The Last Supper Just Got AI Analysis — What’s Hidden in the Paint Changes Everything
The Last Supper Code: What Leonardo Da Vinci Buried Beneath 500 Years of Paint
In the refectory of the convent of Santa Maria delle Grazie in Milan, on a wall measuring approximately fifteen feet by twenty-nine feet, sits one of the most analysed objects in the history of Western art.
Leonardo da Vinci painted it from 1495 to 1498. He did not use the traditional fresco technique. He used tempera and oil applied to dry plaster—a method he developed himself, which gave him more time to revise his work than wet fresco allowed, but which also caused the painting to begin deteriorating within twenty years of completion.
For the next five hundred years, the painting endured:
Through botched restorations.
Through Napoleonic occupation, when the refectory was used as a stable.
Through a 1943 Allied bombing that destroyed most of the surrounding building and left the painted wall standing alone, miraculously intact behind sandbags.
Through six centuries of varnish, retouching, overpainting, and the slow chemical decay of the original pigment layer.
In 1978, an Italian conservator named Pinin Brambilla Barcilon began what would become the longest single painting restoration project in modern art history. It took her twenty-one years.
When she finished in 1999, she had stripped away nearly five centuries of accumulated overpainting and revealed for the first time since the Renaissance what was actually underneath.
What was actually underneath is the part of the story most coverage has never told.
Part One: What the Restoration Found
The Brambilla restoration confirmed something Italian art historians had suspected for decades. The painting was not a single image. It was a layered composition with embedded elements—geometric, musical, and architectural—that had been deliberately encoded by Leonardo into the underdrawing and the early paint layers, and that centuries of subsequent restoration had progressively obscured.
In 2007, an Italian musician and amateur researcher named Giovanni Maria Pala published a book titled La Musica Celata (“The Hidden Music”). He claimed that the positions of the bread loaves and the hands of the apostles, when read as notes on a musical staff, produced a forty-second composition that played coherently as a Gregorian-style hymn.
That same year, an Italian computer technician named Slavisa Pesci took a digital image of the painting, mirrored it horizontally, and overlaid the mirrored copy on the original. What emerged from the overlay was not subtle:
Additional figures appeared at the edges of the table that were not visible in the painting alone.
A face appeared in the upper architecture that had no counterpart in the surface composition.
Both findings were quietly dismissed by mainstream art historians. Both findings have now—with the application of modern computer vision and AI-driven multispectral analysis—been reinforced by something the original 2007 researchers did not have access to.
This raises three questions:
What is the AI actually identifying in the cumulative imaging data set?
Why have previous attempts to expose Leonardo’s hidden compositional layer been suppressed by the institutions that own the painting?
What specific structural elements—geometric, musical, theological—are now appearing in the digitally stripped image that the surface painting was deliberately built to conceal?
These are the questions that one twenty-eight-foot wall in a Milan convent has now placed at the centre of a five-century-old mystery. The field of Renaissance art history has been quietly waiting for the technology to finally answer.
The painting is what Leonardo painted. What is underneath the painting is what Leonardo encoded. And what the AI has just begun to read is the part of the story Leonardo himself—by every measure of the cumulative scholarly evidence—expected someone to eventually find.
Part Two: The Commission and the Method
To understand what Leonardo actually embedded in The Last Supper—and why the technology to decode it has only recently become available—you first have to understand who commissioned the painting, why it was placed where it was placed, and what Leonardo’s working method actually involved.
Ludovico Sforza was the Duke of Milan. He ruled from 1494 to 1499 during one of the most turbulent periods in Italian Renaissance history. Milan was wealthy, powerful, and positioned at the centre of European politics. The Sforza family had risen to power through a combination of military force, political manoeuvring, and strategic patronage of the arts.
Ludovico understood that great art served political purposes. He commissioned Leonardo to paint The Last Supper in the refectory of Santa Maria delle Grazie—the church attached to a Dominican convent that Ludovico intended as the Sforza family mausoleum.
