Resource Hubs: For more long-form features on the people, machines, and discoveries that shaped how the modern world works, browse the Growth List editorial collection.
Think of a pocket watch. Now strip away the spring, the escapement, the glass, the centuries of refinement, and bury what remains in saltwater for two thousand years. Pull it out off a sponge-diving boat near the Greek island of Antikythera, and you have a corroded lump that nobody recognizes as a machine for almost half a century. That lump turned out to contain at least 30 hand-cut bronze gears, a 223-tooth wheel keyed to an eclipse cycle the Babylonians had worked out centuries earlier, and pointers that could show you where the Sun and Moon would sit in the zodiac years into the future.
It is called the Antikythera mechanism. It should not exist.
The shipwreck nobody was supposed to find
In the spring of 1900, a Greek sponge diver named Elias Stadiatis surfaced from a wreck off Antikythera with a story about bronze limbs and rotting timber on the seafloor, forty-five meters down. The ship had gone down sometime in the first century BC, probably carrying loot from the eastern Mediterranean back to Rome. Over the following year, with the help of the Hellenic Navy, divers brought up statues, amphorae, coins, and a fused lump of corroded metal nobody paid much attention to.
In May 1902, the archaeologist Valerios Stais was examining the haul at the National Archaeological Museum in Athens when he noticed a gear wheel poking out of one of the rocks. It was, as Ars Technica recounts, the first hint that the shipwreck had carried something stranger than statues.
For fifty years, the lump sat in a museum case while classicists argued about whether ancient Greeks could have built such a thing. The consensus was: probably not. The fragments were filed away as a curiosity.
Derek de Solla Price and the X-rays
In the early 1950s, a British-born historian of science named Derek J. de Solla Price began studying the fragments, first through photographs, and in 1958, on a research grant from the American Philosophical Society, he traveled to Athens to examine them in person. The following year he laid out his conclusion in Scientific American, in an article titled “An Ancient Greek Computer”: the lump was a geared device that modeled the motions of the heavens. Two decades later, working with the Greek physicist Charalambos Karakalos, who imaged the corroded pieces with X-rays and gamma rays, Price published his full 70-page reconstruction, Gears from the Greeks, in the Transactions of the American Philosophical Society in 1974.
Most classicists were not convinced. The claim was that a Greek workshop had built something with the mechanical sophistication of a medieval European clock, more than a thousand years before any such clock existed. The simpler explanation, for most of the field, was that Price was wrong.
Price was not wrong.
What the CT scans showed
Starting in the early 2000s, a team using high-resolution computed tomography began imaging the fragments at micron-scale resolution. The scans revealed inscriptions hidden inside the corrosion. They revealed teeth on gears nobody had been able to see. They revealed, layer by layer, a machine.
The device, as currently reconstructed, lived inside a wooden case roughly the size of a mantel clock. A hand crank on the side drove the gear train. On the front face, pointers showed the position of the Sun and Moon in the zodiac, the Moon’s phase, and the date in two calendars at once: the Egyptian solar calendar and a Greek lunisolar calendar.
On the back, two spiral dials handled the long cycles. One tracked the Metonic cycle, the 19-year period after which the lunar and solar calendars realign. The other tracked the Saros cycle, an 18-year, 11-day pattern of eclipses the Babylonians had measured by watching the sky for centuries. Turn the crank forward a few times and the device told you when the next solar eclipse was coming.
The surviving fragments contain roughly 30 gears. Estimates of the original total run higher because pieces are missing. The teeth are triangular, hand-cut, individually filed. The largest preserved gear carries on the order of 223 teeth, and 223 is the number of lunar months in the Saros eclipse cycle, the number cut into the gearwork that drives the eclipse dial.
That detail tells you what the maker was doing. He was not building a clock. He was building a physical model of Babylonian and Greek astronomical mathematics, with the cycles cast into the tooth counts. The math was already centuries old by then. The achievement was translating the math into bronze.
For comparison: the earliest verified European astronomical clocks with comparable gear complexity, like the Strasbourg astronomical clocks, appear in the 14th century. The mechanism predates them by roughly 1,400 years.
The day-zero problem
Any gear-driven calculator needs a starting position. You have to set the dials to a known sky on a known date before the cranking means anything. In 2022, a team led by Aristeidis Voulgaris of the Thessaloniki Directorate of Culture and Tourism published a preprint arguing they had identified the calibration date: December 22, 178 BC, the day of an unusually long annular solar eclipse visible from the eastern Mediterranean, with the winter solstice beginning the following day.
