Cameron Blevins

This is @emollick.bsky.social "jagged frontier": Claude Code spent 30 minutes completely failing to transcribe a 4 pg. historical source but spun up a fully interactive web map based on that source in less <10 minutes.

So much for soup to nuts. Next I gave it a transcribed version of the data I had typed into a CSV file years ago. 7 minutes later, I had a fully interactive web map. I hit my rate limit before I could make any further tweaks but even this first autonomous pass is pretty stunning: +

Poor OCR'd version of mail transit table from 1882.

Womp womp. Complete failure. It kept trying to use OCR packages to extract the data rather than vision + reasoning, which was never going to work with this kind of messy tabular source. I finally shut it down after 30 min. of spinning its wheels with gibberish OCR: +

I wanted to test out the full "soup to nuts" agentic capacity of CC with one-shot prompting, so I started by just giving it the four-page PDF file and telling it to build some kind of interactive web map based on the information within it. Drumroll... 🥁 +

I had always wanted to build some kind of interactive map or network viz using this source, but this kind of coding isn't my strong suit so I gave up. The file just gathered dust on my computer for the past 12 years. +

Sample page from a table from 1882 postal guide of time of transit of mails between major US cities

"Just point Claude Code at some files and see what it can do." I wanted to test this on a historian's workflow so I used a small 4 pg. PDF from my dissertation research that never made it into my book: an 1882 table of how long it took the mail to travel between different cities. +

Thanks that’s super helpful! I’ve been thinking about how the amount of time CC saves (5x, 10x, or if you’re an especially mediocre coder like me, 100x?) crosses some threshold where without it you would never actually make the thing in the first place because of those time constraints

Really cool! I’m curious: how long do you think building that same interface without CC would have taken you? Beyond the outputs I think it’s helpful for folks to start to understand more concretely how much time is saved.

A lot happening with VMI again. For those interested to learn more about the Lost Cause at VMI, when I taught there, my cadets and I made a digital history project documenting this history in unflinching detail 🗃️ locatinglegacies.org/s/VMI/page/h...

Friends, can I ask you to spread the word that we have a THREE-YEAR postdoc in American history at Cambridge up for grabs - ANY field, but applications are due March 1 so don't delay - apply, apply, apply! networks.h-net.org/jobs/69790/u...

A new creative writing style bench and leaderboard for LLMs surgehq.ai/blog/hemingw...

AI Preparedness Guidelines for Archivists — Archives & Records Association

"AI Preparedness Guidelines for Archivists" from @lisejaillant.bsky.social & Giovanni Colavizza: www.archives.org.uk/ai-preparedn...

AI-Forced Restructuring Is Top Risk for Higher Ed (opinion) AI-forced restructuring is the biggest risk colleges face.

AI will transform what the university sells and how it sells it. That is why it is the alligator closest to the boat. I welcome the AI alligator, because mentorship paired with expert judgment is the only defensible value proposition left. www.insidehighered.com/opinion/view...

Sign me up for that panel @zoeleblanc.bsky.social! I took a similar approach to @regan008.bsky.social last year with a two-stage approach but I have a feeling teaching those fundamentals are going to get harder and harder to convince students to actually do

The Scientific Analysis of Renaissance Recipes: Proteomics, Medicine, and the Body in the Material Renaissance Abstract. Collaborations between the humanities and sciences allow for novel insights into the material world of Renaissance recipe cultures, and in partic

In “Scientific Analysis of Renaissance Recipes,” @stefanhanss.bsky.social and his research team show how scientific analysis and historical contextualization of biochemical information of early modern recipes reveals new insights into early modern medicine and material practices. #AHR

The annual BYU Redd Center funding season is here and we have INCREASED the $$$ amounts!

PLEASE share this as broadly as possible.

We accept applications from ANY discipline for projects focused on the intermountain states of AZ, CO, ID, MT, NV, NM, UT, & WY.

tinyurl.com/2026ReddAwards

I believe it! I’m guessing most patrons aren’t using all those features that mitigate fabrications (in part because of free vs paid tiers), and the default options are still pretty unreliable.

Do you mind sharing a couple of the queries here or over email? I’m curious why I’m having such a different experience from other folks when it comes to evaluating this stuff…

Thanks @tedunderwood.com! @christorical.bsky.social if you are available Jan 29 @4pm EST @jerielizabeth.bsky.social, @faisdaudau.bsky.social and I are doing a session on demystifying AI and ML for humanists with the @ncph.bsky.social. More info available here community.ncph.org/event/AIHuma...

Some of best applications I've seen for general-purpose historical research are: a) handwriting transcription (ex. generativehistory.substack.com/p/gemini-3-s...), and b) LLM interfaces for search and discovery (ex. newsletter.dancohen.org/archive/the-...).

Hallucinations are an issue but their rate has gone down dramatically in the last 12 months. Still have limitations, but they're (mostly) not doing the blatant falsified citations or complete fabrication if you're using the most advanced models + "thinking" mode (cblevins.github.io/posts/gen-ai...)

The Outward Turn. Geocoding the Expansion of Fictional Space in Russian 19th-Century Literature We examine the large-scale geospatial dynamics of Russian prose literature in the 19th century. Specifically, we analyze how the distribution of location mentions shifts from the early 19th-century romantic era to the late 19th-century realist period. We demonstrate how realist literature moves away from the historical (and often heavily mythologized) landscapes of Russia, Poland, Ukraine, and the Baltics. Instead, it increasingly focuses on the modern Imperial capital, Saint Petersburg, as well as on Western European countries and the expanding eastern and southern peripheries of Russia, reflecting the country's ongoing military and economic expansion.

Editors’ Choice: The Outward Turn. Geocoding the Expansion of Fictional Space in Russian 19th-Century Literature

CFP: Navigating Digital Humanities Careers Beyond the Ivory Tower | Debates in the Digital Humanities Transforming scholarly publications into living digital works

How does DH travel outside the academy? What impact and what careers do DH skills enable? CFP for our next excellent (and necessary) book in the DDH series, edited by Jeanelle Horcasitas, @lisaironcutter.bsky.social, and @kallewesterling.bsky.social

dhdebates.gc.cuny.edu/page/cfp-dh-...

I am so humbled and thrilled to say that my first book, Opium Slavery: Civil War Veterans and America’s First Opioid Crisis, is a finalist for the Gilder Lehrman Lincoln Prize! 🗃️ @uncpress.bsky.social www.gilderlehrman.org/book-prizes/...

Wonderful, congratulations!!

Relevant links:
- Julie Mujic, "Generative AI and Civil War Primary Sources: A Methodology Experiment": doi.org/10.1353/cwh....
- Transcript Accuracy Auditor: digitalomnium.com/apps/transcr...

So much of the substantive discussion about Gen AI and historical research/teaching is happening on Substack, blogs, and social media rather than in traditional journals. Peer-reviewed literature feels increasingly disconnected from those conversations. +

I know this isn't a new issue (and is exacerbated by the cratering reviewer rates), but it seems especially pressing for something like Gen AI. That slow deliberative pace of publication is hindering our ability to keep up with - and critique - Gen AI from a disciplinary perspective. +

I want to be clear: this has nothing to do with the quality of Mujic's research or analysis. The problem is structural. An 18+ month publication timeline is reasonable for most historical scholarship. It just doesn't work for a technology that is changing so quickly. +

A historian reading this hot-off-the-press, peer-reviewed article would conclude: Gen AI is terrible at transcribing handwritten sources. But they should be reaching the OPPOSITE conclusion: Gen AI is approaching parity with human experts for handwriting transcription. +