The Antikythera Mechanism: a Greek computer from a shipwreck

Scene · reconstructed from the paperworkPicture it. A merchant ship, somewhere off the small Greek island of Antikythera, more than two thousand years ago. It goes down with a cargo of Greek treasures. Two thousand years later it is still there, and among what it carries is a corroded lump of bronze and wood that nobody aboard would have thought the most precious thing on the ship.

The wreck was found and worked around 1900 and 1901. The 2006 Nature paper[freeth-2006] dates the discovery of the Mechanism to 1901, in a Roman shipwreck, and the Research Project[amrp-museum] says it has been in the National Archaeological Museum in Athens ever since. Statues and fine glassware from the same wreck went to the museum too. The Mechanism was the odd one out: broken, crusted, easy to overlook.

What came up is not one object but many pieces. Today there are 82 fragments[freeth-2021], and only about a third of the original survives, including 30 corroded bronze gearwheels. The largest fragment, A, holds 27 of those 30 gears. The pieces are damaged and covered in accretions, and yet, as the 2021 team writes, they are rich in evidence at the millimetre level, with thousands of tiny text characters buried inside them.

Those characters are the first witness in this file. There is no one alive who saw the ship sink, and the divers' own accounts are not something the Archive has opened. What testifies instead is the object: its gears, and the Greek words engraved on its plates. The museum[nam-collections] calls it a scientific instrument of the 1st century BC used for astronomical and calendar forecasts. Its own visitor guide goes further and lists it, dated 150 to 100 BC[nam-brochure], as the oldest surviving astronomical and calendrical portable mechanical computer.

That is a large claim for a museum to print, and it is not hype. It is the settled reading of a century of work on the fragments, which the rest of this file follows. The 2006 team[freeth-2006] put it plainly: the device is technically more complex than any known device for at least a millennium afterwards.

So the night of this file is a wreck, and then a very long wait. The ship's cargo sat on the sea floor for two thousand years. The Mechanism then sat in a museum for most of another century before anyone could see inside it. What they found when they did is the subject of the next chapter.

One line is worth keeping in mind from the start. Everything known about the Mechanism comes from the fragments themselves and from the people who have studied them. There is no ancient manual outside the object and no written account of who made it. Its own inscriptions are the closest thing to a manual, and the Archive treats them as documents.

The paper trail of the Antikythera Mechanism is a century of research, and most of it is in peer-reviewed journals.

The early work set the frame. The 2021 team[freeth-2021] records that Albert Rehm's research notes of 1905 to 1906 already imagined a planetarium with ring displays, and that Derek de Solla Price's classic study, Gears from the Greeks (1974), proposed lost gearing for the planets. Michael Wright later built the first workable front gearing. Those are the people the modern papers stand on.

The turn into hard data came in 2005. A British and Greek team, working with the National Archaeological Museum[amrp-museum], imaged the fragments with surface imaging and high-resolution X-ray tomography. Their 2006 Nature paper[freeth-2006] reconstructed the gear function and doubled the number of deciphered inscriptions. It found that the device predicted lunar and solar eclipses using Babylonian arithmetic cycles, and that its gearing realised Hipparchos' theory of the Moon's uneven motion. Two of its authors were museum staff who had catalogued the fragments.

In 2008 the same group read the calendar. The second Nature paper[freeth-2008] identified all twelve month names on the 19-year Metonic dial, and they were Corinthian. That points to the Corinthian colonies of northwest Greece or to Syracuse in Sicily. The paper also showed that a small dial follows the four-year cycle of the Olympiad and its Games, and that the lower back dial is a Saros eclipse dial of 223 lunar months. The museum's explainer[nam-digital] sets out those cycles: Metonic, Callippic, Saros and Exeligmos.

In 2014 Christián Carman and James Evans went after the date with astronomy rather than lettering. Their paper[carman-evans-2014] argues that the eclipse dial would work best if it started with the full Moon of 12 May 205 BCE. As the press report[physorg-2015] explains, letter-style estimates had put construction at 150 to 100 BC.

In 2021 a UCL team turned to the lost front. Their Scientific Reports paper[freeth-2021] proposes a ring display of the Sun, Moon and five planets, matched to inscriptions that describe a Cosmos with little spheres and a little golden sphere for the Sun. UCL's release[ucl-2021] calls it a hand-powered device that predicted the positions of the Sun, Moon and planets and eclipses.

