Seeking a Look at John Hancock

I asked Jeffrey M. Griffith, editor of the Colonial Society of Massachusetts’s new Papers of John Hancock series, about computerized handwriting transcription in part because his website for that project includes elements created with generative A.I.

At the bottom of the home page, for example, are three “John Hancock Videos.” It looks like these started with period images of Hancock—the well known John Singleton Copley painting and a couple of engravings. I presume the engravings were colorized.

Then all three images were fed through the sort of program that animates portraits so the subject appears to move their head, eyes, and mouth. These Hancocks deliver statements on different topics derived from the man’s actual writings through large-language-model A.I.

Those sorts of moving portraits were all the rage online a few years back. The early models showed us George Washington and the like swiveling their eyes and smiling. I wondered then how the results would look if we fed in a photograph of someone we’ve all seen really moving—a twentieth-century President or movie star. Would the output still be impressive, or would the uncanny valley open?

Just this week I saw a similar animated portrait incorporated into a museum exhibit. Now that the technology exists, it might become a standard way to enliven images for the public. Or it may be remembered as a hallmark of the 2020s. At least with figures as old as Hancock, no one will be fooled into thinking it’s genuine footage.

Dr. Griffith alerted me to another way he found an A.I. program valuable:

With that said, a very recent positive use of AI for research that I have come across is Harvard’s Collection Explorer. The metadata and previous digitalization efforts that they have worked on over the past decade has allowed them to create a tool that digs deep into their holdings.

I had a list of a handful of letters that I needed to find from Houghton Library that I could not previously uncover through their traditional search engine, and I was able to find them through this tool. Additionally, I found letters for Hancock that I have never seen cited before, nor did I know they existed. This nearly blew up Volume 2, but fortunately I was able to slip in a letter before the book design was completed.

To summarize, the actual readily-accessible tools to transcribe are not quite ready for wide use…, but the integration of AI searching has done great things for the discovery and acquisition of the materials for me to transcribe.

That speaks to one of the particular challenges of assembling John Hancock’s papers. While some families had the motivation and wherewithal to preserve a famous ancestor’s files, providing a big home base for a documentary project, others didn’t.

Also, I’m pleased to know other people find the Harvard library catalogue requires advanced technology to navigate.

Improving Transcriptions of Handwritten Documents

Even as I muse on the value of A.I. programming for transcribing historical handwritten documents, the Fold3 service announces:

We’ve unlocked 2.4 million pages of Revolutionary War Pension Files with a state-of-the-art, patent-pending Transcription Viewer.

For paid subscribers, the service offers multiple ways to see the scanned documents and its transcriptions at the same time—that seems to be the patent-pending part. But beneath that function are the transcriptions themselves.

Fold3 launched “Full-Text Search” of the Revolutionary War Pension Files in March. However, its staff says, “Handwriting recognition technology has really advanced since” that time, so the service now provides “Dramatically improved line-by-line transcription quality.”

The improved transcriptions were produced, says the staff, “primarily with AI with lots of human interaction.” It’s not clear what form that human interaction took—staff hours, volunteers, feedback from users, advanced programming?

Human oversight still seems to be key. When I asked Dr. Jeffrey M. Griffith of the John Hancock Papers about his impression of the latest handwriting-recognition software, he said:

I have not used AI tools to transcribe the manuscripts, as I have found that widely-accessible large scale models like Chat GPT, Claude, and Grok are not as accurate as I desire. What’s even more troubling for my project in particular, is that the same prompt can result in different outputs. Most problematically, in some instances where I have tested the systems, if the system is uncertain, it just creates contents that may not even be relevant to the particular manuscript.

Knowing how specialized AI systems can get, I am sure there are some extremely powerful (and equally expensive to use) models continue to improve reading handwriting from various decades and centuries, but I have not explored those. Admittedly, even if an AI system is to be used for an initial transcription, the content needs to be thoroughly reviewed and compared to the original text.

Griffith judges the basic transcription tools available now as “much better than those very early efforts by Clements Library at University of Michigan.” Transkribus or a similar service might today produce better results from the Gage Papers. Conversely, if “lots of human interaction” is key to continual improvement, that site might benefit from providing a clear channel for user feedback on texts.

Because the profession of documentary editing puts such value on accuracy, people working in the field would naturally be wary of A.I.-produced transcriptions until they’re much more reliable. A text with errors could mislead readers. More likely, in a time of tight budgets, an institution might decide that a text with errors is good enough and not put funds toward professional-quality transcriptions.

