A gear-driven box built more than 2,000 years ago predicted eclipses down to the hour

A corroded bronze box pulled from a Mediterranean wreck in 1901 could tell whoever turned its hand crank not just that an eclipse was coming, but the hour it would start. Mathematician Tony Freeth of the Antikythera Mechanism Research Project laid out the arithmetic behind that claim in a 2014 paper, showing that the device’s makers built a working prediction scheme into a spiral of engraved bronze small enough to fit in a shoebox.

Freeth’s case rests on a single dial on the back of the instrument, spiraling through four full turns and inscribed with symbols the size of rice grains. Reading it correctly meant reconstructing a system nobody had operated in roughly two thousand years, using nothing but fragments of corroded metal and the faint traces of Greek lettering hidden inside them.

Tony Freeth solves a fifty-year-old glyph puzzle

Freeth’s paper, published in PLOS ONE, worked from small pictograms called glyphs arranged around what researchers call the Saros Dial, a four-turn spiral tracking the 223-lunar-month cycle ancient astronomers used to anticipate eclipses. Each glyph carried a letter for the Moon or the Sun, a note on whether the eclipse would even be visible, and a coded time. Freeth showed how turning the mechanism forward to a given month moved a pointer to a specific glyph: Month 78 on the dial, marked with the index letter T, points to inscriptions describing a solar eclipse timed to the first hour after dawn.

Reconstructing how those times were calculated took a separate model, built on a Babylonian arithmetic technique known as zigzag functions rather than the Greek geometric methods historians had assumed. Tested against the surviving glyph times, Freeth’s model produced a lunar eclipse timing error of 1.4 hours and a solar eclipse timing error of 1.9 hours, for a combined average error of 1.7 hours across the predictions it could check. Freeth was blunt about the result in his own paper, writing that the scheme “was not entirely accurate, but it was an astonishing achievement for its era.”

150 feet down, a Roman cargo hid a Greek calculator

The mechanism did not come from an orderly excavation. Sponge divers from the Greek island of Symi found the wreck in 1900 while sheltering from a storm, resting at a depth of roughly 148 feet off the island’s coast, and a Hellenic Royal Navy salvage operation hauled up bronze and marble statues, glassware, coins, and the corroded lump that turned out to hold the gearing, according to the mechanism’s documented recovery history. It sat overlooked in Athens for months until archaeologist Valerios Stais, examining the museum’s haul in 1902, noticed a gear wheel embedded in the rock-like mass and legible Greek inscriptions beside it, a find later credited as the moment the device’s true nature was recognized.

The ship that carried it has been dated separately from the device itself. The National Archaeological Museum of Athens, which now holds all 82 surviving fragments, places the wreck at roughly 60 to 50 BC, while dating the cargo it carried, including statuary and the mechanism, across a much wider span from the fourth to the first century BC. That gap between when the ship sank and when its most sophisticated passenger was actually built is exactly what let Freeth’s team push the mechanism’s likely origin earlier than the wreck date alone would suggest.

A design that may trace back to Hipparchus of Rhodes

Freeth’s arithmetic pointed toward a strikingly early construction date, calculating that the Saros Dial’s underlying scheme matches a lunar month beginning in May of 205 BC, more than 2,200 years before this article was written and well over a century before the ship went down. Other researchers have proposed dates as late as the early first century BC, but every published estimate places the mechanism’s design more than two thousand years in the past.

Whoever built it was working with theories developed by Hipparchus of Rhodes, the astronomer whose model for the Moon’s uneven orbital speed is embedded in the mechanism’s gearing, which has fed a long-running hypothesis that the device was made in a Rhodian workshop with ties to his circle. Cicero later described a similar geared instrument built by Archimedes, who died in 212 BC, though historians generally treat that as evidence of a broader Hellenistic tradition of astronomical machines rather than a claim about this specific wreck.

What survives is 82 fragments, the largest bearing a bronze gear roughly 13 centimeters across that once carried 223 teeth, a number that matches the eclipse cycle it was built to track. Modern imaging has identified at least 30 of an estimated 35 meshing gears inside the fragments, work that built on X-ray studies begun by physicist Derek de Solla Price in 1971 and continued by later research teams using higher-resolution scanning. No comparably complex geared machine is known to have been built again for well over a thousand years, which is part of why a single spiral dial, deciphered more than a century after divers first brought it to the surface, still counts as a live research problem in 2026.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview