(29-01-2026, 09:07 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view. (29-01-2026, 07:57 PM)DG97EEB Wrote: You are not allowed to view links. Register or Login to view.I'm 99.9% sure it's gibberish.. the problem is it has grammar and structure and weird properties... I reckon Timm-Schinner probably came closest..
But there are many arguments against the "gibberish" theory. Including all the statistical features that are compatible with natural languages, and unlikely to occur in text generated by any plausible gibberish generation method. (No, Thorsten and Timm did not come close, sorry...)
Whereas the only argument for "gibberish" seems to be "we have tried to decipher it, assuming it is an encrypted European language, and we have failed so far". Which has a logical problem...
All the best, --stolfi
Yes, and I agree absence of evidence is not evidence for absence. It's a personal belief, and like all of us, we can be persuaded by evidence, I hope. But from the statistical analysis I've run, it's at least a 4 layer system, which cannot be explained by any cipher and doesn't as you say conform to any European languages, Hebrew or Arabic in any of its forms. The challenge is that at a certain point you're in to special pleading. It looks like a zibaldone/hausbuch from medieval Europe or Southern Germany and seems to contain influences from there in terms of iconography and text (per my post on LJS51 symbols appearing in Italy in the same time period).. so for me the logical conclusion is the pictures have value but the text doesn't. The only remaining question is then the scale of the thing. If it's pedagogical z why bother making 240 folios of the thing, unless that was the point... But like all of us, I seek the truth and will be persuaded as we get more evidence
..
(29-01-2026, 09:36 PM)DG97EEB Wrote: You are not allowed to view links. Register or Login to view.so for me the logical conclusion is the pictures have value but the text doesn't.
I believe exactly the opposite: the pictures are mostly fancy decoration badly copied by the Scribe from random books (each day we find one more), while the meaningful contents is the text.
For one thing, if the pictures are what matters, what would be the point of the Starred Parags section? It is not gibberish; the natural-language-ish features are as strong there as elsewhere, and several details say to me that the Author cared a lot about it.
All the best, --stolfi
(29-01-2026, 10:54 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view. (29-01-2026, 09:36 PM)DG97EEB Wrote: You are not allowed to view links. Register or Login to view.so for me the logical conclusion is the pictures have value but the text doesn't.
I believe exactly the opposite: the pictures are mostly fancy decoration badly copied by the Scribe from random books (each day we find one more), while the meaningful contents is the text.
For one thing, if the pictures are what matters, what would be the point of the Starred Parags section? It is not gibberish; the natural-language-ish features are as strong there as elsewhere, and several details say to me that the Author cared a lot about it.
All the best, --stolfi
Yes, I agree that's a flaw in the argument.. they are mid 15th Century Paduan recipes, but that doesn't mean it's not a skeumorph.
[
attachment=13748]
(29-01-2026, 12:39 PM)Rafal Wrote: You are not allowed to view links. Register or Login to view.How do you build such websites with LLM?
You tell ChatGPT "Make me a website with statistics on word frequencies in Voynich Manuscript"? 
I guess it is more complicated.
Well, it's a tad more complicated, yes. To answer this properly I'll need to give you good news and not so good news.
The good news. I am not an educated man. I have no college degree and to be quite honest, I'm a truck driver. Saying that should clue you in that just about anyone with curiosity and a strong desire can do what I have and lots more.
The less than good news. I have over 40 years of practice writing code so I'm well beyond that 10,000 hour mark. I'm still far from an expert but I'm competent. You won't need to spend 40 years but, even with AI doing the heavy lifting, knowing how to write at least one language is (currently) something I strongly advise. It does make mistakes and if you have no idea how to describe those mistakes or locate them yourself, you'll end up spending a lot of time going in circles trying to tell it what to fix, it not fixing it or worse, breaking something else when it does try to fix it.
The good news. You can, if you wish, teach yourself how to write code. You won't know all the cool phrases that IT pros use but you can develop a deep understanding of how to write code in most any language.
My suggestion to get started, just use the free chat gpt. It's competent in many languages like php, python, c#... it can even work in my old favorite, Pascal. Have it write the code and then you study it. Have it create things you actually want, you'll be more excited when you do get it working.
That's about the best advice and warning I can give with one caveat. Once you do learn, you'll likely never stop writing code.
(29-01-2026, 07:57 AM)DG97EEB Wrote: You are not allowed to view links. Register or Login to view. Did you build it in AI Studio? Btw, it does run on mobile in "desktop mode"
Sorry I didn't answer this earlier but this will take a bit.
I tried AI studio but hated it. To me, Gemini is horrible at writing code. Claude is probably the best but I'm on a 'retirement' budget and it gobbles tokens like a pig eating slop.
What I settled on was a combination of Visual Studio Code with Codex and GPT 5+.
The entire site actually a giant WordPress plugin. I've been writing plugins for around 15 years and I have a boilerplate I created. So, any new plugin I start, all the WP API is already there. The basic method was first, have GPT create a general template for the page. If you'll notice, most of the pages have 2 'HTML cards' up top where the transcriber, folio, section is selected and another below that to show what was selected. Below that, cards and charts are created as needed for the specific page.
Once the template was created, I used GPT to create a set of very loose instructions for Codex, which has full access to the entire plugin. I give Codex the instructions and then wait for it to produce the page and make changes to the other files. I then FTP them up to a test server and check them. Usually, it gets the basic code right. Once thing it sucks at is UI/UX creation. So then, I send the finished code back to GPT and have it do 'surgical' repairs, which it is pretty good at.
I know, it's not the most elegant flow but, I pretty much have it set up and Codex knows what I'm doing now. I get an idea for a page, I give it the instructions and wait to see how cool the page looks when it's done.
And yea, it will work in desktop mode. That's where all the CSS is focused. I'll eventually get around to doing the mobile css but I'm still having fun making pages.
Again, thanks.