All that said, I do use imperfect transcriptions like those on Google Books or the Gage Papers as my starting point when I quote documents. I did that with texts produced through O.C.R. programs, and I’m doing that with texts produced through A.I.

I find those imperfect databases easier to search, read, and quote than online archives without any transcriptions, like the New York Public Library’s scans of the Samuel Adams Papers and Boston Committee of Correspondence Papers.

So for me the bottom line is the more information that’s available in an online archive, the better. Scans better than mere lists of documents. Scans and bad transcriptions better than scans alone. Good, authoritative transcriptions without scans very useful, and good transcriptions with scans the best of all.

TOMORROW: Other ways of using A.I.

Writing the Second Draft of History with Gen. Heath

Another of the Massachusetts Historical Society’s ongoing crowd-sourced transcription projects is Gen. William Heath’s manuscript diary.

When Heath wrote his memoir, published in 1798, he went back to this document as his source. That book is in day-by-day journal form.

The entries for 9 Oct 1776, 250 years ago today, show how he developed his history.

Heath’s diary entry, as newly transcribed:

Early this morning Three ships Two of 40 Guns and one Frigate, With Two, or Three Tenders Stood up the north River, They were Briskly Cannonaded from Mount Washington and Fort Constitution, They However Passed our works & Chevaux De Frise,

our Gallies Small Craft, and Two Large Ships Stood on before them, The Two Ships were Run on shore nears Philips’s mills, and Two Gallies near Dobbs’s Ferry,

The Enemy took Possession of the Two Gallies and got them off, one Boat Landed and Plundered a Store, (Stove the Cask) & set it on fire, but our People Extinguished it,

Colo. Sargent has marched for that Place, with 500 Infantry and 40 Light Horse, Capt. Horton with 2 twelve Pounders, and Capt. Crofts with a Howtzer

the Enemy also took a Schooner Load,d with Rum Sugar Wine etc, and Sunk a Sloop which had on board the machine for blowing up their Ships, a Contrast this to Sink that, which was to Blow them up —

our Division being under arms on the alarm were marched Down over the Broken Ground in front of our Post

The day’s entry in Memoirs of Major-General Heath:

Early in the morning, three ships, two of 40 guns, and one frigate, with two or three tenders, stood up the North River. They were briskly cannonaded from Fort Washington and Fort Constitution. They however passed our works and the chevaux-de-frise; the American galleys, small craft, and two large ships standing on before them.

The two ships were ran on shore near Phillips’s mills, and two of the galleys near Dobbs’ Ferry. The enemy took possession of the two galleys, and got them off. A boat landed a number of men, who plundered a store, stove the casks, and then set the store on fire, and left it. The Americans soon extinguished the fire.

Our General ordered Col. [Paul Dudley] Sargent, with 500 infantry, 40 light-horse, Capt. [Jotham] Horton of the artillery, with two 12 pounders, and Capt. [Edward] Crafts with a howitzer, to march immediately, with all possible expedition, to Dobbs’ Ferry.

The enemy took a schooner loaded with rum, sugar, wine, &c. and sunk a sloop, which had on board the machine, invented by, and under the direction of, a Mr. [David] Bushnell, intended to blow up the British ships. This machine was worked under water. It conveyed a magazine of powder, which was to be fixed under the keel of a ship, then freed from the machine, and left with clock-work going, which was to produce fire when the machine had got out of the way. Mr. Bushnell had great confidence of its success, and had made several experiments which seemed to give him countenance; but its fate was truly a contrast to its design.

Our General’s division was formed in line, with its advance, reserve, flank-guards, and artillery, all in order of battle, when they were moved down over the different grounds, which it was supposed might be the scene of action. Some of this ground was very broken, and there were many fences. These afforded frequent opportunities for the troops to break off and form; for the pioneers to open avenues, &c. and for the whole to become acquainted with every part of the ground, and the best choice of it, if suddenly called to action.

Heath made sure he appeared in his book, as “our General,” as an active participant, issuing orders and strengthening the forces under his command.

He also took the opportunity to add more detail, such as the backstory of David Bushnell’s submarine.

This entry doesn’t reveal any significant information that Heath left out of his book. But now that the whole diary is digitized and transcribed, folks can look for those details.

Visiting with Mercy Warren

Among the Massachusetts Historical Society’s ongoing crowd-sourced transcription projects is the Mercy Otis Warren Papers, including both political and family letters.

I found an example of a document from Mercy Warren showing the wrinkles of this sort of project. It’s a letter dated 16 June 1775. There’s no recipient noted on the paper, nor on the webpage or caption for the images.

