29-08-2026, 09:13 AM
(This post was last modified: 29-08-2026, 10:18 AM by JoJo_Jost.
Edit Reason: I'm always mixing up tokens and glyphs, even though I actually know what's what.... I have no idea why...
)
The Length-Dependent Part of the Cipher
As I’ve been writing for some time now: I realized that I was still missing an important part of the cipher—the missing 20 percent....
I’ve found it. And with that, everything is now much clearer—and, unfortunately, also a bit more complicated:
In addition to the vowel bridge, the VMS also uses a length-dependent ciphering of the cores.
I can now also see how the cipherer went about it. There was likely initially a length-dependent cipher for tokens up to 3/4 glyphs in length and for those with longer tokens. This had to be the case because vowels tend to be the first and last letters of the token. This means that this principle only works if words are longer than three letters. I had already intuitively realized this and treated the short tokens differently from the longer ones—without realizing the logic behind it. This also explains why the short tokens are so very different from the longer ones (something that has been known for a long time).
After all, he naturally needed separator characters as well—for example, when one word ended with a vowel and another began with one, and for other cases that arise automatically due to the VBM.
This gave him the idea to incorporate length dependence into additional tokens as well, and that’s how he came up with the counter letters: which, as in the Eva transcription, are read as “e” and “i.” But in principle, they’re just dashes.
At first, I did recognize the counting structure and ran through every conceivable variation, but I was always missing a part of the structure, and it just wasn’t consistent no matter how I approached this matrix.
Until it dawned on me that he had actually taken the length-dependent game even further: He had different lists for words of different lengths.
Example: So “ked” in a word that’s 5 glyphs long can mean something different than “ked” in a 6-glyphs word. He introduced the counter strings to create a second level, making the words “longer” and thereby placing them in a different length category—where they then have a different meaning.
VMS “ked” = “nd,” but now he needed a different consonant cluster, so he simply inserted a second “e,” resulting in “keed”—this could be (!) a completely different cluster that appears in a different list. This way, he expanded his selection while simultaneously making the cipher nearly unbreakable.
That explains why there are so many words that look the same but are missing a letter or have an extra one. Even though they look almost the same, they are completely different words.
And only in a final step did he think to himself: Oh, I can apply this to the vowels too. Because with them as well, you can see a length-dependent change in their context. That is, the vowel bridges behave differently depending on the length of the word.
And that’s exactly the aspect of the cipher that accounts for the missing 20 percent. Very exciting.
I know: I still have to prove it, but I'm working on it. I've found a few more clues, and I'm now putting it all together like a crossword puzzle. It's going to take some time.
Unfortunately, this theory (and that’s all it is so far) has completely thrown off the basic structure of my generators, so I’ll have to keep working on that as well. But since I’m on my own, that’s obviously a lot for one person to handle.
Conclusion:
But I think we should at least consider the idea that this cipher could habe a component that depends on word length.
It would explain why the cipher has driven so many cryptologists to the brink of madness—in addition to the VBM structure. The two together are nearly impossible to crack.
But it still roughly fits the ciphers of the time; only the complexity is a bit anachronistic.... I’ll admit that, but that’s what the numbers show...
As I’ve been writing for some time now: I realized that I was still missing an important part of the cipher—the missing 20 percent....
I’ve found it. And with that, everything is now much clearer—and, unfortunately, also a bit more complicated:
In addition to the vowel bridge, the VMS also uses a length-dependent ciphering of the cores.
I can now also see how the cipherer went about it. There was likely initially a length-dependent cipher for tokens up to 3/4 glyphs in length and for those with longer tokens. This had to be the case because vowels tend to be the first and last letters of the token. This means that this principle only works if words are longer than three letters. I had already intuitively realized this and treated the short tokens differently from the longer ones—without realizing the logic behind it. This also explains why the short tokens are so very different from the longer ones (something that has been known for a long time).
After all, he naturally needed separator characters as well—for example, when one word ended with a vowel and another began with one, and for other cases that arise automatically due to the VBM.
This gave him the idea to incorporate length dependence into additional tokens as well, and that’s how he came up with the counter letters: which, as in the Eva transcription, are read as “e” and “i.” But in principle, they’re just dashes.
At first, I did recognize the counting structure and ran through every conceivable variation, but I was always missing a part of the structure, and it just wasn’t consistent no matter how I approached this matrix.
Until it dawned on me that he had actually taken the length-dependent game even further: He had different lists for words of different lengths.
Example: So “ked” in a word that’s 5 glyphs long can mean something different than “ked” in a 6-glyphs word. He introduced the counter strings to create a second level, making the words “longer” and thereby placing them in a different length category—where they then have a different meaning.
VMS “ked” = “nd,” but now he needed a different consonant cluster, so he simply inserted a second “e,” resulting in “keed”—this could be (!) a completely different cluster that appears in a different list. This way, he expanded his selection while simultaneously making the cipher nearly unbreakable.
That explains why there are so many words that look the same but are missing a letter or have an extra one. Even though they look almost the same, they are completely different words.
And only in a final step did he think to himself: Oh, I can apply this to the vowels too. Because with them as well, you can see a length-dependent change in their context. That is, the vowel bridges behave differently depending on the length of the word.
And that’s exactly the aspect of the cipher that accounts for the missing 20 percent. Very exciting.
I know: I still have to prove it, but I'm working on it. I've found a few more clues, and I'm now putting it all together like a crossword puzzle. It's going to take some time.
Unfortunately, this theory (and that’s all it is so far) has completely thrown off the basic structure of my generators, so I’ll have to keep working on that as well. But since I’m on my own, that’s obviously a lot for one person to handle.
Conclusion:
But I think we should at least consider the idea that this cipher could habe a component that depends on word length.
It would explain why the cipher has driven so many cryptologists to the brink of madness—in addition to the VBM structure. The two together are nearly impossible to crack.
But it still roughly fits the ciphers of the time; only the complexity is a bit anachronistic.... I’ll admit that, but that’s what the numbers show...
![[+]](https://www.voynich.ninja/images/collapse_collapsed.png)