The Voynich Ninja

(07-09-2026, 07:55 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.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.

A similar theory already exists (from English Wikipedia):
Rafael Mnishovsky, the friend of Marci who was the reputed source of the Bacon story, was himself a cryptographer and apparently invented a cipher which he claimed was uncrackable (c. 1618).You are not allowed to view links. Register or Login to view. This has led to the speculation that Mnishovsky might have produced the Voynich manuscript as a practical demonstration of his cipher and made Baresch his unwitting test subject. Indeed, the disclaimer in the Voynich manuscript cover letter could mean that Marci suspected some kind of deception.

However, it’s unclear why such a specific book needed to be encrypted. One could simply write a bunch of text, encrypt an ordinary fairy tale, or a made‑up text.

(07-09-2026, 07:55 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.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.

It is not crazy at all.  There are historical examples.  That was a natural way for a cryptographer to prove his worth to potential employers.

One such precedent is famous and well-known to cryptographers (Agrippa? Trithemius? Forgot who.)  

IIRC, another one was written by Raphael Mnishovski, Marci's friend who originated the idea that the VMS was a Roger Bacon original bought by emperor Rudolf.

However, for that purpose, the drawings of the VMS would be wasted work and wasted vellum.

All the best, --stolfi

Just to be clear, I’m not attached to this theory—it’s simply one of many possible theories.

But the idea that it was also done to present the cipher does make sense. If you’re creating an entire book for that purpose, you’d naturally include plants and other illustrations; the rather poor quality of the drawings would then be further evidence of that.

There were plenty of people who practiced the forbidden arts who might have been looking for such a cipher. 

But who knows... 

I could advise you to publish more translations in this thread. First, it’s interesting, especially given your conclusions about vascular systems. Second, it will help you better understand your theory.

Well, personally, I’m not very good at understanding your statistical conclusions

A translation of 5 lines (not a solution)

I’ve written about this before: Some of the statistical peculiarities suggest that the VMS might be encrypted in reverse, likely line by line. Take, for example, the strange long-range effect of “ch” and “sh” before the “E” family on the “qo” of the next word. But there’s quite a bit more to it than that. That’s why I keep testing whether generators, among other things, work better in reverse. This test is no exception—it’s simply an attempt to force the AI to read plaintext from the VMS using a VBM cipher that’s optimized for the text but internally coherent, as already described above
 
(As I’ve emphasized several times, this is NOT A SOLUTION).

And indeed, it works better in reverse: Five consecutive lines (!) spanning line breaks were successfully "translated" (or, to put it more accurately and honestly: crammed  ) into a comprehensible and grammatically correct sentence.

Reva = Reverse EVA:

f115v.8 madchp ydechekl rako ydechpl lotoq ydchto seechlo schtoq loesh sorocht 
f115v.9 ror yechl yechkl niatoq ydoch ychtoq ydeekloch niatoch niiadch seechy 
f115v.10 ydech oeeto ydechlad niiako ydoesh osh niias 
f115v.11 matoq ydechpd yechtoq dechtoq lochkch lokch niiadoch liarosh rodcht 
f115v.12 niiaros ydechkl ydechl deeshy

= Wer den schweren Magen bessern will, der nehme etwan warmes Salz, dan Wasser und braue eine Weile lau, so gar, einen Essigtrank. Gib dem Kranken Beeren.

(Whosoever would mend a heavy stomach, let him take some warm salt, then water, and brew it lukewarm for a while fully into a vinegar draught. Give unto the sick man berries.)

