TL;DR
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
Software engineer Jesse Waites reports that an AI-assisted search of digitized historical records surfaced a forgotten meteorite account, three rhinoceros records and possible unrecorded volcanic eruptions. The findings are leads grounded in archival documents, not all confirmed discoveries; specialist catalogues and original scans remain central to verification.
Software engineer Jesse Waites says an AI-assisted search across digitized historical archives surfaced a forgotten meteorite report, three records of rhinoceroses and accounts that may document previously unrecorded volcanic eruptions. The work shows how automated screening can help researchers find material buried in large collections, but the records still need to be checked against original documents and existing specialist catalogues.
Waites describes building a workflow after reading historian Benjamin Breen’s account of a possible new dodo eyewitness record in Dutch East India Company materials. Breen’s post, dated October 1, 2026, described a 1615 ship journal from Mauritius in which sailors reportedly wrote that they caught tortoises and dodos. Waites set out to apply a similar method to several historical questions at once.
The project used transcriptions from GLOBALISE, which Waites says contains 4.35 million pages of Dutch East India Company records spanning the 17th century through the 1790s. He also searched digitized Dutch and American newspapers and some ship logbooks. Waites says his system converted 5.7 million passages from the Company archive into searchable representations of meaning, helping it find relevant passages despite inconsistent historical spelling and errors in text recognition.
Rather than have a large model read every search result, the workflow used a small decision model to screen passages, followed by a larger model to translate and extract dates and places from a smaller set. Waites says the AI agent then opened scans of original pages so he could compare the archival image with the transcription. He also says he checked potential findings against catalogues used by specialists. The source account does not provide enough detail to treat every lead as a confirmed addition to the historical record.
How AI Can Surface Archival Leads
The project illustrates a practical use of AI in historical research: screening enormous collections for material that a person might not find through ordinary keyword searches. Early documents often contain variable spellings, handwriting and imperfect transcriptions, making a search for a modern term an unreliable way to locate every relevant passage.
Waites says that reading the full Company archive at two minutes per page would take an individual decades under a regular work schedule. His system, by contrast, processed the collection overnight. That comparison is his estimate, not an independently measured benchmark, but it conveys why automated triage could change which questions researchers can investigate. The potential value is in directing human attention to promising sources, not replacing expert assessment.
That distinction matters because an AI-generated match is not itself historical proof. A candidate passage must be interpreted in its original language and setting, checked against the scan, and compared with prior scholarship. If the leads hold up, they could add information to specialist records; if not, the screening method may still help narrow searches, but the claims should not be counted as discoveries.
As an affiliate, we earn on qualifying purchases.
From Dodo Search to Broader Records
Waites says Breen’s work prompted him to look beyond a single animal or archive question. The Dutch East India Company operated from 1602 until its dissolution in 1799, leaving a large body of records about its ports, voyages and activities across Asia. GLOBALISE has made millions of handwritten pages available as searchable transcriptions, though text recognition and historical spelling complicate searches.
Waites asked an AI research assistant to suggest historical questions that might be answerable with online records and that could be checked against an original page. The candidates included animals described in Company documents, reports of an unidentified 1808 volcanic eruption, earthquakes mentioned in old newspapers, meteorite falls and missing ships. He says he remained involved throughout, choosing directions, evaluating leads and deciding when a result was too weak to pursue.
“I think there’s something honorable about adding to what we know.”
— Jesse Waites, describing his research goal
historical document transcription software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Which Leads Will Be Verified
The supplied report does not give enough evidence to establish the meteorite, rhinoceros and eruption leads as confirmed historical discoveries. It does not include the complete document references, full passages or specialist responses needed for an independent assessment here. Waites says he compared potential findings with existing catalogues, but the extent and outcome of those checks are not specified in the supplied material.
It is also unclear which candidates have been accepted by historians or relevant scientific specialists, whether the records change established chronologies, or whether all the transcription readings match their original scans. The reported dodo account is described as Breen’s finding; it should not be conflated with Waites’s separate set of leads or treated as independently verified on the basis of this account alone.
As an affiliate, we earn on qualifying purchases.
Original Documents and Specialist Review
The next step is document-level scrutiny: researchers can follow the archive references, inspect the scans, confirm translations and compare each lead with published catalogues and scholarship. That process will determine which records qualify as new evidence and which are false matches, transcription mistakes or material already known under a different description.
Waites’s report sets out a method and candidate findings, but the supplied material does not identify a publication schedule for formal specialist review or further results. Readers should treat the claims as reported leads until the underlying documents and their interpretation have been assessed.
digitized historical records database
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What did the AI-assisted archive search find?
Waites reports finding a forgotten meteorite report, three rhinoceros records and accounts that may relate to unrecorded volcanic eruptions. The supplied account does not establish that every lead has been confirmed by specialists.
Which archive did the project search?
Waites says the search included 4.35 million pages of Dutch East India Company material transcribed by GLOBALISE, along with digitized Dutch and American newspapers and some ship logbooks.
How did the system handle old spelling and handwriting?
Waites says it searched passages by their meaning as well as their wording, then used AI models to screen results and extract information. He says he checked promising passages against scans of the original documents.
Are the reported discoveries confirmed?
Not on the evidence provided here. Waites says he checked potential findings against specialist catalogues, but the supplied material does not include enough documentation or expert responses to confirm each lead independently.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
