The Antikythera mechanism, recovered from a Greek shipwreck and dated to roughly 100 BC, was a hand-cranked bronze computer that tracked the heavens, of unknown origin and mechanically unmatched for more than a thousand years

In the spring of 1902, an archaeologist at the National Archaeological Museum in Athens noticed that a corroded lump of bronze, pulled the previous year from a shipwreck off the island of Antikythera, had a gear wheel set into it. It had come up alongside statues, coins and amphorae, and for a time it drew far less attention than the marble.

The gear changed that.

What the museum held, it turned out, was a hand cranked bronze instrument that modelled the sky. Turn the crank and pointers moved across dials to show the position of the Sun and Moon, the phase of the Moon, the dates of a lunisolar calendar, the timing of eclipses, and the positions of the five planets known in antiquity. Nothing of comparable mechanical complexity survives from anywhere for more than a thousand years after it was made.

What the machine actually did

Only about a third of the original device survives, now split into 82 fragments held in Athens, including thirty corroded bronze gearwheels. The housing was a wooden case roughly 34 centimetres on its longest side. Its dating is not settled. The shipwreck is usually placed around 60 BC, and the object itself is dated variously to the second or first century BC, which is where the familiar figure of roughly 100 BC comes from.

Who built it, and where, remains unknown.

The front face carried a calendar and a display of the planets, Sun and Moon. The back carried two large spiral dials. One tracked the Metonic cycle, the nineteen year period over which lunar months and solar years come back into step. The other tracked the Saros cycle used to anticipate eclipses, and a smaller dial marked a four year cycle of Panhellenic games. This was not a clock in any modern sense but a calculator for astronomical cycles, built to be read rather than programmed, and the “first computer” label it often attracts flatters the analogy more than it clarifies it.

What “unmatched for a thousand years” leaves out

That nothing rivalled it for a millennium is defensible, with one qualification worth keeping. It is a statement about what survives. Geared devices of similar ambition do not reappear in the physical record until medieval astronomical clocks in Europe. But the Roman writer Cicero, within a generation or two of the shipwreck, described mechanical models of the sky built by Archimedes and by Posidonius. If those accounts are accurate, the Antikythera mechanism was not a single object without ancestors or peers. It was the one example that happened to end up on a seabed and stay there.

That distinction matters. A lone marvel invites the story of an isolated genius, while a sole survivor of a lost tradition is a different and, in our reading, more interesting proposition.

What the recent work has added

Modern imaging did most of the decoding. A 2006 study led by Tony Freeth, published in Nature, used X-ray computed tomography to read inscriptions inside the fragments and resolve the back of the machine. In 2021 the UCL Antikythera Research Team, again with Freeth as lead author, published a reconstruction of the front display in Scientific Reports, arguing that it showed the planets in geocentric order, marked by small spheres. The authors called it a “creation of genius.” It is a proposed model that fits the surviving evidence and the inscriptions, not a recovered blueprint, and they present it that way.

A quieter result arrived in 2024. Graham Woan and Joseph Bayley, astronomers at the University of Glasgow, applied statistical methods drawn from gravitational wave analysis to the broken calendar ring, working from hole positions measured by an earlier team. Writing in the Horological Journal, they estimated the ring once held 354 or 355 holes rather than 365, which points to a lunar year rather than the Egyptian solar calendar. A count of 360 was strongly disfavoured. Their analysis also found the holes placed with an average radial variation of about 0.028 millimetres.

The question of whether it worked

Not every recent finding burnishes the object. In April 2025 two engineers at Argentina’s National University of Mar del Plata, Esteban Szigety and Gustavo Arenas, posted a simulation on arXiv asking whether the mechanism would have run at all. Their model combined the effect of its unusual triangular gear teeth with a published estimate of its manufacturing errors. The teeth alone caused little trouble. The manufacturing errors did not, and on their assumptions the device would likely jam before the solar pointer completed about four months of motion.

The paper is a preprint and has not been peer reviewed, and the authors are careful about their own conclusion. They note it cuts two ways. Either the mechanism did not function smoothly, or the error estimate they relied on overstates the real imprecision of the original. A device this elaborate is an odd thing to build if it could not be used, which is itself a reason to treat the jamming result as a prompt for further measurement rather than a verdict.

Roughly two thirds of the mechanism is still missing. Divers and archaeologists have returned to the Antikythera wreck repeatedly, including in expeditions over the past decade, without recovering the rest of it. Until more of it comes up, the reconstructions remain arguments built from a third of the evidence, and the most basic questions, who made it and exactly when, stay open.