https://www.digitaltrends.com/computing/ai-is-turning-its-attention-to-historical-secrets-and-already-decoding-centuries-old-papers/
Yeah, the progress here is incredible. Did I post on Transkribus? I don't recall.
In any case, interestingly, some of the hurdles are more legal/institutional now. As a scholar, your promotions and raises depend on your publications.
As a publisher, you only spend the money to print something if you can assert copyright.
But the courts have rules that AI content cannot be under copyright.
So for people doing the type of work I used to do (in fact, the project is still going), the AI decoders can speed the process up and the results are quite good, but... BUT it means our publisher can't assert copyright or it becomes complex (the text isn't copyrighted but the footnotes are).
I think this should be a non-issue since every critical edition faces the same problem - if you publish an edition of Shakespeare or the Bible, the text is in the public domain.
Anyway, this field is advancing very fast.