Yeah, these notes need to be just openly published. Then I could ask my own Claude for an explanation. (Not /s)
Or just share the Claude sessions as Tao did his.
You could likely "unlock" the manuscript yourself with a competent computer vision algorithm. It's nothing that fancy.
Have you seen the manuscripts? Take a look at slide 10 of https://aitp-conference.org/2026/slides/JvP.pdf [1]. On slide 12 you can at least make out a few words (even before reading the English translation I could clearly read "Jede" and "Systems ist berechenbar"), but any part of the shorthand is completely unrecognizable. Also take a look at slide 6 that shows a bunch of word transcriptions in the shorthand. If I understood the article correctly, you have to intimately know the math involved to have any chance of decoding the content, as there is a lot of interpretation involved.
If one could wish for a best case publishing, I would like to have the page scans side by side with the German reading side by side with the English translation with running commentary by the research group.
[1] via https://news.ycombinator.com/item?id=50487434 , thanks empath75
a computer can't see the letter "c" or any letter. You train it based on data and pixels otherwise it is "completely un-recongizable." What is so special about these squiggles? That it is a shorthand? Are words in other languages not "completely unrecognizable" to you?
loading story #50491845
[dead]
That would be the future of crowdsourcing - by submitting the same prompt as 1000 other people, you're basically donating your tokens to the cause of maybe spotting a valuable insight the others didn't get because of random output sampling :)