The painting would occupy the north wall of the dining hall where the monks took their meals, transforming their daily bread into a perpetual participation in Christ’s final meal with his disciples. The commission was theological, political, and personal.
Leonardo accepted. He worked on the painting for approximately three years, from 1495 to 1498.
His method was unusual. Traditional fresco technique required artists to work quickly. Pigment was applied to wet plaster, which absorbed the colour as it dried. The chemical bond between pigment and plaster created a durable surface that could last for centuries. The technique demanded speed. The artist had to complete each section before the plaster dried, typically within a single day.
Leonardo did not work quickly. He was known for contemplating his compositions for extended periods, sometimes standing before a painting for hours without applying a single brushstroke. He revised constantly. He experimented. He changed his mind.
Wet fresco did not accommodate his process. So he invented something else.
Leonardo applied his pigments to dry plaster using a mixture of tempera and oil—tempera grassa, as it came to be called. The technique gave him the flexibility he wanted. He could work slowly. He could revise. He could layer and adjust and reconsider.
The technique also meant the painting would not last.
Within twenty years of completion, The Last Supper was already deteriorating. The pigment layer was not chemically bonded to the plaster. It sat on top of the plaster, vulnerable to moisture, temperature fluctuation, and the slow decay of the organic binding medium. By 1556, the art historian Giorgio Vasari described the painting as “a muddle of blots.”
The deterioration continued for centuries.
Part Three: Five Centuries of Intervention
Restorers attempted to save the painting. They applied varnishes to seal the surface. They retouched areas where the original paint had fallen away. They overpainted entire sections, replacing what Leonardo had made with what they thought Leonardo had intended.
Each restoration added another layer. Each layer obscured more of the original. By the twentieth century, scholars estimated that less than 20% of Leonardo’s original paint survived on the wall. The rest was the accumulated work of five centuries of well-meaning intervention.
The painting that visitors saw was not The Last Supper Leonardo had made. It was a palimpsest. Leonardo’s original buried beneath centuries of subsequent interpretation.
The 1943 Bombing
The bombing should have destroyed it.
On the night of 15–16 August 1943, Allied bombers struck Milan. The bombs hit Santa Maria delle Grazie directly. The roof of the refectory was blown away. The walls collapsed. The building was reduced to rubble.
But someone had anticipated the attack. Before the bombing, workers had erected scaffolding around the north wall and covered it with sandbags. The protection was improvised, desperate, and barely adequate.
It was enough. When the dust settled, the north wall was still standing. The Last Supper, behind its protective barrier, had survived.
The image of that wall standing alone amid the destruction—the painting intact while everything around it lay in ruins—became one of the iconic photographs of the Italian campaign. The painting had endured.
But what had endured was not Leonardo’s painting. It was the accumulated overpainting of five centuries, with Leonardo’s original work buried somewhere beneath.
Part Four: Brambilla’s Twenty-One Years
Pinin Brambilla Barcilon began her restoration in 1978. She was an Italian conservator with decades of experience restoring Renaissance paintings. She understood the challenges she would face. She understood that the work would take years.
She did not understand at the outset how many years it would actually require.
Twenty-one years. From 1978 to 1999, Brambilla worked on The Last Supper with a patience that bordered on obsession. Her tools were simple: cotton swabs, solvents, magnification, and an intimate knowledge of Renaissance painting technique. Her method was painstaking: removing the overpainting one tiny section at a time, revealing whatever lay beneath, and documenting everything.
What she found changed the understanding of the painting.
Approximately 20% of Leonardo’s original pigment layer survived.
That 20% had been almost completely obscured by subsequent restorations.
The colours were different than what the overpainting showed.
The expressions were different.
The details were different.
And the structure was different.
Brambilla documented compositional elements in the underdrawing and early paint layers that did not match the surface composition. Lines that served no obvious pictorial purpose. Geometric relationships that appeared deliberate but did not correspond to the visible image.