The transcription begins:

With the highest pleasure I Receved the several Letters of my Dearest Friend. & now having Returned to my own Habitation I have Leasure to answer them I Returnd from a Visit to my Father [James Otis, Sr.] yesterday. the two miss Otis’s sister Hannah & Miss Otis Came away with me Designing only to accompany me to Sandwish but it being a very fine Day they Came as far as Plimouth. & after spending the afternoon & Evening & Regaling several times on the finest strawberys I have seen Left me this morning at 10 o clock. but the Beauties of Nature. the Bounties of providence nor the social Intercourse of Friends Can I taste with unalloyed Delight so Long as my Country is threatned with Desolation & Bloodshed. & my Beloved Husband not only Long abscent from me but stationed on Duty in a place where I am Greatly Apprehensive of personal Danger to him.

Eventually the letter is signed “your Mercy Warren.”

Folks who read a bunch of Warren letters will recognize that she often addressed her husband James as “my Dearest Friend.” When she wrote of “my Beloved Husband,” she was being arch and literary, not describing him to a third party. 

Jeffrey H. Richards and Sharon M. Harris recognized that intended recipient in Mercy Otis Warren: Selected Letters, which they published in 2009. Unaccountably, though, they dated this letter to 15 June.

James Warren responded on 18 June. His letter has already been transcribed and published in the Massachusetts Historical Society’s Proceedings and Collections volumes—not because it came from the man in the marriage, but because he had a lot to say about the Battle of Bunker Hill.

In his reply James also told Mercy: “I hope your Strawberries are well taken care of and that you have fine feasting on them.”

TOMORROW: The general’s diary.

Crowd-Sourcing the Documents of Revolutionary Boston

The Massachusetts Historical Society is one of the institutions that’s taken a crowd-sourced approach to transcribing significant collections that aren’t as monumental as the major Founders’ papers.

Traditionally, transcription projects of this size have been undertaken by one dedicated scholar, sometimes with a small team of students as assistants.

That method definitely has advantages: the editors become familiar with the writers’ handwritings, vocabulary, and references. They know the standards of the field. But such a team needs funding.

Scanning a collection and making those digital images available online has made it possible for an archive to recruit and train volunteers to transcribe pages and check each other’s work. Each page becomes a puzzle to solve, and a contribution to scholarship. That sort of project still needs funding, but not as much.

The resulting transcriptions are reliable in the same way Wikipedia has become a useful (albeit not infallible) source of information on many topics. Many sets of eyes on the same information means fewer errors survive. While A.I.-aided transcriptions may be faster, these are still more reliable.

One of the M.H.S.’s first projects of this sort was the diary of John Rowe, the politically malleable merchant. I’ve used that resource many times. Perhaps my favorite entry is Rowe’s comment about a Massachusetts Provincial Congress holiday that he went back and edited, reversing the political sentiment he originally expressed.

It looks like the M.H.S.’s team is nearly finished with the Ezekiel Price Papers. Price was a court official, notary, insurance broker, and all-around networker, with connections and gossip sources everywhere. I haven’t found as much drama in Price’s own life, but he’s a great source and observer on events.

The M.H.S. welcomes more volunteers interested in transcribing more documents from the eighteenth and nineteenth centuries.

The biggest drawback of these projects for me are twinges of guilt. Guilt that I haven’t yet signed up for any. Guilt that when I milk the results for Boston 1775, I follow some of my own formatting preferences (non-typewriter punctuation, paragraph breaks), breaking the rules that the volunteers so carefully follow.

But my main feeling is gratitude to all the staff and volunteers who are making those collections more available and more useable for everyone.

TOMORROW: Sampling some ongoing projects.

Speeding Up “Slow-Motion Scholarship”?

I’ve been doing test drives on the Clements Library’s machine-made transcriptions of documents from the Thomas Gage Papers, seeing how useful those texts from Transkribus are.

Archiving documents already costs money, of course. Systematic scanning of those documents and making them available online costs more.

Producing accurate transcriptions for those documents requires even more labor, with a lot of expertise and patience thrown in. That might be the most expensive stage.

The U.S. government once deemed that work important enough to fund. We can look back at precedents like the Varick Transcripts of Gen. George Washington’s letters made in the 1780s, Peter Force’s American Archives in the ante-bellum period, and the major scholarly projects of the Progressive, New Deal, and Cold War eras.

Originally based on thick printed volumes, that effort has naturally spread to computers and computer networks. Founders Online is not only hosted by the National Archives, but its constituent projects received funding through the National Historical Publications and Records Commission (N.H.P.R.C.).

