The Voynich Ninja

(01-09-2026, 11:15 PM)Mauro Wrote: You are not allowed to view links. Register or Login to view.In Italian too. However EVA 'qo' is arbitrary, I would not rely much on its similarity to Latin 'qu'. On a sideline: if I remember correctly, German is rather special because not only 'qu' is bound nearly 100 percent of the times, but 'ch' is bound too.

Yeah, I don't think “qo” has anything to do with ‘qu’ either, but it shows what the writer “knew.”

And there are several features that are reminiscent of the German language:

“ch” is indeed a bound letter in German. If you look at “sh,” the curve of the first letter could form a sort of “s,” so that it could stand for “sch,” which is the soft form of “ch” and is also a bound glyph sequence in German.

Also striking is “aiin”—“ain” is a distinctly German word, derived from “ein,” and “ain” gave rise to the English “a.”

“dain” = “dein” = “your” is also a real German word.

That’s why it looks as though the writer knew the German language, but that’s just an assumption, too.

But if anyone here has an idea on how to crack this specific—but actually simple—VBM approach, I'd really love to hear it

(01-09-2026, 09:18 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.As I understand it, the “Stars” text is the most fully developed part of this cipher. It appears to have been further developed as it was being written.

I would very much like to see the evidence for this claim, for it being 'developed as written'.

(02-09-2026, 07:49 AM)dashstofsk Wrote: You are not allowed to view links. Register or Login to view.I would very much like to see the evidence for this claim, for it being 'developed as written'.

The reason isn't that the system is changing—but rather that it isn't actually changing, yet something within the system is changing:

For example, this oddity: Initial glyphs

Section    qo     d       Total
Herbal-A  6.6   13.1   19.7
Herbal-B  8.7   10.7   19.5
Stars-B   16.3   2.8    19.1

The qo channel rises while the d channel falls—but, and this is the crucial point: the sum remains constant at ~19. At the same time, the overall density of space transitions across all sections is practically identical (85–90 per 100 tokens), and the ch/sh class remains stable at 22–26.

So something seems to be “evolving” here—which is why I’m bringing it up.

This is also consistent with the fact that the od/ed ratio is almost completely reversed between A and B (which has been known for quite some time), and another interesting point: the daiin proportion drops from 3.7% (Herbal) to 1.1% (Stars).

From this, I conclude: The basic framework (the cipher) is the same in A and B—otherwise, the totals and the transition density wouldn’t hold so precisely—or am I mistaken?

What is shifting, however, is the choice of certain glyphs and glyph clusters within the system:

That would be exactly the pattern one would expect if someone uses a method over a long period of time and identifies where it falls short. The basic rules remain the same, but some “habits” or perhaps list entries, etc., shift.

Why do I classify Stars as “the most developed”? It’s simple: Stars is at the endpoint of this trend—lowest d-channel, lowest daiin, highest qo and l usage. All of these characteristics, which increase from A to B, are most pronounced there; all those that decrease are least pronounced.

Well, you have to realize: This is a drift across the sections, not a proven timeline. And, of course, this thesis stands or falls on that.
So it doesn’t refute the thesis that there were simply different scribes.

But if you look at it closely, this idea of development/error correction is probably the simplest explanation with the fewest assumptions, which thus also accounts for all three findings at once.

But that's just my humble guess.

(02-09-2026, 08:13 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.highest qo

Not so. The attached output shows that it is quire 13 that has a higher frequency of prefix qo .

But also it has been seen for a long time that there are many language clusters. Different sections of the manuscript have differences in the writing. It is not just an A/B difference.

[attachment=17527]
Interrogative words in German always begin with a ‘W’: wo, was, wie, wer, wann, warum, etc.
In Latin, on the other hand, they begin with ‘qu’: qua, qui, que, quo, etc.
When I look at recipes, ‘lot/lott’ is often used in German to indicate quantity.
But so are ‘quentin’ and ‘quantum’.
In Latin, however, ‘lot’ is not used (I’ve never seen it), but ‘quentin’ and ‘quantum’ are.
Now there’s a ‘qu’ explosion in recipes. But only in recipes.