Note: "etwan" (= somewhat/at times) and "dan" are attested 15th-century forms (etwan 3×, dan 103× in my Bavarian reference corpus; dan 18× in Ortloff, so gar = vollends - nicht "sogar")

Broken the german stream down by vowels and consonant clusters:

W-e-RD-e-NSCHW-e-R-e-NM-a-G-e-NB-e-SS-e-RNW-i-LLD-e-RN-e-HM-ee-TW-a-NW-a-RM-e-SS-a-LZD-a-NW-a-SS-e-R-u-NDBR-aueei-N-e-W-ei-L-e-L-a-∅-u-S-o-G-a-R-ei-N-e-N-e-SS-i-GTR-a-NKG-i-BD-e-MKR-a-NK-e-NB-ee-R-e-N

(∅ marks the token osh: it has no token-internal material, so it contributes an empty cluster. Normally, a maximal vowel run is assigned to one bridge (as with ee, ei). The L2 token osh is the special case here: because it has no internal core, 
the vowel run au in lau is distributed across its two adjacent bridges, sh.o = a and sh.n = u.)

The vowel classification:

p.y=e | l.r=e | o.y=e | l.l=e | q.y=a | o.s=e | q.l=e | sh.s=i | t.r=e | r.y=e | l.y=ee | l.n=a | ch.y=e | ch.n=a | ch.s=e | y.y=u | ch.o=aueei | d.n=ei | sh.o=a | sh.n=u | s.m=o | d.y=ei | q.d=e | ch.l=i | sh.r=e | t.n=a | s.y=e | l.d=e

Multiple Use (all consistent):
q.y=a = 4×
o.y=e = 3×
ch.n=a = 3×
o.s=e = 2×
q.l=e = 2×
ch.l=i = 2×
l.y=ee = 2×

Remaining weak spot: ch.o=aueei — a single bridge carrying the whole vowel run of "braue eine". Internally consistent with the scheme (every maximal vowel group = one bridge, same as ee, ei), but clearly not a good choise.

Consonants:

adch=w | dechek=rd | ak=nschw | dechp=r | oto=nm | dcht=g | eechl=nb | chto=ss | oe=rnw | oroch=lld | o=rn | ech=hm | echk=tw | iato=nw | do=rm | chto=ss | deeklo=lzd | iato=nw | iiad=ss | eech=r | de=ndbr | eet=n | dechla=w | iiak=l | doe=l | osh=∅ | iia=s | ato=g | dechp=r | echto=n | echto=n | ochk=ss | ok=gtr | iiado=nkg | iaro=bd | odch=mkr | iiaro=nk | dechk=nb | dech=r | eesh=n

Multiple Use (all consistent):

chto=ss = 2×
dechp=r = 2×
echto=n = 2×
iato=nw = 2×

The actual conclusion from this small example is as follows:

Of course, this arrangement has weaknesses, which I also recognize, including too many degrees of freedom and other issues. But the point isn’t that this is already the actual translation. It’s about something else:

With a consistent vowel bridge/core scheme, it’s actually possible to create a coherent German text spanning multiple lines—without having to completely reinvent the system after every line!

This proves that the mechanics of the Vowel Bridge Model are fundamentally powerful enough to construct meaningful plain text from the VMS, even across multiple lines.

This example also demonstrates that the statistical peculiarities of VMS (such as its peculiar entropy or rigid word architecture) do not necessarily contradict the existence of a genuine human language.

----

And if you also show that this model is capable of translating such stange strings like: 

<f75r.13,+P0> pchedy.keedy.qokedy.qokedy.qokedy.qokedy.qokain.olshedy

You are not allowed to view links. Register or Login to view.

or olcholcho

You are not allowed to view links. Register or Login to view.

The VMS can quite simply be a coded language, and the VBM could, in principle, be the system behind it.

P.S. If anything, I’m aware that this isn’t a solution. The questions are about the theory in general, but they’re based on this example

Thank you very much! My guess was confirmed — you’re treating this cipher as a substitution, where a space can also be part of the cipher alphabet. I thought it would be more complicated...
But I still have some questions:
1). Why do you need to reverse the text? Does it significantly change the translation? If you translate the non‑reversed text, will you get a translation that is reversed in the same direction?
2). Do these consonant combinations cover the entire VMS text? This seems rather far‑fetched. In that case, the cipher alphabet would look very strange.
3). I would like to ask if you could describe the operation of the cipher in general terms and, in particular, send me the cipher alphabet? I would like to try this procedure myself; it looks very interesting  I will publish the result in the thread with the phonetic experiment.

