The GDELT Project fed five images into Gemini 3.0 Pro on December 26th and got back a plausible explanation for handwritten annotations that had baffled experts for years. Total cost: $0.026008.
The images showed a two-page spread from a lavishly illustrated 1493 copy of the Nuremberg Chronicle, plus zoomed shots of four hand-drawn circles at the page bottom. Each circle contained Roman numerals and abbreviated Latin. Scholars who'd examined the page could tell the circles related to dates (they start with "anno" for "year"), but nobody could piece together what the previous owner was actually doing.
So what was in the circles?
Gemini's answer, which I've spent more time than I'd like verifying against the original GDELT post: the anonymous annotator was building a conversion table. The Nuremberg Chronicle gives two dates for Abraham's birth, one from the Septuagint (the Greek Old Testament) and one from the Hebrew Bible. These dates are wildly different, separated by over a thousand years in the "Anno Mundi" (Year of the World) system.
The annotator extracted both dates, then calculated the corresponding "Before Christ" years for each. Circle one: 3184 AM according to the Septuagint. Circle two: 2015 BC. Circle three: 2040 AM according to the Hebrew Bible. Circle four: 1915 BC.
The math checks out. The difference is exactly 100 years regardless of which system you use, which is suspicious in that satisfying way that suggests real intent behind it.
What Gemini actually did
This wasn't transcription. Gemini read the abbreviated Latin shorthand (things like "xpi" for "Christi" and "mdi" for "mundi"), then went back and read the printed page text to understand why someone would write those specific numbers. It identified the printed passages that contained the dates, worked out which dating system each referred to, and then explained how the annotator converted Anno Mundi to BC using medieval assumptions about when the world was created.
The model made some errors, mostly small misreadings of numerals. But the overall interpretation holds together. According to GDELT, this is the first plausible explanation for these margin annotations.
There's also a handwriting analysis, if you're into that sort of thing. Gemini thinks the annotations date from 1500-1550, written in "Cursiva Currens" with Humanist influence, probably by a German scholar or cleric. The writer used a quill with a narrow square-cut nib. Whether you find that convincing depends on how much you trust a model's paleographic expertise, which is a sentence I didn't expect to write.
The prompt was simple
Here's what GDELT sent:
Attached are two pages from the Nuremberg Chronicle. At the bottom of the page a previous owner of the book drew four circles and wrote Latin text in them. It was probably written sometime from 1493 to 1600s for context. I've also attached zoomed up images of the four circles. For each of them, transcribe the Latin and translate into English and tell me its meaning and importance. They relate to the text in the page, so use the page text to decipher and contextualize them.
No special prompting tricks. No chain of thought scaffolding. Just images and a straightforward request.
Why this matters
The Nuremberg Chronicle is famous. It's one of the earliest illustrated printed encyclopedias, with woodcuts by artists who trained the young Albrecht Dürer. Libraries have digitized copies. And yet these margin annotations sat uninterpreted because the combination of faded handwriting, abbreviated medieval Latin, Roman numeral arithmetic, and theological context was too much for individual specialists to handle quickly.
Gemini combined all of it. Vision for the handwriting. Language for the Latin. Historical knowledge for the dating systems. Arithmetic for the conversions. Each skill alone exists in various tools and humans. The synthesis is what's new.
The catch: human paleographers should verify this. The model's small errors could accumulate in less obvious ways. GDELT acknowledges this but seems pretty confident in the core interpretation.
What's next is probably more interesting. The Vatican Library and British Library have millions of digitized manuscript pages. Most have marginalia. Most of that marginalia has never been systematically analyzed because the labor cost was prohibitive. That constraint just changed.
