In the spring of 1900, sponge divers sheltering from bad weather near the Greek island of Antikythera found a shipwreck on the seafloor. It was loaded with the sort of ancient treasure people know how to recognize: marble and bronze statues, jewelry, coins, glassware. Divers and archaeologists spent the next year bringing it up.
Among it all was a corroded lump of bronze and wood. It didn’t look like much until it broke apart and someone noticed teeth inside it. Gear teeth.
The bronze wheels were cut with surprising precision and packed together in a mechanism built sometime around the second or first century BC. The fragments eventually revealed what we now call the Antikythera mechanism: a hand-operated astronomical computer that could represent the motions of the heavens, track calendars, predict eclipses, and model the irregular motion of the Moon.
It had been sitting underwater for roughly two thousand years.
Fragment A of the Antikythera mechanism. National Archaeological Museum, Athens. Photo: Marsyas/Wikimedia Commons.
A universe in a box
Calling the mechanism a computer can sound like we’re trying to make an ancient artifact seem more modern than it was. In this case, though, the description is pretty literal.
The mechanism seems to have been housed in a wooden case roughly the size of a large shoebox, with dials on the front and back and inscriptions explaining how to use it. Turn a crank or knob to select a date and the gears inside translated that input into several astronomical outputs. Pointers showed the positions of the Sun and Moon. A small black-and-white ball showed the Moon’s phase. Other displays tracked longer cycles and indicated when eclipses might occur.
It wasn’t modern astronomy. The mechanism worked from the geocentric model of its time. But its maker had figured out how to turn astronomical theories into physical ratios. If one celestial cycle took a certain number of days relative to another, gears with the right numbers of teeth could reproduce the relationship mechanically.
That is, in a very real sense, computation—just done with bronze wheels instead of transistors.
The Moon is probably my favorite part. Its apparent speed across the sky isn’t constant, and the mechanism reproduced that variation with a clever pin-and-slot arrangement coupled to gears. A mathematical account of lunar motion became a changing mechanical speed. More than a thousand years before the great mechanical clocks of medieval Europe, someone was building a machine that could imitate the Moon’s uneven progress across the sky.
Modern reconstruction of the mechanism’s front. Photo: Mogi Vicentini/Wikimedia Commons.
The back was just as elaborate. Two large spiral dials represented long calendars. One tracked the 19-year Metonic cycle, in which 235 lunar months come remarkably close to 19 solar years. Another tracked the Saros cycle of roughly 223 lunar months, which could be used to anticipate eclipses. Inscriptions inside the eclipse dial seem to have described characteristics of the expected event.
There was even a calendar for athletic games, including the Olympics. I like this detail. The same little machine that modeled the heavens also reminded its owner when it was time to go watch people throw javelins.
Only about a third of the original mechanism survives, in 82 fragments, so every reconstruction is partly an argument. Much of what we know has come from looking inside metal that can’t simply be taken apart. X-ray and CT imaging have exposed hidden gear teeth and inscriptions. Researchers have spent decades matching fragments, counting teeth, reconstructing missing wheels, and checking the results against what we know of Greek astronomy.
A 2021 reconstruction by researchers at University College London proposed a solution for much of the missing front gearing, including displays for the five planets known in antiquity. It’s an impressive reconstruction, but it isn’t a recovered blueprint. Large parts of the front are gone, and other interpretations remain possible.
Even the number of holes around one damaged calendar ring is still being debated. In 2024, researchers at the University of Glasgow applied Bayesian analysis originally developed for gravitational-wave astronomy to measurements of the surviving holes. Their analysis favored 354 divisions, consistent with a lunar calendar, rather than the 365 one might expect for a solar calendar. Other Antikythera researchers have challenged the conclusion. There is something wonderful about scientists using techniques developed to study ripples in spacetime to argue over how many holes a Greek machinist drilled into a bronze ring two thousand years ago.