And I apologize, but I can’t help but say this…

(09-09-2026, 02:57 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.pchedy.keedy.qokedy.qokedy.qokedy.qokedy.qokain.olshedy

This can be done using Roman numerals. Again, if you align them in order in the text, you will easily end up with such sequences.
I’m not saying my theory is better - it’s just more trivial. This, one might say, is the theory of a banal substitution: Letter – Number – Voynichese. But so far, I don’t have any meaningful translation, or even a hint of one.
And in a very simple version, some features of the VMS text also appear in my mini‑cipher (yes, it’s not perfect, but it’s created for the sake of clarity).

Among other things, there is the Naibbe cipher. In my opinion, it shows that the same results can be achieved using simple and modern methods for the author of VMS.

No, it’s a homophonic cipher (!), meaning a plaintext letter can be replaced by several different ciphertexts. So it’s not a pure substitution cipher.

There are many small clues suggesting that the reverse makes more sense. As I mentioned earlier, the phenomenon where “ch” and “sh” before “e” predict how often “qo” appears at the beginning of the next word. This very strange long-range effect is resolved under Reva (Reverse EVA).

The “qo” atom.

Forward: “q” follows “o” 97.7% of the time. Thus, “q” carries almost no information about the following character in the forward direction → “qo” as a unit is justified. Additionally: 99.4% token-initial.

In Reva (reverse reading direction), the “q” is, so to speak, merely a choice at the end of the word (where it appears in Reva); there are then other glyphs that follow “o,” making it much more “natural”—

But the most important reasons for the VBM are:
The VMS families—the “e” family, for example—often seem to have a glyph (e.g., K) preceding them, but consist of several glyphs themselves. In REVA, this is reversed, and that fits much better with the distribution of consonants between vowels in German.

The “aiin” families fit perfectly with the typical endings in German (e.g., “en”), but they’re in the wrong place—they should actually be at the beginning—and Reva balances this out perfectly as well.

There are also several minor statistical effects here, but overall, the VBM—assuming it uses a German cipher—fits the backward-read variant much better. And, by the way, that was also a typical cipher variant of the Middle Ages.


The five translated lines were still just a proof of concept; the actual cipher will consist of several tables, with about 60 to 80 basic patterns that are combined.

I’m nowhere near finished with this cipher—or rather, with several variants of it—so I don’t want to release anything until I’m at a point where it makes sense to do so.

The VBM cipher is much simpler than the Naibbe cipher.

And regarding the VBM, I’ve attached the VBM PDF again; hopefully, it explains it very clearly. But it’s also a bit outdated.

Regarding Reva:

Something I forgot to mention: In VBM, vowels are defined by the two letters surrounding the space.

In reverse, the structure is clear: First comes a highly variable letter that introduces the vowel bridge and defines the vowel. These are all the letters with which a token begins in eva (i.e., in forward order).

In Reva, this is followed by the transition into the families: y = into the e family, n = into the aiin family (niia in reverse), the r family, and the l family.

This looks like a clear structure that makes much more sense in reverse. The last token of the word branches off via the bridge into the various families. It looks like a grammar...

[attachment=17582]

I’ve looked through the file. It’s interesting and detailed, but for the most part, I didn’t learn anything new about VBM.

(10-09-2026, 07:08 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.The VBM cipher is much simpler than the Naibbe cipher.

I don’t know how many symbols you have in your cipher alphabet, but it seems to me that they’re roughly on the same level. For Naibbe, you need to constantly throw cards; for VBM, you need to manipulate the alphabet.

And another question: in the examples above, you say that this is not a solution. I can’t understand whether you have any translations at all? Or is there a reason to hide them that I don’t know about yet?
Again, as I understand it, you translated part of the Pharma. I don’t mind keeping it secret, but nevertheless, I just want to know whether there are any translations as such...