A mathematical framework underlay the composition—a framework that centuries of overpainting had progressively buried.
Her findings were published in 2001 in collaboration with the Italian art historian Pietro C. Marani in a book titled Leonardo: The Last Supper. The book documented what Brambilla had found. It did not explain what the findings meant.
The compositional anomalies sat in the scholarly record—documented but not interpreted.
And then Giovanni Maria Pala read them.
Part Five: The Hidden Music
Pala was not an art historian. He was an Italian musician and computer technician—a man with training in musical theory and a fascination with hidden patterns. He had studied Leonardo’s notebooks, which contain extensive evidence of the artist’s interest in mathematical proportion, geometric encoding, and the relationships between visual and musical composition.
Leonardo was not merely a painter. He was a polymath: a scientist, engineer, anatomist, architect, and musician. His notebooks—including the famous Codex Leicester and Codex Atlanticus—demonstrate a mind that saw connections across every domain of human knowledge.
He designed flying machines.
He studied the flow of water.
He dissected corpses to understand human anatomy.
He investigated the mathematical principles underlying musical harmony.
The notebooks also demonstrate that Leonardo encoded information deliberately. His famous mirror writing—scrittura speculare—is the most obvious example. Leonardo wrote most of his notebooks in reverse so that the text could only be read when reflected in a mirror.
The reason for this practice has been debated for centuries. Some scholars argue it was simply a consequence of his left-handedness. Others argue it was deliberate concealment—a way of hiding his ideas from casual observers.
The pattern of deliberate encoding extends beyond the mirror writing. Leonardo embedded mathematical relationships in his compositions. He used geometric principles to structure his paintings in ways that were not immediately apparent to viewers. He built layers of meaning into his work—layers that required specific knowledge to decode.
Pala believed The Last Supper contained one of these hidden layers.
His theory was musical. The twelve apostles in The Last Supper are arranged in four groups of three, with Christ at the centre. This compositional structure is well documented in art historical analysis. It creates visual rhythm and balance, organising the complex scene into manageable units.
Pala noticed that the groupings could be read as musical notation. The hands of the apostles—their positions, their gestures, their relationships to the table and to each other—could be mapped onto a musical staff. The loaves of bread on the table corresponded to specific notes.
The entire composition, when read from right to left in the manner Leonardo used for his mirror writing, produced a coherent musical sequence. The result was a forty-second composition—a Gregorian-style hymn, playable on period instruments, harmonically coherent.
Pala published his findings in 2007 in a book titled La Musica Celata (“The Hidden Music”). The book included the musical notation he had derived from the painting and recordings of the composition being performed.
The mainstream art historical response was dismissive. Critics argued that Pala had imposed a pattern that was not actually present. That the musical reading was a projection rather than a discovery. The positions of hands and bread could be read in many ways, they said. Finding a musical sequence in them was a form of pareidolia—like seeing faces in clouds.
Pala’s findings were not refuted. They were set aside.
Part Six: The Mirror Overlay
The same year, Slavisa Pesci conducted a different experiment.
Pesci was also an Italian computer technician—not an art historian, not an academic, not a credentialed expert in Renaissance painting. He was a man with digital image processing skills and a hypothesis.
His hypothesis was based on Leonardo’s mirror writing. If Leonardo encoded his text in mirror reverse, perhaps he encoded his images the same way.
Pesci took a high-resolution digital image of The Last Supper. He mirrored it horizontally, creating a reversed copy. Then he overlaid the mirrored copy on the original, adjusting the transparency so that both images were visible simultaneously.
What emerged was extraordinary.
The composite image showed elements that were not visible in either image alone.
At the edges of the table, figures appeared that had no counterpart in the surface painting.
Between Christ’s hands, a vessel emerged—a chalice or cup that was not painted, but that appeared in the overlay as a clear geometric form.
In the architectural elements above the scene, a face materialised that corresponded to no figure in the visible composition.
Pesci published his findings online and in Italian media. The mainstream response was the same as the response to Pala: dismissal.