The inventory: a museum object in 82 pieces, the museum's own descriptions, four major peer-reviewed papers, and a research project that keeps the bibliography. For an object two thousand years old, that is an exceptionally good file.

The turn in this file is not a single discovery. It is the moment the Mechanism stopped being a mystery about what it was and became a set of questions about how and when.

That moment is 2006. Before the X-ray work, the gears and inscriptions were only fragmentary, and its specific functions had remained controversial, as the 2006 paper[freeth-2006] says in its own abstract. After it, the back of the machine was solved: eclipse prediction on Babylonian cycles, and a pin-and-slot gear that models the Moon's changing speed. The 2021 team[freeth-2021] states it as settled background: the 2005 imaging decoded the structure of the rear of the machine.

Once that was done, the popular framing changed too. The Mechanism is often presented as an impossible object, an out-of-place artefact that proves someone else, ancient or otherwise, must have built it. The research record points the other way. The 2021 paper[freeth-2021] describes it as combining cycles from Babylonian astronomy, mathematics from Plato's Academy and ancient Greek astronomical theory. Its month names are Corinthian Greek[freeth-2008]. Its eclipse scheme, on Carman and Evans's reading[carman-evans-2014], fits Babylonian-style arithmetic better than trigonometry. Everything about it is Greek and Hellenistic in origin. What is astonishing is the level, not the source.

That is why this file is labelled Confirmed and graded Recorded. The core claim, that the Mechanism is an ancient Greek geared astronomical calculator, is documented by the museum that holds it[nam-collections] and by the peer-reviewed work of the teams that imaged it. There is no official UFO file here, and none is needed.

The arguments that remain are serious and scholarly. The date is one. The museum gives 150 to 100 BC[nam-brochure]; the eclipse dial's epoch points to 205 BCE[carman-evans-2014]; and Alexander Jones, quoted by Phys.org[physorg-2015], notes that 205 BC could be a historic start date chosen by the maker rather than the year it was built. The 2008 team, working from the calendar, linked the device to Syracuse, home of Archimedes, but did not claim he made it.

The other argument is the front of the machine, where most of the evidence is missing. The 2021 model is the best fit yet to the surviving data, and its authors say plainly that they cannot claim it is a replica[freeth-2021] of the original. Our Other Side[freeth-2021] box takes that caution at full strength.

What is still open about the Antikythera Mechanism is real research, and the questions are sharp.

First, when it was made. The museum dates it to 150 to 100 BC[nam-brochure] and calls it a 1st-century BC instrument[nam-collections] on another page. The 2006 team put it around the end of the 2nd century BC[freeth-2006]. The eclipse dial's start date, on Carman and Evans's analysis[carman-evans-2014], is 205 BCE. Those are not contradictions so much as different clocks: the style of the letters, the date of the wreck, and the date the dial was set to. Which of them is the building date is not settled, and Alexander Jones[physorg-2015] is on record as unconvinced by the earliest one.

Second, where it was made, and by whom. The Corinthian month names point to northwest Greece or Syracuse[freeth-2008]. Syracuse brings in Archimedes, who died there in 212 BC, and Cicero's description of an Archimedean device, quoted by the 2021 team[freeth-2021]. That is a possible workshop tradition, not an identification. No maker's name has been read on the fragments.

Third, what the front looked like. The 2021 ring model fits the inscriptions and the physical clues, but much is missing, and some features in it, such as a hand showing the Moon's nodes, are labelled by the authors themselves as hypothetical. The paper[freeth-2021] is explicit that the model is not a replica.

Fourth, the rest of the wreck. More fragments may lie on the sea floor. The Archive has not opened any official report of new finds, so this stays a question.

Fifth, the record of the dive itself. The divers who first worked the wreck around 1900 left accounts that are often retold. The Archive has not opened a primary text of them, so no diver's words are quoted in this file.

The honest position: Confirmed, the Mechanism is an ancient Greek geared astronomical calculator, held by the National Archaeological Museum[amrp-museum] in Athens since 1901, that tracked calendars and predicted eclipses on Babylonian cycles. Recorded, its month names are Corinthian and one dial follows the Olympiad. Argued, its exact date and maker. Modelled but not recovered, its front display.

• When was it built, and when was it set?
• Which workshop made it?
• What did the front really show?

The folder stays open and the file stays a draft for the editor.