iiinova — where your digital self keeps evolving

Josh Hadro @hadro.bsky.social · 03/10/2026

Also fun fact: the first step in this drive is to take a ferry across the Hudson from 42nd St to Weehawken ferry terminal, which replaces the Lincoln Tunnel drive people would do now

010

Josh Hadro @hadro.bsky.social · 03/10/2026

Elder millennials love waxing nostalgic about printing out MapQuest directions, but 1917 has us all beat: people bought books of fun drives that helped them navigate an era before named highways, let alone interstate highways. Here’s the 1917 Automobile Blue Book 1917 vol 3 trip from NYC to Philly

Screenshot of the drive from NYC to Philly form automobile blue book of 1917 vol 3

172

Josh Hadro @hadro.bsky.social · 02/10/2026

I broke this story in 2011, about the first major publisher-imposed loan restrictions on ebooks (the only truly big story of my trade journalism career). It’s been 15 years, and the only major difference is that public library budgets are lower and what publishers charge libraries has gone up.

libraryjournal.com

HarperCollins Puts 26 Loan Cap on Ebook Circulations

11815

Josh Hadro @hadro.bsky.social · 02/10/2026

Link to the original story is here: www.libraryjournal.com/story/harper...

libraryjournal.com

HarperCollins Puts 26 Loan Cap on Ebook Circulations

000

Josh Hadro @hadro.bsky.social · 02/10/2026

It took me a long time to realize that the particular shape of my career wasn’t due entirely to my skills, but rather in large part because I was a technology editor at a magazine who was all in on Twitter in ~2008 and met people on there that to this day are wonderful friends and colleagues

0130

Josh Hadro @hadro.bsky.social · 01/10/2026

6.) Do this more or less cleanly at a scale of millions of pages and finally, 7.) do this at a service level cost that is at all feasible

030

Josh Hadro @hadro.bsky.social · 01/10/2026

3.) Not hallucinate plausible text into the page in a way that's almost impossible to detect 4.) Not get itself into VLM OCR psychosis when it encounters a run of dot leaders or dashes 5.) Provide line-level bounding boxes

120

Josh Hadro @hadro.bsky.social · 01/10/2026

I have yet to see a VLM that can do the following: 1.) handle the entire page at this level of complexity, on what are often degrades page scans, not just clean clips: tile.loc.gov/image-servic... 2.) Preserve the reading order, or at least not jumble it beyond recognition

The full page 4 of The New Hampshire gazette, September 23, 1800, with some complex tables with text going in many directions, and brackets, and numbers

140

Josh Hadro @hadro.bsky.social · 01/10/2026

Fun fact: I also posted this on LinkedIn, but over there it has escaped containment and I'm now getting a stream of tech industry dudes sharing their one-shot VLM output and saying all it needs is "a good prompt." 🙄

1114

Reposted by Josh Hadro

Daniel van Strien @danielvanstrien.bsky.social · 30/09/2026

Uploaded a dataset of 98,877 historical newspaper pages (1700s–1940s) to the Hub, each with its original OCR text, word boxes and confidence scores. huggingface.co/datasets/big...

Sample grid of newspapers in the dataset

24917

Josh Hadro @hadro.bsky.social · 30/09/2026

You: "OCR is a solved problem" Me:

A screenshot of a page from The New Hampshire gazette, September 23, 1800, with text going in every possible direction, and brackets, and numbers

35611

Reposted by Josh Hadro

Melanie Walsh @mellymeldubs.bsky.social · 30/09/2026

You can make flipbook GIFs 🤩 Play with it here: hadro.github.io/flipbook/#m=...

1177

Josh Hadro @hadro.bsky.social · 30/09/2026

So this is why I was asking for examples -- hope it's ok to have featured your example! bsky.app/profile/hadr...

000

Josh Hadro @hadro.bsky.social · 30/09/2026

Oh yeah, agreed -- this was more curiosity mixed with a resurfaced memory of the Internet Archive experiment, I don't think it's a viable solution except in maybe some very narrow circumstances. But I had fun trying to source the examples!

