I have built a virtual museum exhibition. This was not the plan.
Judith and Holofernes: The Fatal Image is now open at judith.jameshoward.us. It brings together 58 works depicting Judith, Holofernes, and the remarkably unstable meanings artists have found in their story.
This began at the Walters Art Museum in Baltimore with Trophime Bigot’s extraordinary candlelit painting of Judith beheading Holofernes. The scene emerges from darkness around a single flame. Judith works with grim concentration while her maid holds the light. Holofernes is still alive, but not for long.
I did not know then that I was looking at one image in a conversation that had already been going on for centuries.
The Book of Judith tells the story of a Jewish city facing destruction by an invading army. Its leaders have nearly exhausted their courage when Judith, a wealthy and devout widow, decides to act. She enters the enemy camp, allows Holofernes to believe he understands her intentions, waits until he is drunk, and kills him with his own sword. She returns to her people carrying his head. The invading army, deprived of its commander and apparently also its capacity for independent thought, collapses.
The story is short, brutal, and almost infinitely reusable.
Judith becomes a pious widow, civic heroine, tyrant-slayer, political allegory, model of chastity, seductress, dangerous woman, national symbol, femme fatale, feminist icon, or nightmare. Sometimes the artist shows the killing. Sometimes Judith appears after the sword has fallen, serenely displaying the head. Sometimes her maid is barely present. Sometimes the two women operate as a highly efficient team. Sometimes Holofernes dominates the scene even while dying. In other works, he has been reduced to the object Judith carries away.
The narrative remains recognizably the same while almost everything it is made to mean changes.
At first, I wanted to place the famous versions together. Caravaggio, Artemisia Gentileschi, Botticelli, Donatello, Giorgione, Klimt, and a few others would have made a perfectly respectable little exhibition.
Respectability did not survive contact with the research.
An exhibition assembled from the dozen works everyone already knows says something about the canon. It does not necessarily tell the history of the subject. Institutional fame is not a proxy for artistic importance, and search-engine visibility is not a curatorial principle. Once the search moved outward from the object tradition rather than downward through lists of famous museums, Judith began turning up everywhere.
The resulting gallery includes paintings, sculpture, drawings, prints, manuscript illumination, tapestry, photography, terracotta, ceremonial art, and decorative objects. There is a Safavid album painting made in Isfahan, a southern-Netherlandish tapestry, a carved powder flask, a nineteenth-century terracotta Judith, and a Polish Hanukkah lamp crowned by Judith holding the head of Holofernes. The lamp is the sort of object that confirms the search has finally escaped the familiar survey.
The famous works are still there. They are simply no longer permitted to pretend they constitute the whole tradition.
Each object has its own catalogue record with information about attribution, date, medium, dimensions, creation place, present location, interpretation, and sources. Disputed dates and attributions are identified as disputes. Surrogate sources are identified as surrogate sources. The site does not silently convert the best available evidence into certainty.
The exhibition also includes two complementary maps. “Made here” shows where the objects were created. “Held here” shows where they live now. The distinction matters. A Brazilian artist may have painted a work in Florence, while an Italian painting may now hang in Baltimore, Tehran, Melbourne, or São Paulo. The maps can be zoomed, panned, and used to open the works associated with each location.
This reveals two different geographies. One is the geography of artistic production. The other is the geography produced by collecting, trade, inheritance, conquest, donation, and the modern museum.
The story itself is part of the exhibition. The site includes two complete English translations of the Book of Judith, chapter crosswalks, and interpretive material connecting the text to the objects. This helps separate elements actually found in the book from later artistic and religious traditions. Judith’s unnamed maid, for instance, is often called Abra, although the name comes from later tradition rather than the biblical text. The familiar association with Hanukkah and cheese also has its own complicated history.
Of the 58 objects presently included, 43 have reusable images. The remaining 15 are deliberately presented as text-only records because suitable images are not available for reuse. Each still includes the fullest defensible record and a link to the institution, holder, or rightsholder possessing the image. The complete image-rights audit explains the status and source of every displayed or withheld image.
Omitting those works would make image licensing determine art history. That seemed unacceptable.
Instead, the exhibition explains why the image is absent and includes a tactfully pointed invitation for the holder or rightsholder to make one available. This is museum-grade shade, but it is also a serious point. Public collections cannot simultaneously lament their limited reach and prevent the public from using images of the objects they preserve.
The exhibition does not claim completeness. Its open research ledger identifies what remains missing, what could not yet be verified, and which areas of the tradition remain underrepresented. The revision history records factual and structural corrections rather than quietly pretending the first version was flawless.
My work in history and genealogy has repeatedly taught me that absence from the familiar story is not the same thing as absence from history. A life reduced to birth, marriage, and death often becomes astonishing once someone bothers to look for what happened between those dates. The same principle applies here. Fame-weighted research produces thin histories. Object-first research recovers things the familiar narrative discarded.
This exhibition was built with extensive assistance from artificial intelligence. ChatGPT assisted with collection research, source organization, writing, site engineering, data validation, and deployment. Claude was then invited, repeatedly and with considerable enthusiasm, to attack the result. It found factual errors, dead links, contradictory records, weak explanations, and several things it merely disliked. The valid criticisms were researched and corrected. A few findings were rejected when the underlying collection records supported the exhibition.
That process is disclosed as part of the exhibition because it is part of the project. Artificial intelligence made an exhibition of this breadth practical for one person to construct. It did not make skepticism, judgment, verification, or responsibility unnecessary.
The complete site is deployed from its GitHub repository, with every change automatically tested and published through GitHub Actions. The exhibition now has a permanent home, a reproducible build, and a structure that can continue growing as additional objects and better evidence emerge.
A physical exhibition of this scope would be nearly impossible. The objects are scattered among national museums, municipal collections, churches, university museums, libraries, specialist institutions, private collections, and collections on several continents. Some could travel. Many could not. Some are famous. Others have nearly disappeared from the familiar account because their holders are less prominent, their catalogues are difficult to search, or their images are unavailable.
A virtual exhibition has different rules.
It can place Rome beside Baltimore, Florence beside Tehran, and a celebrated oil painting beside a manuscript leaf, tapestry, photograph, terracotta sculpture, or Hanukkah lamp. It can connect an image to the passage that inspired it. It can place works created centuries apart immediately beside one another. It can preserve unresolved questions rather than concealing them behind a wall label.
I wanted to see what happened if we put as many Judiths as we responsibly could into the same room.
And, yes, it absolutely slays.