Critics argued that overlaying any complex image on its mirror would produce artefacts. That the appearance of additional figures and objects was a consequence of the technique rather than evidence of deliberate encoding.
The criticism was reasonable, but it did not address the specificity of what Pesci found. The figures that emerged in the overlay were not random artefacts. They appeared in compositionally significant positions:
The chalice between Christ’s hands corresponded to the theological significance of The Last Supper—the institution of the Eucharist, the cup of the new covenant.
The face in the architecture appeared at a position that corresponded to the painting’s vanishing point.
The vanishing point is one of the documented mathematical features of The Last Supper. The architectural perspective lines in the painting—the ceiling beams, the wall panels, the tapestries—all converge at a single point. That point is precisely located at Christ’s right temple.
This is not accident. It is deliberate mathematical construction. Leonardo positioned Christ at the geometric centre of the composition and made his head the focal point of the entire architectural framework. Every line in the painting points to Christ. The mathematics enforces the theology.
The face that Pesci identified in his mirror overlay appeared at the same geometric centre. A face that was not painted. A face that emerged only when the encoding was reversed.
Part Seven: The Astrological Reading
Sabrina Sforza Galitzia added another layer in 2010.
Galitzia was a researcher who had served as a curator at the Vatican Library. Her credentials were more traditional than Pala’s or Pesci’s. Her claims were equally extraordinary.
Galitzia argued that The Last Supper encoded an astrological and zodiacal cycle. The twelve apostles corresponded to the twelve signs of the zodiac. Their arrangement around the table mapped to the celestial calendar. The composition was not merely a depiction of a biblical scene. It was a cosmological diagram encoding the relationship between the earthly and the celestial.
Her findings were published in Italian media and academic venues. The mainstream response was familiar: dismissal.
But the pattern of dismissal itself had become notable. Three independent researchers working from different methodologies had identified hidden compositional elements in The Last Supper:
One found music.
One found mirror-encoded figures.
One found astrological symbolism.
All three were dismissed. All three had documented their findings in detail. All three based their work on Leonardo’s demonstrated interest in encoding: his mirror writing, his mathematical compositions, and his integration of multiple symbolic systems into single works.
The dismissal did not engage with the specific evidence. It rejected the premise that Leonardo would have encoded anything at all.
But Leonardo did encode things. His notebooks prove it. His mirror writing proves it. His mathematical compositions prove it.
The question is not whether Leonardo was capable of embedding hidden layers in his work. The question is what those hidden layers contain.
Part Eight: The Disembodied Hand
The disembodied hand is the anomaly that mainstream art history has never satisfactorily explained.
In the composition of The Last Supper, the Apostle Peter is shown leaning toward John, who sits immediately to Christ’s right. Peter’s left hand rests on John’s shoulder, and his right hand is behind his back.
But there is another hand at Peter’s neck.
This hand appears to emerge from nowhere. It holds what appears to be a knife or dagger. It is positioned in a way that suggests threat—the blade pointed toward Peter’s throat or chest.
The hand does not clearly belong to any figure in the composition. Some art historians have argued that it belongs to Peter himself—that Leonardo depicted Peter with his right hand twisted behind his back in an anatomically improbable position. Others have argued that it belongs to an unseen figure, partially obscured by the crowded composition.
Neither explanation is fully satisfactory. The hand remains an anomaly—a compositional element that does not fit the surface narrative of the painting.
In the mirror overlay that Pesci produced, the hand takes on different significance. When the composition is reversed and overlaid, the relationship between the hand and the surrounding figures changes. The knife points in a different direction. The threat is redirected. The meaning shifts.
The question of what Leonardo intended with the disembodied hand has never been resolved. The question of what the hand means in the encoded composition has barely been asked.
Part Nine: Multispectral Imaging and the Missing Data
Modern multispectral imaging has transformed the analysis of Renaissance paintings. The technique involves photographing artworks under multiple wavelengths of light: visible, infrared, ultraviolet, and X-ray. Different wavelengths penetrate to different depths.
