Perhaps the most important story behind our discovery of a rare engraving earlier this week is the crucial role the Internet Archive's digitized historical book archive played in the discovery. While Gemini's visual reasoning was able to determine that our engraving was a perfect mirror-image of the original Audran, that left the most important question: what precisely was it? The credit line to Kleinschmidt was of little use, since all searches for his engraving of the chapel cupola returned his later one-sheet engraving after Picart, not his monumental 5-sheet exact replica of the original Audran engraving of Le Brun's painting. It was only after Gemini tracked down a 1912 Louvre inventory and connected its Heinecken citation back to the original text of that 1778 volume, cross-referenced it against Le Blanc's 1854 Handbook for the Collector of Prints, was it able to finally unravel our mystery and determine that our engraving was a long-forgotten exceptionally rare wanddrucke: essentially a laid tissue paper template designed to be destructively copied onto a wall or ceiling as an artist's template.
Think about that for a moment. Like a human scholar, Gemini went searching through the Internet Archive's library of centuries-old digitized historical books, identified and deciphered references to yet other works it tracked down and ultimately pieced together all of these disparate puzzle pieces spanning four different centuries (1681, 1778, 1854 and 1912) to finally uncover the true identify of our engraving. Without the Archive's digitized library centralizing all three books into a single searchable and browsable repository, where Gemini could keyword search and then visually read through each matching work, we would never have been able to unmask this work.
Most importantly, Gemini didn't simply keyword search and excerpt paragraphs from three historical books. It iteratively traced from work to work and wove all of the details it uncovered into a final origin narrative.
As agentic AI research tools like Gemini and Gemini Deep Research increasingly learn how to navigate and utilize digitized historical book archives, the ability of museum curators and scholars to identify new collection discoveries like this will grow dramatically. Who can even guess what myriad hidden treasures remain to be discovered in museum collections through this kind of AI-assisted digitized-book-powered research!