But that was all before the current administration set out to slash funding for historical projects that don’t feed its leader’s ego, both within the budget process and illegally outside of it. In June the Organization of American Historians issued a report titled “The Federal Assault on History: A Record of Executive Actions,” (P.D.F. download). It says:

Despite the impressive work done with the assistance of NHPRC grant awards, the program created by Congress often finds itself eliminated in presidential budget requests. FY 2026 and 2027 were no different, with the Trump Administration attempting to zero out the funding. As in prior years, Congress recognized the value that NHPRC brings to their districts and rebuked the effort, providing $5 million in grantmaking funds, though this represents a 50 percent cut to the program from FY2025.

Funding is only one of the challenges faced by NHPRC in 2025 and now in 2026. NHPRC is convened twice a year by the Archivist of the United States, its chair, to review grant proposals, recommend awards, and approve new calls for proposals. Without an Archivist of the United States duly confirmed by the Senate, the Commission has no chair to convene it and has not been able to meet since November 2024. This little-understood effect of the firing of the Archivist of the United States in February 2025 has blocked Congress’s intention to award millions of dollars in grant funding to archives and other historical institutions throughout the country.

That’s yet another part of our national government and polity that will need repair.

To be sure, there were limitations in that approach to transcribing archives. For example, there’s always been more funding available for the papers of elite American Founders than for other sorts. (Gen. Thomas Gage came from an elite class, too, but he was on the Wrong Side.)

Those monumental projects also took decades. Back in 1997 Lingua Franca magazine published an article by Jack Hitt titled “In the Franklin Factory: Slow-Motion Scholarship in an Age of Academic Obsolescence.” It concluded with a glimpse of the Benjamin Franklin Papers editors looking ahead to the end of their work after four decades. They’re still at it.

Given that environment, it’s easy to see the appeal of artificial-intelligence software as a way to ease, speed up, and democratize the transcription of handwritten documents. If, that is, the technology actually works.

TOMORROW: Alternative approaches.

Reading Documents about the Governor and the Towns

I plucked out a couple more documents from the digitized Thomas Gage Papers to see how well the automated handwriting-recognition program Transkribus has worked there.

Here’s a bit of “Intelligence” dated 25 Feb 1775. It’s a casual note, written quickly for Gov. Gage’s files, rather than an official letter. The transcription reads:

St.
It having been reported that aquantity of Shan Outs &c is
confesting of persons
now buying up for the use of the Army a Committee, from
all the Towns round met last might to consult measures to
prevent this supply from being bro: down & effectually
hinder the people from selling any hereafter & also to him,
der any large quantity of any kind of Provisions beyond
what may be daily expended from being sent into Town

I read it as:

It having been reported that a quantity of Straw Oats &c is now buying up for the use of the Army a Committee consisting of persons from all the Towns round met last might to consult measures to prevent this supply from being brot: down & effectually hinder the people from selling any hereafter & also to hinder any large quantity of any kind of Provisions beyond what may be daily expended from being sent into Town

The phrase “consisting of persons” is an insert between lines marked by a caret, and the second “hinder” fell over a line break, so it’s less surprising the program glitched on those. The “St.” at the beginning of the automatic transcription appears to be an echo of “It”—in other words, the program transcribed those marks twice. All told, not a bad start, but still falling short.

Here’s a crowded document with bleeding through the paper: a 3 Feb 1775 petition from the selectmen of six towns urging Gov. Gage not to send troops to Marshfield. It begins:

We his Addutys loyal Subjects, Selectmen of the Sevent
town of Otimouth, Kingston, Dubord, Pembrook Hanour
& Lastrate, alenly affected with a Sence of the Encreasincy
Danger Realamities, which menace one of the most prom-
ifing country upon tenth, with Political Action, cannot
but lement, that while we are Endeavouring to preserve
peace & maintain the authority of the laws; at a perod
When the bunds of Government are relieved, by Violent
Infructions on the Chinster of the Province, our Enemies
are prutifing Every Mydim, Strutages, to Seduce the
people into act of Violance, & ontrage.

My reading of those lines:

We his Majestys loyal Subjects, Selectmen of the Several towns of Plimouth, Kingston, Duxborô, Pembrook, Hanover & Scituate, deeply affected with a Sense of the Encreasing Danger & calamities, which menace one of the most promising countrys upon Earth, with Political Exition; cannot but lament, that while we are Endeavouring to preserve peace & maintain the authority of the laws; at a period When the bonds of Government are relaxed, by Violent Infractions on the Charter of the Province, our Enemies are practising Every Insidious Strategem to Seduce the people into acts of Violance, & outrage.

