Prefer to watch? Here's the 3-minute version:

For a hundred years, people walked across the Nazca Desert in Peru and saw nothing unusual. Beneath their feet, scraped into the reddish ground sometime between 200 BC and 650 AD, was a 72-foot killer whale clutching a knife. They missed it. So did the survey planes and the drones. Then researchers pointed an AI at the same desert — and it surfaced the orca, plus 302 other ancient figures, in about six months.

It had taken humans roughly a century to catalog the first 430 of these giant drawings. The AI nearly doubled that number before a single field season was out.

TLDR: Researchers trained an AI to spot faint, eroded patterns across terabytes of aerial imagery — and it found hundreds of 2,000-year-old artworks the human eye kept walking past. The same trick is now uncovering lost cities under jungle and desert worldwide. And the underlying superpower — finding the signal you have gone blind to in your own mountain of data — is one you can borrow today.

Why a Century of Experts Missed It

The Nazca figures are not hidden because they are small — some run over a thousand feet. They are hidden because the desert is enormous and many of them are faint, eroded relief carvings that read as little more than a slight shift in soil color. Combing that much ground on foot would take a lifetime several times over. The data was always there; the bottleneck was human attention. A person can study a handful of images carefully. An AI can study millions and never blink.

So the team — Masato Sakai's group at Yamagata University, working with IBM — fed the model thousands of high-resolution aerial photos, then turned it loose on satellite and drone datasets measured in terabytes. It learned the difference between a deliberate ancient scrape and a random one, and flagged candidates. Then humans hiked out to confirm each one. AI did not replace the archaeologist — it told the archaeologist where to walk.

The Desert Was Just the Beginning

Here is where it stops being a Peru story. The same recipe — let lasers and AI sift impossible volumes of terrain, then send humans to verify — is redrawing the map everywhere:

• In southern Mexico, laser mapping has revealed thousands of Maya structures hidden under jungle.

• In Ecuador's Amazon, it exposed a cluster of cities that once held at least 10,000 farmers, buried under canopy for 2,000 years.

• In the mountains of Uzbekistan, drone scans surfaced two abandoned Silk Road cities that had vanished for centuries.

And the AI did more than count. The new Nazca figures tend to line up along old footpaths — evidence the smaller drawings were meant to be walked past and seen up close, not just admired from the sky. The machine did not just find more dots; it helped explain the picture.

There is a clock running, too: erosion, wildfires, and climate events are erasing sites faster than people can find them. AI is, in a real sense, racing to recover human history before it disappears — and winning.

The Superpower You Can Borrow

Strip away the desert and the lasers, and the breakthrough is simple: AI is freakishly good at finding the one meaningful pattern buried in a pile too big for a human to scan. You almost certainly own a pile like that — the months of receipts you never reconcile, the customer emails you skim, the spreadsheet you have stared at so long you have stopped seeing it. Your own knife-wielding orca is probably in there. You have just been walking over it.

So we built a prompt that turns AI into your personal pattern-hunter. It interviews you about your world first, then coaches you to surface the anomalies and trends you have gone blind to — no pasting, no setup.

The Excavation →

Same prompt, four very different people — four very different orcas surface:

A hundred years of experts walked over a giant orca in the sand. What is hiding in plain sight in your data — just waiting for the right set of eyes?

Same prompt. Your pile of data. Your own orca is in there — go find it.