A Smithsonian-led project called Revolution Crossroads has started using artificial intelligence to stitch together scattered artifacts, documents, and newspaper records from the American Revolution, surfacing everyday people whose stories had slipped through the cracks of history. The effort began applying AI models in spring 2026 as the United States marks its 250th anniversary of independence, and it has already identified roughly 10,000 objects centered on people who lived between 1770 and 1810.
One early find involved a silversmith who advertised a creamer around 1774 in a notice that AI tools matched to a nearly identical creamer still held by the Smithsonian's National Museum of American History, according to ClickOnDetroit | WDIV Local 4, which cited reporting from the Associated Press. The Smithsonian has been digitizing objects specifically to make them more accessible to AI systems, and it is making Revolution Crossroads records available for online public search, allowing the public and scholars to search collections and connect records across the Smithsonian and partner organizations.
The scale of the underlying data is considerable. The project released an open-source dataset on Hugging Face comprising more than 77,000 page-level historic newspaper records spanning 1770 to 1810, totaling 247 gigabytes, according to Hugging Face. That open-data approach means outside developers and historians can build their own tools on top of the archive rather than waiting on the Smithsonian alone.
Pension Files and Library Records Fill in the Gaps
The Library of Congress is also involved in Revolution Crossroads, and its director of digital strategy, Natalie Buda Smith, said the collaboration can help researchers determine whether to begin their search at the Smithsonian or at the Library of Congress. Beyond those two institutions, the initiative pulls in extensive Revolutionary War veteran pension files from the National Archives and Records Administration to cross-reference historical figures, according to a YouTube interview cited in the research record. Pension files often contain detailed personal narratives that help connect physical artifacts to military and genealogical records.
The Library of Congress has paired that automated work with old-fashioned human effort. Its “By the People” volunteer crowdsourcing program reached more than 1 million completed transcriptions of handwritten historical documents by March 2026, the agency reported. Volunteers transcribe handwritten historical documents through the program.
Human Historians Still Have the Final Say
Becky Kobberod, the Smithsonian's first leader devoted to a pan-institutional digital strategy, said the project could add to the public record about people whose lives have faded over the centuries. The institution, founded in 1846, now spans 21 museums and galleries, including the National Zoo. Kobberod said the Smithsonian does not use current frontier AI models that have drawn scrutiny in recent weeks, and that the institution instead relies on human oversight throughout its AI research, with trained historians reviewing the data AI tools surface.
“AI cannot recognize some historical nuances that trained historians can,” Kobberod said, per the ClickOnDetroit report, even as she argued the Smithsonian must embrace AI's possibilities. Smithsonian data scientist Mike Trizna worked directly on Revolution Crossroads, and the project was photographed in Washington in late September, around the same time the Smithsonian displayed the first printed publication of the Declaration of Independence. The research record also notes a Delaware Gazette edition dated July 8, 1809, among the historical newspapers folded into the project's archive.
Governance Rules Guide the Smithsonian's AI Use
To manage artificial intelligence across its museums and research facilities, the Smithsonian established an internal AI Community of Practice co-chaired by Under Secretary for Science and Research Ellen Stofan and Chief Information Officer Deron Burba, according to testimony submitted to the U.S. Senate Committee on Rules and Administration. That governance structure is meant to keep AI tools tested for reliability while preserving human scholarly oversight over what gets published.
Revolution Crossroads is part of a larger digital push tied to the Semiquincentennial. The Smithsonian announced in June 2026 the launch of “MySmithsonian,” an online platform meant to deliver personalized museum content to audiences who cannot travel to Washington, D.C., according to GovCIO Media & Research. The Library of Congress has also begun testing a synthetic DNA data storage pilot with the University of Washington, aimed at preserving digitized national collections far beyond the next 250 years.
Politics Shadow the Project's Rollout
The technology push is unfolding against a backdrop of political tension over how federal institutions portray the nation's founding. President Donald Trump has questioned some of the Smithsonian's portrayals of history, a dynamic that echoed through a separate controversy at the Smithsonian's George Gustav Heye Center in New York, where an exhibition's characterization of George Washington as “Town Destroyer” sparked federal debate, as Hoodline reported last month. That dispute underscored a broader national debate over how federal institutions represent Founding Fathers and Native American history ahead of the 250th anniversary.
AI policy itself remains a live political fight in Washington. AI industry leaders have met with Trump and signed a voluntary accord for reviews of their systems, even as Democrats and some Republicans continue calling for stronger AI regulation. The broader 2026 commemorations trace back to the Semiquincentennial Commission Act of 2016, which established a federal commission tasked with orchestrating nationwide commemorations leading up to July 4, 2026 — a timeline Hoodline has traced through America250 events nationwide.
For now, the Smithsonian's bet is that AI can act less like a replacement for historians and more like a high-speed filter, flagging connections between newspaper ads, silversmith receipts, and military pension claims that a human researcher might otherwise never cross-reference. Whether that filter changes how the public understands the Revolution's lesser-known figures will depend, Kobberod and her colleagues suggest, on keeping trained historians firmly in the loop.