Translated with DeepL.com (free version)

(02-09-2026, 06:02 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.This was only a different approach to cracking the cipher, since a homophonic cipher—which may also have a length dependency—cannot be cracked using pure frequency analysis alone.

But still, if it’s a cipher, the decryption should be unambiguous. A well-known example of a homophonic cipher is Copiale — this essentially confirms it. After removing the Latin letters, it instantly succumbed to cracking, given its complexity.
Could you clarify whether the cipher provides different translation options for a single string (for example, two meaningful translations), or whether all translations except one are meaningless?

(02-09-2026, 03:48 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.But still, if it’s a cipher, the decryption should be unambiguous. A well-known example of a homophonic cipher is Copiale — this essentially confirms it. After removing the Latin letters, it instantly succumbed to cracking, given its complexity.

The key lies in the structure:

If there are few homophones and clear word boundaries, it’s usually relatively easy to decipher.

If there are many homophones with a well-balanced frequency, it’s already much more difficult.

But if there are also variable spellings, zero characters, obscured word boundaries, or context-dependent selection of homophones, it becomes considerably more difficult.

If the choice of homophone itself depends on position, word length, or neighboring characters, strictly speaking, it is already more than a simple homophonic substitution—and it can become extremely difficult.

This is especially true when the underlying language is subject to highly flexible spelling conventions, as was the case in Middle High German.

VBM breaks down word boundaries and operates at the level of consonant clusters and vowel bridges—that is, not at the token level. If both are interpreted as homophones, it becomes nearly impossible...

(02-09-2026, 03:48 PM)ololololo Wrote: You are not allowed to view links. Register or Login to view.Could you clarify whether the cipher provides different translation options for a single string (for example, two meaningful translations), or whether all translations except one are meaningless?

I've done a lot of calculations and analysis, but without plain text, it's hardly realistic to answer such questions. However, it seems that, when it comes to vowels, multiple bridges represent a single vowel, and the most common consonant clusters can be represented by several different VMS Cluster.
In addition, there’s a second level—besides the families—and that seems to have been encoded differently yet again.

But, to be honest, this is all just speculation, of course.

(02-09-2026, 09:04 AM)dashstofsk Wrote: You are not allowed to view links. Register or Login to view.Not so. The attached output shows that it is quire 13 that has a higher frequency of prefix qo .
But also it has been seen for a long time that there are many language clusters. Different sections of the manuscript have differences in the writing. It is not just an A/B difference.

Yeah, you're right—I did the math a bit too roughly. I hadn't really considered “Quire 13”; as far as I remember, “Stars” was the highest, so I checked ‘Stars’ and “B” instead of the individual sections...

So, sorry. I’ll think about it some more.

I ’ve been thinking about it, so here goes: Crazy idea:

Maybe the VMS isn’t a special, encrypted work at all, but rather a cipher experiment. Someone developed a cipher—perhaps in collaboration with others—and they tested it on the book by encrypting another book or sections of it, to see what worked and what didn’t.

Perhaps this also explains the “qokedy” repetitions (among other things)—they were trying to see how far they could take it.

In that case, the various differences across the sections would be, in part, tests to see what works better and, in part, actual further development.

That would explain why the vellum is of such poor quality—they were simply scraps. The plants are probably just figments of the imagination, too, because the point wasn’t to draw real plants; the fact that the nymphs ended up naked was perhaps just a bit of fun on the part of the people working on the cipher. And the other drawings may have been nothing more than tests as well. Referencing astronomy with the cipher, plants, etc...

I know this isn’t proven by anything, so it’s just another crazy theory , but it could explain the differences between the individual sections. I still feel that the Stars section is the “most advanced,” but maybe that’s just my imagination.