Some reveal the surface paint layer.
Others reveal underdrawings.
Others reveal layers that were painted over—corrections the artist made, earlier versions of the composition that were subsequently changed.
The technique has been applied to major works throughout the world. Pascal Cotte and Lumiere Technology conducted multispectral analysis of the Mona Lisa, published in 2015. Their work revealed multiple underlying portrait layers—evidence that Leonardo revised the composition substantially, and that the painting visible today is the final version of a work that went through multiple earlier stages.
The Vatican has applied multispectral imaging to works in the Sistine Chapel and other collections. The Louvre, the National Gallery London, and major museums worldwide have incorporated the technology into their conservation and research programmes.
The technique has not been fully applied to The Last Supper.
Partial imaging has been conducted. Fragments of the underdrawing have been documented. But the comprehensive multispectral survey that modern technology makes possible—the kind of analysis that revealed the Mona Lisa‘s hidden layers—has not been completed. Or, if completed, has not been publicly released.
The institutional control of access to The Last Supper is extraordinary.
The painting is owned by the Italian state and administered through a complex arrangement involving the Dominican order, the cultural ministry, and various supervisory bodies. Access is tightly controlled. Visitors are admitted in small groups for strictly limited viewing periods. Researchers require extensive permissions.
The digital imaging that has been conducted is not freely available. The high-resolution scans that would allow independent researchers to conduct the kind of analysis Pala and Pesci attempted are not in the public domain.
The institutions that control the painting also control the data.
Part Ten: AI and What Comes Next
AI-driven computer vision has opened new possibilities. Machine learning algorithms can identify patterns in complex images that human observers miss. They can analyse geometric relationships across entire compositions. They can detect regularities in colour, form, and structure that suggest deliberate encoding rather than random patterns.
The application of these tools to Renaissance art is still in its early stages. But preliminary results suggest that Pala, Pesci, and Galitzia may have been identifying real patterns rather than projecting imaginary ones.
The musical structure Pala identified can be tested mathematically. The positions of hands and bread loaves can be mapped precisely. The resulting notation can be analysed for harmonic coherence. If the pattern is random, the analysis should show randomness. If the pattern is deliberate, the analysis should show structure. Initial computational analysis suggests structure.
The mirror overlay technique Pesci used can be refined with modern image processing. The artefacts that sceptics attributed to the technique can be filtered. The elements that remain can be analysed for compositional significance. Initial computational analysis suggests significance.
The astrological mapping Galitzia proposed can be tested against the documented positions of celestial bodies during the period Leonardo was working. The correspondences can be evaluated statistically. Initial analysis suggests correspondence beyond chance.
None of this is conclusive. All of it is suggestive.
The technology that could resolve the questions definitively—comprehensive multispectral imaging analysed with modern AI tools—has not been applied. Or if applied, the results have not been released.
The silence is itself meaningful.
Part Eleven: The Pattern of Dismissal
Carlo Pedretti was one of the most prolific Leonardo scholars of the twentieth century. His work on Leonardo’s compositional methods documented the artist’s systematic use of mathematical proportion, geometric structure, and symbolic encoding.
Martin Kemp, the Oxford art historian, has written extensively on Leonardo’s integration of scientific and artistic thinking—the ways in which Leonardo’s paintings embody the same investigative intelligence that drove his scientific work.
Both scholars have acknowledged that Leonardo worked in layers. That the visible surface of his paintings is not the complete content. That underdrawings, pentimenti, and structural elements carry meaning that the surface image does not fully reveal.
Yet neither scholar has fully engaged with the specific claims of Pala, Pesci, or Galitzia. The claims have been dismissed without being addressed. The evidence has been set aside without being refuted.
The pattern is familiar. It is the same pattern that has characterised institutional response to anomalous findings throughout the history of scholarship. When evidence challenges accepted narratives, the evidence is not evaluated. It is dismissed.