Then software had a much tougher time with this one. Though I can’t blame a large language model for not recognizing the word “Exition,” which the Oxford English Dictionary says went out of use by 1820. I had to look that up myself.

Perhaps the biggest flaw in the automatic transcription is the town names. Those are important since researchers might scan for those terms. Granted, the selectmen didn’t help by using unorthodox spellings. To their credit, the cataloguers at the Clements Library listed all the towns in their standard forms in the metadata for this document, so it will come up in a local-history search.

Peter Force included this petition in his American Archives, changing the spelling, capitalization, and punctuation to conform to the standards of 1837. So historians have already had the chance to see local office-holders refer to Massachusetts as “one of the most promising countrys upon Earth.”

A quick scan of the automated transcription would probably bring the phrase “we are Endeavouring to preserve peace & maintain the authority of the laws” out of the unrecognizable gobbeldygook. That might lead one to think these selectmen were supporting Gov. Gage’s policies.

In fact, these selectmen were protesting the idea of stationing soldiers in Marshfield. They said those redcoats would provoke “the people,” already upset by the Massachusetts Government Act. That law-and-order language was a way to head off criticism and turn it against the local Loyalists asking for military protection. Another lesson in the value of reading the whole document in its historical context.

TOMORROW: The present and future of transcription.

“Even the best quality HTR will contain errors”

Yesterday I described how the Clements Library in Michigan has provided transcripts of handwritten documents in the Thomas Gage Papers.

The library website explains:

The plain text we provide for this collection was created using Transkribus, a tool that uses machine learning and artificial intelligence to transcribe handwritten documents. Handwritten Text Recognition (HTR) “reads” the page images and makes a best guess at identifying the letters and words on a page. This is an automated process and subject to error, particularly if the original handwriting is difficult to read.

Even the best quality HTR will contain errors. This means that the plain text is not a fully faithful reproduction of the scanned original document. Proper nouns such as names of individuals, groups, and places are especially subject to recognition errors. Text written at an angle or vertically on the page will likely not be recognized at all. In addition, any spelling errors, line breaks, or abbreviations in the original document will be represented verbatim.

We provide this uncorrected plain text in the interests of improving accessibility and searchability of the collection. Please keep in mind that full text search results may not be fully accurate for specific keywords and we encourage you to use multiple search strategies to find relevant materials.

I’m very grateful to the Clements Library for providing scans online. As for the transcriptions, I chose an incoming document at random to assess.

On 29 Dec 1774, New Hampshire governor John Wentworth (shown above) wrote to Gen. Gage about how his authority was crumbling. The transcription of that document currently reads in part:

It is very certain that the peoples
minds, are too generally prepared, by
factions artifices and falwhoods wickedly
disseminated by the popular Laders, for any
violences whatever. This Spirit, I observe,
has been exceedingly angmented since the
return of the Delegates, from the continental
Congress; And having broken forth into
overt Acts, in the late repeated assaults on
His Majesty's Castle William and Mary,
those persons & all they can influence are
less restraind, and more readily disposed
to pursue any violences that may be
recommended by their Party in Boston,
or in contemplation hers, among which

I have strong reasons to beleive, Reprisals
on the King's Servants mentiond by the
Congress, will be the earliest & most sudden,
His Majesty's Ships Searborough and
Cancanx are not sufficient to frustrate
such intentions, and at the same time
cover of sure the Castle Wm. and Mary,
which is situate three mites from this
Town, which wou'd receive protection
from one additional Sloop or small Rigate
of war; with an angmentation of about
Forty or Fifty Mariness to their Compliment

To quote those lines here on Boston 1775, I’d begin with this text but carefully read the scans to confirm or correct it. The result is:

It is very certain that the peoples minds, are too generally prepared, by factious artifices and falsehoods wickedly disseminated by the Popular Leaders, for any violences whatever. This Spirit, I observe, has been exceedingly augmented since the return of the Delegates, from the continental Congress; And having broken forth into overt Acts, in the late repeated assaults on His Majesty’s Castle William and Mary; those persons & all they can influence are less restrain’d, and more readily disposed to pursue any violences that may be recommended by their Party in Boston, or in contemplation here, among which I have strong reasons to beleive, Reprisals on the King’s Servants, as mention’d by the Congress, will be the earliest & most sudden;

His Majesty’s Ships Scarborough and Canceaux are not sufficient to frustrate such intentions, and at the same time cover & secure the Castle Wm. and Mary, which is situate three miles from this Town, which wou’d receive protection from one additional Sloop or small Frigate of war, with an augmentation of about Forty or Fifty Marines to their Compliment.

The A.I. transcription has some errors, but Gov. Wentworth’s main topic and message come through: Send in the Marines!