010

Josh Hadro @hadro.bsky.social · 30/09/2026

And for the sake of full disclosure and transparency, the blog post is 100% human written the old fashioned way, by me, but Claude Code generated nearly all the code, and also was surprisingly helpful in creating the screen recording in the first skeet (Claude really knows how to use FFmpeg!)

161

Josh Hadro @hadro.bsky.social · 30/09/2026

Digital libraries over the years have tried to solve this in a bunch of ways, and I go into that in the blog post above. I also tell a fun little memory lane story from during my time in NYPL Labs of how we tried to come up with generative ebook covers for a large corpus of public domain works.

131

Josh Hadro @hadro.bsky.social · 30/09/2026

The backstory is here: hadro.github.io/blog/flipbook/ It's inspired by an Internet Archive feature from ~2013 where book thumbnails cycled through pages from inside each volume. Not everyone loved the feature, and it didn't last long, but it stuck with me.

hadro.github.io

How Do You Represent a Book with a Single Image?

The first page of a digitized book is often its least interesting one. A look at how digital libraries have tried to pick a better thumbnail, and Flipbook, a small tool that revives the Internet Archi...

261

Josh Hadro @hadro.bsky.social · 30/09/2026

How do you show a whole digitized book in one thumbnail? Modern books have illustrated covers, but in research collections the cover is often a plain library binding. So I made Flipbook: hover over a book from any #IIIF source and flip through its pages. hadro.github.io/flipbook/ #GLAMsky 📚📜

49630

Josh Hadro @hadro.bsky.social · 26/09/2026

Happy to report that the 5yo has no problem answering the question, “what is it like to be a bat?” Checkmate, philosophy 101, next problem please 🦇

141

Josh Hadro @hadro.bsky.social · 26/09/2026

Do you have any favs from this collection? Putting together a playful IIIF thing and looking for some examples to use…

100

Josh Hadro @hadro.bsky.social · 25/09/2026

If you believe that libraries should be allowed to buy books and lend them out, then you also believe in the First Sale doctrine, for better or for worse

051

Josh Hadro @hadro.bsky.social · 25/09/2026

Fun fact, I am working on a fun work project that has an output sorta like this. Don’t worry, it’s a dataviz, not a national library wet lab

010

Reposted by Josh Hadro

William B. Fuckley @opinionhaver.bsky.social · 25/09/2026

Counterpart not being on the top 100 shows list is an outrage. All 12 of us Counterpart fans are outraged.

301339

Josh Hadro @hadro.bsky.social · 25/09/2026

I have ranted about this to everyone who will listen since the show ended

010

Josh Hadro @hadro.bsky.social · 25/09/2026

This is perfect, thank you!

010

Josh Hadro @hadro.bsky.social · 25/09/2026

Ha, yes, it’s widespread, I guess I’m looking for standout examples among a common practice

010

Josh Hadro @hadro.bsky.social · 25/09/2026

Ha, fortunately/unfortunately that’s _exactly_ the kind of collection I’m looking for!

110

Josh Hadro @hadro.bsky.social · 25/09/2026

Library friends (📚📜): I'm seeking examples of digitized books where the "cover image"/first image of the work poorly represents the contents — e.g. maybe there's boring cloth binding that shows as the first image, hiding something like Audubon illustrations just a bit deeper inside. Any examples?

311

Josh Hadro @hadro.bsky.social · 24/09/2026

*nods in a digital library-knowing way*

020

Reposted by Josh Hadro

Daniel van Strien @danielvanstrien.bsky.social · 23/09/2026

The Europeana Newspapers dataset on @hf.co now has an `alto` config: the raw ALTO XML for all 5.9M pages. The coordinates for every word, line and block, per-word OCR confidence and font info that the flattened text dropped are back. huggingface.co/datasets/biglam/europeana_newspapers

huggingface.co

biglam/europeana_newspapers · Datasets at Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

1164

Josh Hadro @hadro.bsky.social · 23/09/2026

What’s your location-specific delusion? I know I could not win a fight against a chimpanzee, and I doubt I could land a plane, but based on no evidence or experience whatsoever I still feel positive I could pilot an East River ferry safely to port if circumstances called.

