In 1901, sponge divers hauled a lump of corroded bronze and wood out of a Roman-era shipwreck off a small Greek island. For two years nobody paid it much attention. Then someone noticed a gear wheel inside. That lump turned out to be the Antikythera mechanism, a hand-cranked box of bronze gears built more than 2,000 years ago to show where the Sun, Moon, and planets would be on any given day, and to predict eclipses years ahead. It is often called the world’s oldest known computer. More than a century later, researchers are still arguing about how it worked, what was on its dials, and whether it ever ran smoothly at all.
A lump of rock with a gear inside
The story starts with a wait for the wind. Around Easter 1900, a crew of sponge divers from the island of Symi, led by Captain Dimitrios Kontos, stopped at Antikythera, a rocky island between Crete and the Greek mainland, to wait for better winds on their way to sponge grounds off North Africa. Diving there to pass the time, they found an ancient wreck about 45 meters (148 feet) down. Later that year they reported the find, and from November 1900 through 1901, working with the Greek navy, they brought up bronze and marble statues, pottery, glassware, jewelry, and coins.
The mechanism came up in 1901, probably in July. It did not look like much. Its bronze had turned into a crumbly green mineral called atacamite, and it cracked and shrank once it was out of the water. Museum staff in Athens were busy piecing together the statues, so the lump sat for two years.
On May 17, 1902, the archaeologist Valerios Stais and his cousin, the politician Spyridon Stais, noticed a gear wheel set into one of the pieces. Valerios Stais thought it might be an astronomical clock. Most scholars at the time thought that was impossible. Gearing that complex was not supposed to exist until many centuries later.
The ship itself sank around 70 to 60 BC. The mechanism was built before that, though experts disagree on exactly when. Estimates range from about 205 BC to about 87 BC.
What it did when you turned the crank
Today the mechanism survives as 82 fragments, kept at the National Archaeological Museum in Athens. Only about a third of the original is left. Inside those pieces are 30 bronze gears. The biggest is about 13 centimeters (5 inches) across and once had 223 teeth. Each tooth is a small hand-cut triangle, filed out of a bronze disc.
It all sat in a wooden box about the size of a thick book, roughly 34 by 18 by 9 centimeters. On one side was a hand crank. Turning it drove the big gear, which turned all the others at once, like the hands of a clock moving together.
The front had a large round dial marked with the 12 signs of the zodiac, the band of sky that the Sun, Moon, and planets travel through. Pointers showed where the Sun and Moon were on that band. A small ball, half pale and half dark, turned to show the phase of the Moon, from new to full and back again.
The cleverest part is how it handled the Moon’s speed. The real Moon moves faster across the sky at some times of the month and slower at others, because its orbit is not a perfect circle. The Greek astronomer Hipparchus worked out a theory for this in the second century BC. The mechanism copies it with a pin on one gear that slides in a slot on another, so the Moon pointer speeds up and slows down on its own. It is the oldest known use of this kind of gearing.
The back held two big spiral dials. One followed the Metonic cycle, the fact that 235 lunar months add up almost exactly to 19 years, which ancient calendar makers used to keep lunar and solar calendars in step. The other followed the Saros cycle of 223 lunar months, a little over 18 years, after which eclipses repeat in a similar pattern. Its marks warned of coming eclipses of the Sun and Moon. A smaller dial kept track of the four-year cycle of the big athletic games, including the Olympics.
Reading a 2,000-year-old label
For decades, most of what researchers knew came from the outside of the fragments. In 1951, the Yale historian of science Derek de Solla Price took up the puzzle. In 1971 he and the Greek physicist Charalampos Karakalos took X-ray and gamma-ray images of the pieces, and Price published his findings in 1974.
The big leap came in 2005. A team led by Mike Edmunds of Cardiff University used X-ray computed tomography, a kind of 3D X-ray scan, to see inside the corroded lumps. The results, published in the journal Nature in 2006, mapped the gears in detail. The scans also showed thousands of tiny letters engraved on the plates, hidden under the corrosion.
By 2016, a team led by the classicist Alexander Jones had read about 3,400 characters, some only about a millimeter tall. Jones thinks the full device may have carried up to 20,000. The text had been called a user manual, but it turned out to be closer to a museum label. It does not tell you how to run the machine. It explains what each dial shows, and it assumes the reader already knows a fair bit about astronomy.