It helped that this document was clearly written. Wentworth had received an elite education at a time when handwriting was part of a gentleman’s training. He had time to compose and write out this letter; I saw only one word squeezed in as an afterthought.

As for the peculiarities of eighteenth-century writing from our perspective, Wentworth’s spelling was rather modern. He used the long s less often than many contemporaries. However, he wrote many of his lowercase e backwards to our eyes—the root of several transcription errors.

TOMORROW: What about less elegant documents?

“To Write a Good Hand Is a fine Accomplishment“

I don’t think Boston 1775 would exist without Google Books, which launched the year before I began blogging.

Being able to search for phrases and names in the vast corpus of nineteenth-century books that Google scanned has let me identify early sources, track local lore, and find rare sources.

That search depends on optical character recognition (O.C.R.) translating the printed marks into digital text.

I’ve learned how to unearth that digital text from Google Books pages for long quotations. At first this involved some tricks to maneuver to the pages I wanted. Then Google reprogrammed that site, so it became easier to capture particular text—but without any punctuation marks.

Either way, I learned to go over the O.C.R. transcription carefully to make sure it was complete and accurate. There are still a few bugs in the system. But it’s good to stay on my toes with historical sources.

Handwritten Text Recognition (H.T.R.) is the next step in this process, converting cursive writing into digital text. The University of Innsbruck’s Transkribus platform was the early leader in this field. The complexity of our writing means people started applying artificial intelligence (A.I.) software to the problem.

Indeed, some historians have said that H.T.R. is now the most advanced and reliable product of A.I. for the field, especially as it’s improved in the last couple of years. 

In September 2025, Civil War History published a study by Prof. Julie Mujic of Denison University on how ChatGPT transcribed a handwritten poem from 1864, finding many discrepancies and problems. Then in January 2026, Prof. Cameron Blevins of Colorado University Denver revisited the same text with the most recent version of Google’s Gemini 3.0, finding it produced a complete and reliable transcription.

Likewise, in November 2025 Dan Cohen, dean and professor at Northeastern, found that the Gemini Pro 3 program did a very good job transcribing some difficult letters. Even letters with cross-writing, that bane of nineteenth-century correspondence. Cohen concluded:

It allows human beings to focus their time on the important, profound work of understanding another human being, rather than staring at a curlicue to grasp if it’s an L or an I.

There’s still room for improvement, as Prof. Mark Humphries of Wilfrid Laurier University discussed in May 2026, but this application of A.I. will probably continue to get better. As with a lot of technology, one big question will be whether proprietors will be able to resist adding features that actually reduce its practicality for many users.

The transcriptions of handwritten texts that I’ve worked with the most come from the Thomas Gage Papers at the Clements Library. That site states: “The plain text we provide for this collection was created using Transkribus, a tool that uses machine learning and artificial intelligence to transcribe handwritten documents.” How do the results hold up?

TOMORROW: Reading the general’s mail.

Inventing “The ‘Preserved Clap’ carriage”

Last May, I wrote a series of postings about Dr. Preserved Clap of Hadley.

This was probably the most that anyone had written about Dr. Clap in two hundred years.

While doing that research, I ran some web searches for Clap’s name to feel sure I wasn’t missing additional sources. Google Books in particular has been hugely valuable for unearthing leads in state and local histories.

However, Google has prioritized its generative A.I. programming, putting that output at the top of search results. Here are some ways Google—once the most powerful, most useful way to search for information that humanity had yet devised—explained Preserved Clap:

The "Preserved Clap" carriage is a specific, modified light ambulance-style carriage used to transport the sick and wounded, famously used during the Revolutionary War siege of Yorktown and developed by Preserved Clap, a Continental Army artificer known for his mechanical ingenuity. [1, 2] . .

The Inventor: Preserved Clap was a Massachusetts-born craftsman with great mechanical abilities. During the Revolutionary War, he worked at the Continental artificer shops and also claimed to have discovered an improved method for boring cannons.

* The Carriage Invention: In 1780, Clap designed and constructed a specialized medical transport carriage. This carriage was later detailed by Dr. James Thacher in his military journals and drawn by an artillery officer present at the siege of Yorktown.

That response was accompanied by a thumbnail photo of the “Chariot d’Orsay” made in Paris in 1875–1885, featured in Art & Architecture Quarterly’s article about the Long Island Museum of American Art, History & Carriages, as shown above.

Dr. Preserved Clap had nothing to do with ambulances. The “carriage” he invented was a gun carriage for mounting cannon. Dr. James Thacher didn’t mention Clap in his Military Journal or his American Medical Biography. The carriage that Google showed has no link to Revolutionary America.

