At least 30 interlocking bronze gears. A hand crank. A pin-and-slot mechanism that reproduced the Moon’s varying speed across the sky, more than 1,600 years before Kepler explained why it varies. Thousands of tiny Greek letters etched onto the surfaces, functioning as an instruction manual. X-ray CT scans completed in the mid-2000s revealed all of this inside a shoebox-sized device of corroded bronze that could predict solar and lunar eclipses, track the irregular motion of the Moon, and chart the positions of the planets known to the ancient Greeks.
The object had been sitting in a museum in Athens for more than a century by then. It was hauled up in 1901 by sponge divers working off the rocky Greek island of Antikythera, pulled from a Roman-era shipwreck lying beneath the Aegean. The lump of corroded bronze sat in a wooden crate at the National Archaeological Museum for months before anyone noticed the gear teeth poking through the crust. What the divers had recovered, archaeologists would eventually realise, was an analogue computer built around 100 BC.
It is roughly 1,400 years older than the next device of comparable mechanical sophistication. Nothing like it appears in the archaeological record again until the astronomical clocks of medieval Europe.
A shipwreck full of marble and one strange box
The wreck itself was a slow-motion discovery. Sponge divers had been driven by a storm to anchor off Antikythera, a barren scrap of rock between Crete and the Peloponnese. When the weather cleared, a diver descended in a copper helmet and canvas suit and surfaced babbling about a pile of rotting corpses on the seafloor. They were bronze and marble statues, half-buried in silt, along with amphorae, glassware, and luxury goods that dated the cargo to the first century BC. The ship had probably been carrying loot from the Greek east to Rome when it went down.
The bronze lump was almost an afterthought. It was only in 1902 that archaeologist Valerios Stais, examining the corroded mass after it had split open in the museum’s humid storerooms, spotted a gear wheel embedded in the rock-like calcite. He suggested it might be an astronomical clock. Most of his colleagues dismissed the idea. Clocks with gear trains were not supposed to exist in the ancient world.
Half a century of arguing about gears
For decades, the device sat in three main fragments under museum glass while scholars argued about what it was. A planetarium? A navigational tool? A modern fake somehow smuggled into the wreck? The breakthrough came in the 1950s, when the British physicist and historian of science Derek de Solla Price began studying the fragments at the Athens museum. Price took X-rays of the corroded bronze and began counting teeth on gears that had not been visible to the naked eye for two thousand years.
In a 1959 paper for Scientific American and a longer 1974 monograph titled Gears from the Greeks, Price proposed that the device was a geared calculator of astronomical cycles. He counted at least 30 gears, identified a differential-style gear train, and dated the mechanism to roughly 80 BC. His tooth counts matched the ratios used by Babylonian and Greek astronomers to model the Moon and Sun.
Even so, much of the device remained unreadable. The corrosion was too thick, the writing too faint, and the gears too tangled to map by eye.
What the CT scanner saw
The real reckoning came in 2005, when an international group called the Antikythera Mechanism Research Project brought a custom eight-tonne X-ray CT scanner to the museum in Athens. The machine could see through bronze that had spent two millennia on the seabed. It produced three-dimensional reconstructions of the gear train and revealed thousands of tiny Greek letters inscribed on the surfaces. An instruction manual, essentially, etched into the bronze.
The scans confirmed what Price had guessed and added a great deal more. The mechanism carried at least 30 surviving bronze gears, with evidence that the original had more. Turning a hand crank on the side rotated a system that drove pointers across dials on the front and back faces. The front dial showed the position of the Sun and Moon against the zodiac and an Egyptian calendar of 365 days. A small rotating sphere, half black and half silver, showed the phase of the Moon.
One of the most elegant pieces of engineering was a pin-and-slot mechanism that reproduced the Moon’s varying speed across the sky, a phenomenon Greek astronomers knew about but explained with eccentric circles rather than ellipses. The bronze gearing recreated the effect mechanically, more than 1,600 years before Kepler worked out why it happened. Detailed reconstructions of the gear train show how the irregular motion was generated by one gear riding eccentrically on another.