One reconstruction of the mechanism’s gearing, including proposed gearing for the five known planets. Wikimedia Commons.
The machine and its maker
We don’t know who built it.
The ship probably sank sometime in the first century BC while carrying luxury goods through the Mediterranean, though the mechanism itself may have been older. Its inscriptions and design place it firmly in the Greek intellectual world. Scholars have proposed origins around Rhodes and connections to traditions associated with Hipparchus, Posidonius, or Archimedes. None of that gives us the name of the person who actually sat at a bench and cut the gears.
I find that person almost as interesting as the machine.
Astronomy supplied the theory, but theory doesn’t make a gear turn. Someone had to choose the tooth counts, lay out the wheels, cut bronze teeth small and regular enough to mesh, fit several gear trains into a compact case, engrave instructions, and make the finished device usable by somebody else. Mathematics, astronomy, metalworking, and ordinary workshop skill all meet inside this one object.
It doesn’t look like a crude first experiment either. The density of the design and the range of functions suggest a craft that had already developed some sophistication. There are signs of mistakes and repairs, and recent research has raised questions about how smoothly reconstructions using the surviving geometry would actually have run. Ancient precision was still ancient precision. If anything, that makes the machine easier to appreciate. Its maker was solving a difficult mechanical problem with hand tools, bronze, geometry, and whatever knowledge could be passed around a workshop.
I keep trying to imagine seeing it when it was new: opening a wooden case, turning a handle, and watching the cosmos move.
What happened to the others?
Eventually the mechanism raises a question its gears can’t answer: where are the other ones?
No other geared machine of comparable complexity survives from antiquity. That absence once made Antikythera seem almost impossible, as though somebody had dropped an object from a much later century onto a Roman ship. But ancient writers left hints that devices like it weren’t unimaginable. Cicero described mechanical celestial models associated with Archimedes and, in his own lifetime, Posidonius. Those accounts don’t tell us what gears were inside them. They do suggest that mechanically representing the heavens was a recognizable kind of Greek technology rather than an idea modern archaeologists invented to explain one strange wreck.
And this is where I think the wreck itself is worth remembering.
Bronze is valuable. Machines get dismantled. Metal gets melted down and reused. Wood rots. Workshops burn. Devices become obsolete. Instructions disappear because everyone who knows a craft assumes someone will be around to teach the next apprentice. A complicated bronze machine has almost every quality you could want in an artifact that won’t survive two thousand years.
Antikythera survived because the ship carrying it sank. The sea put it beyond the reach of generations that might otherwise have repaired it, scavenged it, recycled it, or simply thrown it away.
We can count what survived. We can’t count what disappeared.
That doesn’t mean the ancient Mediterranean was full of mechanical computers, or that some forgotten industrial revolution was about to happen. One extraordinary object can’t support that story. But the absence of other surviving machines can’t tell us they never existed either. Before the divers reached the wreck, we had zero Antikythera mechanisms. After they pulled up one ugly lump of bronze, our estimate of what an ancient craftsman could build had to change.
Surviving fragments of the Antikythera mechanism at the National Archaeological Museum in Athens. Photo: Peulle/Wikimedia Commons.
It’s a small caution about the smooth lines we tend to draw through technological history. Those lines are made from whatever evidence happened to reach us. Stone temples survive better than wooden workshops. Pottery survives better than textiles. Monumental inscriptions survive better than everyday instructions. A bronze machine on land is useful raw material for somebody else. A bronze machine sealed under the sea can become a message to people two millennia later.
But I don’t think the mechanism needs that larger lesson to be worth staring at. Someone in the ancient Greek world built a portable machine of dozens of gears that turned astronomical knowledge into motion. It tracked the Sun and Moon, anticipated eclipses, coordinated calendars, and may have displayed the wandering planets. Then it disappeared into the sea.
Two thousand years later, we’re still figuring out exactly what it could do—and I suspect that’s part of why I like it so much.