010

Josh Hadro @hadro.bsky.social · 23/09/2026

A true galaxy brain marketing maneuver: Brooklyn Public Library adding a free ebook PSA onto the end of the Jury Duty orientation video Well played @bklynlibrary.bsky.social! Love the suggestion to visit the nearby Center for Brooklyn History as well.

Promo on a video monitor suggesting jurors should visit the Brooklyn Public Library on their lunch break

010

Reposted by Josh Hadro

Jim Casey @jimccasey1.bsky.social · 22/09/2026

We've made a lot of progress on OCR + HTR, but nothing breaks a layout parser like Cattle Text Recognition (CTR)

Screenshot of a newspaper page from 1908 with little pictures of cows and their brands.

0145

Reposted by Josh Hadro

Mike Thompson-Brusstar @mthompsonbrusstar.com · 22/09/2026

no one who studies the present believes me! the past was full of sickos with their own layout conventions for printed and handwritten material!! it is not trivial to deal with even if "all you want is the text"! which by the way, it shouldn't be!!!

3365

Josh Hadro @hadro.bsky.social · 22/09/2026

If I had a time machine, after I finished all the moral imperative work etc etc., I would go back to 1930s-1950s and figure out how to be a fly on the wall for all the conversations about microfilming library collections to try and grok how it all went down

An essentially impossible to read microfilm image of a 1913 Manhattan and Bronx city directory title page, under exposed to the point of being illegible

1110

Josh Hadro @hadro.bsky.social · 21/09/2026

And once you get into tight multi column pages of city directories, phone books, and more? With indents and ditto marks and ads that cut across columns etc? Digitized from sketchy microfilm? Yowza. Still, fun problems to work on!

A page from a 1906 Upington Brooklyn city directory with two full columns of text and ads that cut across the columns

070

Josh Hadro @hadro.bsky.social · 21/09/2026

You could OCR this page in 2013 just fine — but would the word order be useable? Even now, it’s not simple! That one ad box that cuts across two columns is a beast.

Image of a page from the 1948 national directory of morticians that presents ocr and reading order challenges

2120

Josh Hadro @hadro.bsky.social · 21/09/2026

With new VLM OCR models ushering in an OCR renaissance of sorts, it's time to once again laugh at the fact that many of the challenges for OCR with historical cultural heritage materials are actually challenges of layout recognition & segmentation, which are often harder to solve than the OCR itself

2272

Josh Hadro @hadro.bsky.social · 21/09/2026

“For anyone who needs reminding: The ability to do this is, and has always been, the promise of open data.” This could apply to basically every useful and interesting thing I’ve done in my career!

Screenshot of the post from which the quote above was sourced

162

Josh Hadro @hadro.bsky.social · 20/09/2026

😻 bestill my local news loving heart, this is a hyperlocal dream!

052

Josh Hadro @hadro.bsky.social · 18/09/2026

Spent the last few days at #ff2026 talking about neat advancements in OCR technology, among other things. Here’s a pic of 2016 me showing the time I used GNU parallels to chain together 10 computers in NYPL Labs over LAN to do mass distributed tesseract jobs. I basically created an OCR botnet.

Josh standing in front of a display screen with command line output on it showing the results of a large scale Tesseract job

160

Reposted by Josh Hadro

Paul Ford @ftrain.com · 17/09/2026

Unsolicited recommendation: The @beta.nyc Civic Tech newsletter is a really good news roundup—if you live here and work around the sector you definitely want to look into it. www.beta.nyc/category/new...

beta.nyc

Newsletter - BetaNYC

1173

Josh Hadro @hadro.bsky.social · 15/09/2026

I look like I’m delivering a sermon but really I’m talking about the nerdiest possible uses of things like travel guides and city directories

040

Josh Hadro @hadro.bsky.social · 14/09/2026

I think about that story a lot.

100