hunting for a different lost city when he stumbled onto Machu Picchu in 1911. Miotti wondered what an explorer like that would do with today’s tools.

He put the question to several chatbots, naming Gemini, ChatGPT, Claude and Grok. All of them gave the same answer: lidar. From a satellite, an old ruin in the jungle is just a sea of treetops. Lidar fires laser pulses from an aircraft that slip through gaps in the leaves and bounce off the ground, so you can strip the forest away and see the shape of the land underneath.

Peru didn’t have much public lidar, so he switched to the Maya region, which NASA had scanned years ago for forest research. That night he told the agents to analyze it and went to bed.

When he woke up on Oct. 19, the agents had turned roughly 1,200 square kilometers of jungle scans into images and gone through them one at a time, the way a very patient grad student would. Their job was to spot pyramids, platforms and plazas while tossing out natural hills, sinkholes and roads. Miotti stresses there was no special archaeology software involved. The agents simply looked at the pictures and followed a written guide.

They flagged plenty of villages. He asked them to check which ones hadn’t been documented. A few hours later, one agent came back with a candidate: a hilltop on the edge of a cliff with straight edges and sharp corners. As Miotti puts it, nature doesn’t really do straight lines.

What the laser saw under the trees

Zoomed in, the scan shows a leveled plaza, an oval pyramid at least 8 to 12 meters tall, a raised courtyard attached to it and a flat-bottomed hollow that may have held rainwater. The agents also went back to the raw laser points to make sure the shapes were on the ground and not a trick of the canopy, which runs 10 to 17 meters thick there.

Miotti’s best guess, which he labels speculation, is a small civic center from the Classic Maya period, roughly A.D. 250 to 900. Think of it as the downtown of a local ruling family: a plaza for gatherings, a temple on the pyramid, maybe the ruler’s house on the courtyard and a few homes on the platforms around it.

Next, he asked the agents what to do. They told him to contact Mexico’s National Institute of Anthropology and History, known as INAH, explained the reporting process and wrote the report. Because he doesn’t speak Spanish, he had them translate it, keeping the exact coordinates in a confidential attachment so looters couldn’t use them. He sent it on Oct. 20. On Oct. 21, INAH wrote back that it found no registered archaeological site at that location.

Archaeologists were already there

Not being in the government’s register is not the same as being unknown. Archaeologist Sarah Parcak pointed out on Bluesky that the site had already been published by Francisco Estrada-Belli of Tulane, as The Verge reported.

Miotti owned it. In an update on Wednesday he wrote that the site appears in a 2025 study by Estrada-Belli and colleagues, which the agents missed while combing the literature. “They rediscovered it rather than discovered it,” he wrote. That study pulled together public and privately commissioned lidar across a 95,000-square-kilometer stretch of the Maya lowlands and estimated the area was once home to 9.5 million to 16 million people, according to the paper’s summary.

So the agents did not beat the professionals. They matched them. A non-expert with off-the-shelf SI tools and free government data landed on the same hilltop that trained archaeologists had already mapped, in about three days. Miotti calls that a good sign, and he has a point.

It also shows the weak spot. The agents were great at looking at pictures for hours without getting bored. They were worse at the boring library work of checking whether anyone had already written it up, and that is the step that decides whether something counts as a discovery.

Some archaeologists bristle when outsiders with new gadgets parachute in, and that frustration is fair. Fieldwork, permits and respect for the people who did the slow work still matter. Miotti handed the site to INAH and the researchers who mapped it first, which is the right call.

The bigger takeaway is not about pyramids. For the cost of some computer time, one curious person with no training in the field can now run a search that used to take a funded team. Miotti thinks the only limits left are curiosity, imagination and computing power. He wants to scan the whole planet and let the agents look.

After this week, a better plan might be to let the agents look and then let them call an archaeologist before anyone hits post.