Those inscriptions held a surprise. In the sections about the planets were two numbers, 462 years for Venus and 442 years for Saturn. They are very accurate cycles for each planet’s movement as seen from Earth, and nothing in older Babylonian astronomy explained where they came from. In 2021, a team at University College London led by Tony Freeth used those numbers to design new gear trains for all five planets the Greeks knew: Mercury, Venus, Mars, Jupiter, and Saturn. Their model, published in Scientific Reports, shows the planets moving on rings around the front dial, marked by small beads. It is the first reconstruction that its authors say fits all of the surviving evidence. It is still a proposal, because the planet gears themselves have not survived.
A calendar ring that was counted twice
One small detail shows how much detective work is still going on. Beneath a ring on the front dial is a circle of tiny holes, only some of which survive. For about a century, people assumed there were 365 of them, one for each day of the Egyptian solar calendar written on the ring.
In 2020, a team led by Chris Budiselic, who runs a YouTube channel and was trying to build an accurate replica, measured the surviving holes from the scans and found the count was probably closer to 354. That is the length of a lunar year, twelve months of the Moon. In 2024, two physicists at the University of Glasgow, Graham Woan and Joseph Bayley, ran the same measurements through statistical methods borrowed from gravitational-wave research. Both came out at 354 or 355 holes. They found 354 about 229 times more likely than 360, and they called 365 not plausible. Another paper in the Horological Journal in 2025 reached the same broad answer, a count somewhere between about 350 and 360, and not 365.
If that is right, the ring tracked the Moon’s calendar rather than the Sun’s. It is a reminder that some of the most basic facts about the machine are still being worked out.
Did it actually work?
Here is the awkward question. In April 2025, Esteban Szigety and Gustavo Arenas, two researchers at the National University of Mar del Plata in Argentina, posted a computer simulation of how the gears would mesh. They found the triangular teeth were not a problem on their own. But when they added the manufacturing errors that earlier scans had measured, uneven tooth spacing and gears slightly off center, the model kept jamming or slipping. By their estimate, it would stop before the Sun pointer moved about 120 days, roughly four months.
The authors did not conclude that the machine was a fake or a toy. They pointed out that it seems unlikely anyone would build something so complex that could not move. They suspect the errors in the damaged fragments are bigger than the original errors were, since 2,000 years under the sea warped and shrank the bronze. The paper was posted as a preprint, meaning it had not yet been through peer review when it came out.
There is a separate question of accuracy. Freeth and Jones found that the Mars pointer in their reconstruction could be as much as 38 degrees off at times. That is not a gear problem. It comes from the Greek theories of planetary motion the machine was built on. The mechanism could only be as good as the astronomy of its day.
Why nothing like it shows up for 1,400 years
The strangest thing about the Antikythera mechanism may be how alone it is. No other geared device of anything like this complexity survives from the ancient world. Machines that come close, such as the astronomical clocks of Richard of Wallingford in England and Giovanni de’ Dondi in Italy, did not appear until the 1300s.
It almost certainly was not the only one ever made, though. The workmanship is too sure for a first try. The Roman writer Cicero described a bronze sphere built by Archimedes that showed the motions of the Sun, Moon, and planets. It was carried to Rome after Archimedes died in the siege of Syracuse in 212 BC. Cicero also mentioned a similar device made by his friend Posidonius on the island of Rhodes. Neither is the machine from the wreck, but both suggest there was a small tradition of these devices that has otherwise vanished. Bronze was valuable, and it was usually melted down and reused. The one that survived did so only because it sank.
The bottom line
The Antikythera mechanism is a book-sized model of the sky, with gears that tracked the Sun and Moon, warned of eclipses, kept a calendar of the games, and probably showed all five known planets. After more than a century of study, researchers agree on what most of it did. They are still debating who built it, exactly when, whether its calendar ring counted 354 days or something close to it, and how smoothly it ever ran. If you are in Athens, the fragments are on display at the National Archaeological Museum, alongside modern reconstructions that show how they once fit together.
Further reading
Decoding the Heavens: A 2,000-Year-Old Computer and the Century-Long Search to Discover Its Secrets — Science journalist Jo Marchant tells the full detective story, from the sponge divers of 1900 to the X-ray scans that finally revealed the gears.
A Portable Cosmos: Revealing the Antikythera Mechanism, Scientific Wonder of the Ancient World — Alexander Jones, who led the work on the inscriptions, explains each dial and what it tells us about Greek astronomy, in clear and careful prose.