About Our Shared History
Family letters and oral histories float around many families without a real home. The ancestral family tree is useful, and I wanted to build something more with what artificial intelligence can now do.
My sister, Summer Bibbee, started me down this road. She wanted something more than a family tree to share with the family, and her vision inspired the first version of this site.
The family trees I found told me who was related to whom. They gave me less of the family history I was looking for, and they were hard to add our shared experience to. Most of the last generation of genealogy work converted documents into relational databases: someone pulled the names and dates out of each document and put them into a structure that could be searched by keyword.
As co-author of The AI Conundrum (MIT Press, 2024), I have been thinking a lot about the new capabilities AI brings. One is a different way to work with archival documents. We can keep each document whole, put it into a corpus and vectorize it, so the search understands what a passage means as well as the words it uses. Say I want to learn about crops and farming. My family's letters rarely use the words “crops” or “farming.” They talk about the price of corn and the drought on the cotton. A search of the vectorized letters finds those passages anyway, because they are close in meaning. That is a big step forward, and it got me excited about what could happen if many people joined in and added their own family letters and documents.
I was also curious about my own family: generations of mostly poor farmers, some who held enslaved people, some who fought in wars, some who worked for the railroads as America industrialized. I wanted to understand more about the context of their lives.
My first attempt, in 2024, stalled. Handwriting recognition did a poor job with the letters, and transcribing them myself or with family help was slow, painful work. By 2026 the AI systems from OpenAI and Anthropic had improved enough that I could run each letter through them independently and check whether the transcriptions converged. I could draw on their deep knowledge of history to build a narrative, and use AI agents to research further, check the facts, and mark what is documented and what is speculation. And I could build a system flexible enough that anyone can upload their own documents and have the AI connect them to a shared family graph and draft its part of the history.
From there we can build up, from many family histories to the broader story of a place and a time. We can also drill down, from a place and a time to the people who lived there.
History mostly follows the handful of people and events that shaped the world we live in today. There is real pride in finding a connection to someone remembered for doing something important. I am most interested in our shared humanity: the everyday struggles, and the slow bend of history as ordinary people lived it, people like our own family members.
If you have letters, diaries or family stories, please add them, and add your branch of the family tree. Let’s see what we can learn about each other.
Rex Briggs