In about ten weeks, without a background in software development, Thanyehténhas (Nathan Brinklow) built a suite of digital tools to support his research. He did it by telling an artificial intelligence (AI) model what he needed, testing the results, and asking for changes.

Brinklow, from Indigenous Knowledges and Perspectives and Languages, Literatures, and Cultures, is approaching the current artificial intelligence (AI) boom with both skepticism and curiosity. He set out to test whether AI could extend his capacity as a researcher.

Instead of writing code, Brinklow used natural-language prompts with a large language model (LLM), a practice popularly known as “vibe coding”. He applied the approach to create Common Prayer on Uncommon Ground, a digital humanities project that makes ten editions of the Church of England’s Book of Common Prayer, translated into Kanyen’kéha (Mohawk) and published between 1715 and 1875, more accessible for research and exploration.
 

Bringing rare Indigenous-language texts together

Brinklow has studied translations of the Book of Common Prayer since he began learning Kanyen’kéha in 2010. He values the books for the linguistic and historical knowledge they preserve, including what they reveal about the colonial relationships that shaped their production and harmed Indigenous communities.

The ten editions are rare and scattered, with no known complete collection. That scarcity became personal at the British Library, where Brinklow examined its 1787 edition that came from the library of George III. As far as Brinklow can determine, it was the only Kanyen’kéha edition published in London. Since the book never returned to Kanyèn:ke (Mohawk territory), he realized he might be the first person to open that copy and understand what it said.

“This project examines the history surrounding these books through the evolution of the text itself,” says Brinklow. “By bringing together all editions into one place, it enables close study of how orthography, language, and theology developed across time, and enables a closer examination of the shifting intersections of politics, history, spirituality, language, and translation within Mohawk-Crown relations.”
 

Building digital research tools with AI

To build Common Prayer on Uncommon Ground, Brinklow first asked the LLM for a tool to display editions side by side. That success led to a wider platform for organizing, studying, and sharing the texts.

The result combines a public-facing site with backend research tools. The site pairs digitized page images with text, allowing users to compare editions with the original English source from 1662. Users can search for words and phrases across editions, consult a dictionary with information on variations in translation and spelling, and explore an interactive timeline connecting the texts with historical people and events.

Backend tools support data management and textual analysis, while customized optical character recognition (OCR) models accelerate the conversion of page images into text.

“With vibe coding, instead of knowing how to code, I had to know what I wanted, and I learned how to evaluate whether I got it or not,” says Brinklow. “The path from ideas to working features happened quickly, and I was able to develop tools I genuinely could not do myself.”

But speed does not guarantee accuracy. Brinklow watches for hallucinations, requests changes, and knows when to start again.  

“Vibe coding with an LLM is like being the parent of a four-year-old piano savant,” he says. “When the music is beautiful, it’s wonderful. The rest of the time, it’s still just a kid that won’t listen and lies straight to your face.”

Work continues as Brinklow refines customized OCR modules that generate text from digitized pages. The LLM helps him build tools, while a person checks each page against the original, supporting accuracy and responsible data management while keeping the research in human hands.

For more information about the project, visit Common Prayer on Uncommon Ground.