Are you interested in Automated/Handwritten Text Recognition of historical sources? This September 7-9, we are organizing the 2nd Exchange Meeting of SCOOP (Sources Codes of the Past) in Vienna, bringing together researchers and developers. More information: www.scoop-network.org#programme #htr #atr

SCOOP: Source Codes of the Past

A conference and exchange network for the international community working on Handwritten Text Recognition (HTR). Vienna, 7-9 September 2026.

scoop-network.org

A fun fact: A couple of months ago, I actually returned to Roam Research as my PKM, back from Obsidian. I just opened it one day and it was like before - even my weird CSS still worked. Everything since I left it in 2021 (?) is still in its place, it simply works.

There was Transkribus at the beginning and then eSriptorium, Pero, OCR4All, Teklia, Calfa Vision but it was still calm. Now, we have a new vibe-coded transcription app like every month? Unsubstantiated claims, simply using some VLM. #ocr #htr #atr

Tweeting about your minor find in 19th century.

Making again incomprehensible memes. #latin #palaeography

After one and half hour of vibecoding with Github Copilot Pro. Okay.... That became really unusable for some small experiments.

Does anybody have this book? :)

My idea for a quantitative study of the flammability of medieval manuscripts doesn’t sound so outrageous now, does it?

Göktuğ Kayaalp@igoktugk.bsky.social · last mo.

This is an assault on humanity. "AI vendors are destroying rare books to feed the chatbot Strip-mining the world of old books by Amy Castor and David Gerard" pivot-to-ai.com/2026/07/23/a...

Hey, we are happy to announce HTRMoPo Web at @inriaparisnlp.bsky.social ! htrmopo.inria.fr HTRMoPo is a HTR Model Repository that allows you to describe your Kraken Models. It has been built by Ben Kiessling, the researcher behind Kraken. What does the web version do ? Let's see:

So it seems that Google became totally unusable for searching images. I wrote "medieval scribe" into the search field and maybe 50 % of images is AI generated.

If the manuscript contains medieval logic, then a scholarship for a logician, if theology, then for a theologian, if necromancy, then for a necromancer.

Yvette Hoitink@yhoitink.bsky.social · 2mo ago

New tax: if you buy a pre-1500 manuscript, 10% of proceedings goes to fund medieval scholarship.

My night train from Brussel to Vienna is already six hours delayed and we still have to go through half of Germany. Night trains going through this country are so bad that their purpose (you save time by travelling at night) is completely obliterated.

It's live: github.com/ed-nico/vers... Local-first notes app. Your notes are plain Markdown files with wikilinks, backlinks, Bases, journal, PDF annotation, and much more. We (Claude) did an amazing job!

GitHub - ed-nico/verso_notes: Local-first notes app. Your notes are plain Markdown files with wikilinks, backlinks, Bases, journal, PDF annotation, and much more

Local-first notes app. Your notes are plain Markdown files with wikilinks, backlinks, Bases, journal, PDF annotation, and much more - ed-nico/verso_notes

github.com

One of the interesting developments at #imcleeds2026 is the number of presentations generated by Claude. I have seen already four, I think. And to be honest ... they look fine. Just that it is difficult to unsee them.

Today I learned that Chinese made facsimiles of letters already in around 3rd/4th c. CE (if I understood the speaker well). Fascinating. Do we have facsimiles from elsewhere that would be so old?

MUFI before MUFI

Digital humanities seem to have one key advantage: one has always something to present: A graph, a network, equations, a bunch of numbers. This seems to be much easier than with traditional research. But sometimes I am left to ask myself: what real knowledge are we gaining? #digitalhumanities #dh

By chance, does anybody have this article?

Will there be critical editions of old websites in the future? #dh

So I was already forced to change booking with Eurostar. Will I arrive today to Leeds? Exciting

For the first time, my ÖBB nightjet will arrive so late that I will not catch Eurostar to London (160 minutes).

On a night train to Amsterdam and then to London and then to #imcleeds . Looking forward to it :).

I am still proud of the hitchhiking signs I made back in around 2013.

Original unfinished initial vs. AI-assisted finished initial.

After I extracted the initial Q with LLM, I tried also some modifications with LLM. Would I recognize this as a modern "fake", if I saw it somewhere?

One of the uses of LLM which is tempting for me, is trying to do some small extractions and transformations, like here:

Back at creating AI slop for my presentations.

Preparing a new presentation on ATR/HTR. #atr #htr

Me when I criticize a specific Latin HTR Model output. #htr #atr #digitalhumanities

Preparing presentation slides about Latin HTR for a workshop.