A century after divers pulled corroded bronze gears from a Roman-era shipwreck off the island of Antikythera, a team at University College London announced, in March 2021, that it had finally cracked the part everyone had given up on: the front. Lead author Tony Freeth declared theirs "the first model that conforms to all the physical evidence and matches the descriptions in the scientific inscriptions engraved on the Mechanism itself." The headlines said "mystery solved." The problem is that the mystery did not stay solved.

The Antikythera Mechanism is not in dispute as an object — it is, beyond argument, a hand-cranked bronze analog computer from roughly the 2nd century BC that predicted eclipses and tracked the Moon's irregular orbit, with a mechanical sophistication that wouldn't reappear for fourteen centuries. The 2005 X-ray CT campaign by the Antikythera Mechanism Research Project decoded the rear dials. The front — a display the inscriptions describe as a moving model of the cosmos, with the Sun, Moon, and five visible planets — was always the unsolved face, because most of its gearing was simply never recovered from the seabed.

The 2021 paper, published in Scientific Reports, reconstructed that lost front gearing using clever shared gear-trains and rational approximations of the planets' synodic cycles. It is genuinely elegant. It is also, by the team's own admission, partly speculative: the planetary gears have never been found, and Freeth and Alexander Jones have themselves acknowledged the simulated mechanism is inaccurate at points — the Mars pointer can be tens of degrees off near the planet's retrograde loops, a limit of Greek planetary theory, not just the model.

Then came the harder evidence, and it cut against a load-bearing assumption. The front display rides on a calendar ring. The 2021 model and most prior work assumed that ring carried 365 holes — one per day of an Egyptian solar year. But independent investigators, beginning with Chris Budiselić and colleagues, measured the surviving hole positions and argued the count was wrong. In 2024, two gravitational-wave physicists at the University of Glasgow, Graham Woan and Joseph Bayley, took those measurements and ran them through the same Bayesian inference and Markov Chain Monte Carlo tools they use on LIGO black-hole data. Their result, posted to arXiv: the ring almost certainly had 354 holes — a lunar year — not 365. They put the lunar count as roughly 229 times more probable than 360 holes, and overwhelmingly more probable than 365.

This is the Inverted World detail that the "mystery solved" coverage glides past: the front of the Mechanism may have been counting the Moon, not the Sun. And even the experts who accept the 354-hole measurement don't agree on what it means. Mike Edmunds, who chairs the Antikythera Mechanism Research Project, said there's "no obvious reason to doubt" the 354 figure — but added it is "not at all clear how it would work" as a lunar calendar or how it relates to the markings on the dial. Separately, a 2025 reconstruction in the Journal of Astronomical History and Heritage argued the zodiac ring used twelve unequal divisions to encode the Sun's varying seasonal speed — a different front-dial geometry than the clean model implies.

The skeptical-but-fair reading is that none of this makes the device alien or magical, and nobody serious claims it is. The reconstructions disagree precisely because the front gearing is missing and the surviving fragments are ambiguous — this is normal, healthy science arguing over a hard inverse problem with incomplete parts. "Solved" was always too strong a word for a model built around components that are at the bottom of the Aegean.

The unresolved question is the one the 2021 announcement obscured rather than answered: we still do not know how the front of the world's first computer was actually divided, whether it tracked a solar or lunar year, or which of several competing reconstructions — if any — matches the machine the Greeks actually built. The gears we have don't fit the model cleanly, and the gears that would settle it were never raised from the wreck.