Of course, that text doesn’t reflect how the Google A.I. program would respond to the same query now. Even back then the service would put out different responses from one day to another according to a completely opaque process. Today the programming is using my writing from May to provide more accurate answers about Dr. Preserved Clap.

That’s because generative A.I. programs based on large-language models aren’t really designed to supply true answers to questions. They’re designed to provide plausible answers based on the statistical likelihood of how sentences flow. By releasing a few articles about Dr. Preserved Clap into the world, I altered what sentences about him are likely to say.

Back when I first ran this search, Google’s programming constructed what seemed like the most probable answer about a doctor and a “carriage,” gussied up with references to another Revolutionary doctor and a big Revolutionary event. It roped in a pretty picture of a carriage. The result was what’s come to be called a “hallucination,” a false description of the world offered with seeming confidence and authority. It’s one of the unavoidable pitfalls of this programming.

TOMORROW: Handwriting analysis.

Seeing an Unfamiliar Face

In June of 2025 I attended the Sestercentennial reenactment of the Battle of Bunker Hill in Gloucester.

While walking from one site to another I passed a series of temporary lawn signs profiling various participants in that historical event.

One showed Col. Richard Gridley, the provincial artillery commander. It was a portrait in John Singleton Copley’s style.

By then I’d been studying Gridley for over twenty years. I’d never come across a portrait. But since Gridley’s not one of the most famous or influential figures of his time, I thought it was possible that a portrait had only recently been identified or made available to the public.

I took a photo with my phone and walked on. Then I saw more of those signs, showing other men in the unmistakable style of generative artificial-intelligence (A.I.) imagery. Those pictures didn’t try to replicate photographs or period art—they were overly dramatic, flat digital images with the patina of light brown produced from the mixture of all the colors. You probably know the look.

“Wait a minute,” I thought. I went back to the Gridley sign. Only then did I notice that at the lower left corner of the picture was the tiny label “AI Generated Image.” 

Clearly the folks publicizing that big reenactment had used one of the generative A.I. programs to pump out a bunch of pictures—some in its standard style, some made to look like period portraits. And those pictures did catch the eye. For an expensive event designed to attract people not already interested in Revolutionary history, that might be a valuable function.

The risk is, of course, that those images could fool people into believing they’re real. That’s unlikely for the pictures in the obviously A.I. style, but this picture programmed to look like a Copley fooled me for a couple minutes.

So far this image hasn’t escaped the confinement of one lawn sign in Gloucester in June 2025. I certainly hope my photo taken at an angle doesn’t loose it onto the world. But I can think of plenty of examples of false or falsely attributed portraits proliferating on the internet almost as quickly as false quotations do.

In the past month I had a couple more encounters with A.I. and historical material. I received an email from an “AI agent”; I wrote about that on my Ko-fi page, one of my monthly addenda for generous supporters. I watched an online presentation using images adapted from authentic period artifacts but run through an A.I. program that changed details—apparently without the presenter realizing it. Those experiences made me think it was time to share some thoughts about the use of generative A.I. in presenting history.

TOMORROW: Hallucination nation.

Hopestill Capen’s Last Decades in Boston

Before the Revolutionary War, Hopestill Capen regularly advertised British dry goods in Boston newspapers.

After his two years in jail on suspicion of being inimical to the newly independent state, Capen kept a lower profile. He ran no ads—though he may not have had many goods to sell, either.

Nonetheless, Capen’s shop sign, the Cornfields, continued to be mentioned in the newspapers as a landmark in central Boston.

On 26 July 1786, Boston’s selectmen met to draw names of citizens in good standing to serve as jurors “for a Maritime Court” the next month. The first name chosen was Paul Revere, now with “Esq.” after his name.

The second name was Hopestill Capen, showing he had been fully reintegrated into Boston society.

Interestingly, the third name was David Spear. Capen and Spear had married sisters, Patience and Sarah Stoddard.

Hopestill and Patience Capen had had eight or nine children between 1761 and 1774. Most died in their first year, and only one, Thomas, was still alive when the British left.

Sarah Spear and Patience Capen died within a few days of each other in January 1791. Patience was 57 years old.

On 21 Nov 1795, Hopestill Capen placed this advertisement in the Columbian Centinel:

A Theft.

On Tuesday Evening last, between the hours of 6 and 10 o’clock, was stolen from the Subscriber’s house, the following PLATE, marked as follows, viz

Three table Spoons, one H.C. to P.C. made by Mr. HOMES, one d[itt]o A.B. to P.S. made by A. OLIVER; one d[itt]o. T. S. S. and 1 Porringer, marked T. S. A. Engish mode. Whoever will discover the person, so that he or she may be brought to justice, and the property restored, shall receive TEN DOLLARS reward.