The back of the device: eclipses on a spiral
The two dials on the back were even stranger. The upper one was a spiral with 235 divisions, representing the 19-year Metonic cycle that aligns lunar months with solar years. The lower spiral covered 223 lunar months, the Saros cycle that Babylonian astronomers had used for centuries to predict eclipses. Glyphs scattered around the Saros dial marked the months in which a solar or lunar eclipse was likely, often with notes on the time of day and the expected colour of the eclipsed Moon.
Small subsidiary dials tracked the four-year cycle of the Panhellenic games, including the Olympics, and the longer Callippic and Exeligmos cycles used to refine eclipse predictions. A user turning the crank could spin the mechanism forwards or backwards through time and read off which eclipses would happen, when, and roughly what they would look like. A 2024 analysis of the calendar ring’s hole spacing suggested it tracked a 354-day lunar year rather than the 365-day solar one, refining earlier assumptions about how the front dial worked.
Who built it, and for whom
No name is inscribed on the surviving fragments, and the workshop where it was made has never been identified. The favoured candidates, based on the mechanism’s astronomical content and Greek inscriptions, are the school of Hipparchus on Rhodes, the Greek colony of Syracuse where Archimedes had worked a century and a half earlier, or a workshop on the island of Cos. Ancient sources describe a bronze planetarium made by Archimedes that showed the motions of the Sun, Moon and five planets, a description that matches the Antikythera device unsettlingly well.
The wreck has been excavated repeatedly, most recently by teams using advanced diving suits and remotely operated vehicles, and continues to yield bronze fragments, marble statuary, and ship’s timbers. The cargo dates the sinking to the first century BC. The mechanism itself, based on the inscriptions and the astronomical configuration it shows, was probably built between 150 and 100 BC and was perhaps a few decades old when the ship went down.
How well did it actually work
Two thousand years of corrosion makes it hard to say. A 2025 study by researchers at the National University of Mar del Plata in Argentina ran computer simulations of the gear train using the measured tooth shapes and triangular hole spacings on the calendar ring. Their model suggested the mechanism may have jammed frequently if the holes were as irregular as they appear today, raising the possibility that the surviving fragments were distorted by sea-floor pressure, or that the device was partly ornamental, or that the ancient gear-cutter’s tolerances were simply looser than modern reconstructors assume.
Other researchers, including the team behind the most complete physical reconstructions, argue the gearing works smoothly when built to the proportions implied by the inscriptions. The debate is still live. What is not in dispute is the conceptual achievement: someone in the Hellenistic world built a working physical model of the heavens, with the gear ratios chosen to match cycles refined over centuries of Babylonian and Greek observation.
An object out of its time
The strangest thing about the Antikythera mechanism is the silence that follows it. No comparable geared astronomical device survives from antiquity, the Byzantine period, or the early Islamic world for more than a thousand years. The Islamic astrolabes of the ninth and tenth centuries have geared calendars, but nothing approaching this complexity. The first European astronomical clocks of comparable sophistication appear in the fourteenth century, in cathedrals at Strasbourg and Padua. Historians of Greek science have argued that the mechanism implies a lost tradition of precision bronzework. Workshops, apprentices, suppliers of metal stock, all of it vanished without leaving other examples behind.
The device has also become a cultural object in its own right. It inspired the McGuffin in Indiana Jones and the Dial of Destiny, and in 2026 the Swiss watchmaker Hublot announced a partnership tied to the mechanism’s heritage, drawing on its status as the ancestor of mechanical horology.
What it leaves us with
The fragments themselves sit in a climate-controlled case at the National Archaeological Museum in Athens, a few rooms away from the bronze statues recovered from the same wreck. The largest piece is about the size of a paperback book. Stand close to the glass and the gear teeth are visible, small, regular, machined with care by a craftsman whose name was lost two thousand years before anyone thought to look for it.
And that is the part that nags. If a single shipwreck off a barren rock between Crete and the Peloponnese held a working planetarium, what else went down with other ships, in other centuries, that no sponge diver happened to bump into? How many workshops did this tradition contain before it disappeared, and how much further might the gear-cutters have gone if the Hellenistic world had not been overrun and its libraries scattered? The dive teams keep returning to Antikythera. The silt has already given up one object that should not exist. Why would it be the only one?
