AI Successfully Reads First Complete Herculaneum Scroll, Revealing Lost Ancient Text

On June 25, 2026, a 2,000-year-old bargain was finally broken. For centuries, the carbonized library of Herculaneum—buried by the eruption of Mount Vesuvius in 79 AD—offered a cruel ultimatum: the scrolls survived, but only by becoming too fragile to open. To read one was to destroy it.[4][5]

That paradigm shifted dramatically when researchers gathered at the Biblioteca Nazionale di Napoli Vittorio Emanuele III to announce a historic milestone. For the first time in nearly two millennia, an entire Herculaneum scroll had been read from beginning to end without ever being physically touched.[1][3][4]

The scroll, cataloged as PHerc. 1667, was virtually unwrapped using a combination of phase-contrast X-ray microtomography and advanced machine learning. The resulting digital extraction revealed nearly 1.5 meters of continuous ancient Greek text across 20 columns.[2][4][5]

The achievement marks the culmination of the Vesuvius Challenge, an open-science competition launched in 2023 that mobilized researchers, engineers, and citizen scientists worldwide. By crowdsourcing the algorithmic heavy lifting, the project achieved in three years what traditional papyrology had struggled to do for centuries.[4][5]

The story of the scrolls begins in the opulent Roman town of Herculaneum, which was engulfed by a superheated avalanche of volcanic ash and pumice during the same eruption that destroyed Pompeii. Deep inside a luxury villa—thought to have belonged to Julius Caesar’s father-in-law—a vast library of papyrus scrolls was flash-fried into blackened, brittle lumps of carbon.[2][5]

When farmworkers discovered the buried villa in the 1750s, the scrolls were initially mistaken for lumps of coal or petrified wood. Once their true nature was realized, the temptation to read them proved disastrous. Early attempts to unroll the papyri involved rose water, liquid mercury, and physical peeling, which inevitably caused the fragile layers to crumble into ash.[5]

PHerc. 1667 itself bears the scars of these early efforts. Attempts to open it in the 19th century, and again in the 1980s, damaged its outer layers, leaving only a compact inner core measuring about eight centimeters high. It was this surviving core that researchers targeted for the 2026 breakthrough.[3][5]

The modern effort to read the scrolls non-destructively was pioneered by Brent Seales, a computer scientist at the University of Kentucky. Seales realized that physical unwrapping was a dead end and instead turned to high-energy physics and digital geometry to peer inside the carbonized rolls.[1][4]

The first step required unprecedented imaging. The scrolls were taken to the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, where a particle accelerator bombarded them with intense X-rays. This phase-contrast microtomography generated massive 3D datasets—up to 300 terabytes per scroll—capable of resolving the wafer-thin, densely packed layers of papyrus.[4][5]

However, seeing the layers was only half the battle. Because ancient scribes used a carbon-based ink made of soot and water, the writing was chemically identical to the carbonized papyrus it sat on. To standard X-rays, the ink was completely invisible.[1][3]

This is where artificial intelligence bridged the gap. Seales’ team discovered that the ink, when applied to the papyrus, subtly altered the microscopic topography and density of the surface. They trained machine learning models to detect these invisible patterns, essentially teaching the AI to "see" the raised texture of the ink against the charred background.[1][4]

The digital pipeline then required "virtual unwrapping." Researchers used computer vision algorithms to trace the complex, spiraling geometry of the scroll's internal layers within the 3D scan. These traced layers were then digitally flattened into a 2D sheet, onto which the AI ink-detection models were applied.[4][5]

The results from PHerc. 1667 have stunned the classical world. The recovered text is a philosophical treatise on ethics, focusing on human nature, impulse, and moral progress. It is written in ancient Greek and dates to the second century BC, making it one of the oldest scrolls in the collection.[2][3]

Crucially, the final preserved column of the scroll names Aristocreon, a nephew and disciple of Chrysippus. Chrysippus was one of the most influential thinkers of the early Stoic school, yet almost nothing of his vast written output survives today. The discovery places the text firmly within Stoic philosophy, a significant departure from the Epicurean works that dominate the rest of the Herculaneum library.[2][3]

One passage, read for the first time since the days of the Roman Empire, translates to: "We will inquire into something, but we will not grasp it, if in some way we depart from ourselves and from our own nature." For scholars, gaining access to extended, continuous arguments rather than isolated fragments is a transformative leap forward.[2][3]

The June 2026 announcement also included secondary breakthroughs. Researchers recovered more than 70 columns of text from another scroll, PHerc. 172, identifying it as "On Vices, Book 1" by the Epicurean philosopher Philodemus. A third scroll, PHerc. 139, yielded the title "On Gods, Book 8," also by Philodemus.[4][5]

These discoveries validate the unique open-science model of the Vesuvius Challenge. By releasing the high-resolution scans and open-source software to the public, founders Seales, Nat Friedman, and Daniel Gross bypassed traditional academic silos. The $1 million prize pool incentivized rapid iteration, turning a niche archaeological puzzle into a global machine-learning race.[1][4][5]

The implications of this technology extend far beyond a single villa in Italy. More than 600 unopened scrolls remain in the collections of the Biblioteca Nazionale di Napoli, the Institut de France, and Oxford’s Bodleian Library. Until now, they were considered lost to history.[2][5]

With the virtual unwrapping pipeline now proven on a complete scroll, the Vesuvius Challenge is setting its sights on scaling the process. The ultimate goal is to automate the extraction and read the entire remaining collection, a feat that could significantly increase the total corpus of literature surviving from classical antiquity.[4][5]

For centuries, the ashes of Vesuvius were thought to have destroyed the library of Herculaneum. Today, it is clear that the volcano actually preserved it, sealing the thoughts of the ancient world in a time capsule that technology has finally learned how to open.[1][5]