I rather hope Hopestill Capen got his family silver back, including the spoon with his and his late wife’s initials on it.

On 2 Mar 1807, Hopestill Capen died at the age of 76. His funeral proceeded from the house his son Thomas had inherited, 47 Hanover Street—now the Union Oyster House.

The Capens were buried in the Copp’s Hill cemetery. Patience and Hopestill’s gravestones were still readable in the late 1800s, but they no longer appear to exist.

“To be a liege subject of the State of Massachusetts Bay”

In May 1778, the Massachusetts General Court began a new legislative year.

On 17 September, Council member John Pitts brought the lower house “a Petition of Hopestill Capen, a Prisoner in Boston Goal, praying that he may not any longer be confined.”

This was a breakthrough. For the first time, Capen was addressing the General Court as a legitimate government. Previously he had insisted that doing so would violate his religious scruples.

Once again, the legislators formed a joint committee to consider Capen’s petition, along with a similar plea for release from the shorthand expert John Hodgson.

On 19 September, the Council approved a resolution “liberting them on certain Conditions.” This time, the House fully concurred. The official resolve stated:

that the prayer of Hopestill Capen’s petition be so far granted, that he be liberated from his confinement upon his giving bond to the Treasurer of this State for the sum of Two thousand pounds, with sufficient sureties that he will subscribe and closely adhere to the following declaration by him proposed in his petition, vizt. “I, the subscriber, acknowledge myself to be a liege subject of the State of Massachusetts Bay, and hold myself obliged to assist in supporting peace and good order in this State and when called on by lawfull authority I am ready so to do.”

Back in February 1778, the legislature had stipulated this oath of allegiance:

I, A.B., do swear (or affirm as the case may be) that I will bear faith and allegance to the state of Massachusetts Bay, and will faithfully support and maintain and defend the same against George the Third, king of Great Britain, his abettors and all other enemies and opposers whatsoever, and will discover all plots and conspiraces that shall come to knowledge, against said state or any other of the United States of America. So help me God.

Hodgson was ready to take that oath, but Capen still shied from explicitly renouncing his loyalty to George III. After all, what if the tide of war turned, and the Crown prevailed? According to his faith, he would have to be a Loyalist again. But for now, Capen would acknowledge the state of Massachusetts as his sovereign. (It’s notable he used the term “subject,” not “citizen.”)

In the fall of 1778, after twenty-five months of confinement, Hopestill Capen got out of jail.

TOMORROW: The last years.

Hopestill Capen Still in the Boston Jail

We left the Boston merchant Hopestill Capen locked up in the Boston jail in mid-1777.

As a Sandemanian, Capen had stayed loyal to the British Crown—but then he’d stayed in Boston after the British military left. He refused to pledge allegiance to the colonies declaring independence in July 1776, rendering himself suspicious.

As I previously quoted, Capen tried a public plea for his release. His wife Patience petitioned the state government. His friends in the merchant community vouched for him.

On 20 June 1777, the Massachusetts Council approved a petition from merchant Robert Pierpont and others for Capen’s release. The lower house of the General Court “concurr’d as taken into a new Draught.”

A resolution is printed in Massachusetts Acts and Resolves, stipulating that Pierpont and friends would give “Give bond to the Treasurr. of This State in the sum of five hundred Pounds” that Capen would “appear before any Court in the County of Suffolk.”

But apparently the House never agreed to that wording, or those terms. Capen remained in jail.

On 24 Feb 1778, the Council took up a new petition from Patience Capen. That board named Benjamin Austin to a committee to consider it along with whatever members the House named.

On 3 March, the Council accepted that committee’s recommendation. Moses Gill (shown above) delivered the news to the House. And the House disagreed, telling the committee instead to “enquire of the Committee of Correspondence, &c. of the Town of Boston, respecting Mr. Capen.”

A week later, there was a new report. The Council again agreed with the committee, Jabez Fisher reported. The House again disagreed, and voted to set the petition aside.

The Council, made up of richer men, was apparently inclined to be more lenient, but the House, dominated by rural delegates, was still not in a forgiving mood.

The Massachusetts government had exchanged some of its Loyalist prisoners to bring captives back from New York. That wasn’t an option for Capen since he didn’t want to go over to the enemy.

And thus the legislative term ended in the spring of 1778 with Hopestill Capen still in the Boston jail.

TOMORROW: Change of heart.