“Any measuring system, from a thermometer to the Antikythera mechanism, needs a calibration in order to [perform] its calculations correctly,” Voulgaris told New Scientist, as Ars Technica reported. The device would not have been a precision instrument in the modern sense, the team argued, but its Saros spiral carried enough astronomical information to forecast solar and lunar eclipses years ahead.
If the date holds up, it means a specific person, in a specific workshop, on a specific winter day in the second century BC, sat down and set the pointers.
There is a complication. In 2025, two Argentine engineers, Esteban Guillermo Szigety and Gustavo Francisco Arenas of the National University of Mar del Plata, posted a preprint arguing the mechanism may never have worked.
Their simulation, reported by Gizmodo, combined two known problems. The triangular tooth profile produces non-uniform motion, with each tooth accelerating and decelerating as it engages. Modern gears use an involute profile to avoid this. And the manufacturing tolerances measured from the corroded surviving teeth are loose enough that a working gear train should jam.
“Our findings indicate that while the triangular shape of the teeth alone produces negligible errors, manufacturing inaccuracies significantly increase the likelihood of gear jamming or disengagement,” Szigety and Arenas wrote. Their conclusion offered two options: either the device never worked, or the original tolerances were tighter than the corroded fragments suggest.
The second option is more likely. Two thousand years of seawater does not preserve the original geometry of a tooth face. “Any attempt to apply precision measurements on the current condition of the gears [and] axes includes the effect of the deformation,” Voulgaris told New Scientist in response. The Argentine engineers themselves note their results are speculative and that intentionally building an unusable device makes no sense for the level of craftsmanship involved.
Cicero’s hint, and a thousand-year gap
There is one written clue from the period. The Roman statesman Cicero, writing about a century after the mechanism was built, described mechanical models of the heavens constructed by Greek thinkers, attributing one to Archimedes. He treats these as remarkable but real objects. He is not describing a fantasy.
If Cicero is accurate, the Antikythera mechanism was not unique. It was one survivor of a small tradition of astronomical instruments built in Hellenistic workshops, probably on Rhodes or in the Greek-speaking eastern Mediterranean. Every other example has been lost. The bronze was melted down, the wood rotted, the ships made it home.
This one sank.
What it did, as reconstructed: turn the crank and the pointers advance through the days. The Sun pointer creeps along the zodiac at the right speed. The Moon pointer moves faster, completing a full circuit roughly every 27 days. A small rotating ball, half black and half white, showed the Moon’s current phase.
The Saros dial on the back face carried glyphs marking predicted lunar and solar eclipses, with letters indicating the time of day. The Metonic dial handled the calendar. There is evidence of a separate dial tracking the four-year cycle of Panhellenic athletic games, including the Olympics. The mechanism could tell you when the next Olympics were and where the next eclipse would fall.
That second function mattered more than it sounds. In the Greek world, eclipses were political and religious events. Knowing one was coming was a form of power.
After the mechanism sank around 65 BC, nothing mechanically comparable shows up in the archaeological record for more than a millennium. The Islamic world built astrolabes and water clocks. Chinese engineers built escapement-driven clocks in the 11th century. European cathedral clocks appear in the 13th and 14th centuries. None of them, for a very long time, modeled the heavens with the specific gear-train approach the Antikythera mechanism used.
The knowledge did not transfer. Whoever knew how to cut 223 teeth into a bronze disc and key it to the Saros cycle took that knowledge with them. The workshops closed. The apprentices died. The math survived in Ptolemy’s books, but the metal disappeared.
Still corroded, still being read
The fragments remain in Athens, in the National Archaeological Museum, in climate-controlled cases. Every few years, a new imaging technique pulls another inscription out of the corrosion. In 2016, a multinational team identified about 3,500 characters of text on the surviving surfaces, including what appears to be a user manual carved directly into the bronze plates.
Divers have returned to the wreck site repeatedly, most recently in expeditions run with the Greek authorities and Woods Hole Oceanographic Institution. They have brought up more bronze pieces, ceramic, human remains, and parts of the ship’s structure. Pieces of the mechanism may still be down there, buried in sediment a hundred and fifty feet below the surface.
On December 22, if the Voulgaris calibration is right, the dials were set for the first time, by someone whose name nobody knows, in a workshop nobody has found, on the eve of the longest night of the year.