The Voynich Ninja

@ oshfdk

No! The difference is significant: Most of the older models cited simply reads the two letters across the gap as a bigram pair and derives a substitute alphabet from it. That approach is outdated, well-known, and cannot work. I agree with you.

I perform a phonological classification. The VBM assigns a phoneme class to the bridge position (bridge = vowel, internal glyph = consonant). This is not a statement of the form “there is a hidden bigram in the gap,” but rather a vowel-consonant role reversal across the entire token chain. The older model lacks this class-based logic. But it is essential for deriving a linguistic structure from the VMS - otherwise it doesn’t work, because the CVCV structure typical of language would not emerge.

(08-06-2026, 10:29 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.why people don't realize

The problem, JoJo, is that we can't understand your claims. This thread had already run to more than 26 pages. You are overloading us with too many posts. It is awkward for anyone to have to read them all.

(09-06-2026, 06:58 AM)dashstofsk Wrote: You are not allowed to view links. Register or Login to view.The problem, JoJo, is that we can't understand your claims. This thread had already run to more than 26 pages. You are overloading us with too many posts. It is awkward for anyone to have to read them all.

I totally get it! But sorry, what can I do? I would have liked to start the VBM in a new thread. But since I’m sticking with MHD as the background language, the forum rules don’t allow it... Maybe I can ask Tavie to make an exception, since it is a new theory?

The VBM starts on page 22! Beitrag You are not allowed to view links. Register or Login to view.

(08-06-2026, 10:29 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.[..]—where spaces are not word boundaries but rather delimit vowel bigrams, combined with the well-known slot system—

Because this is just another ˋmethod´ to magically add some extra characters (in this case, „splitted vowels“) to the very limited character collection VMS.
In the desperate hope to produce something recognizable and familiar; others tried this by summing up completely different meanings in each 1 VMS-character, „rotating“ translations etc.

Are you sure your ˋmethod´ is new…?

@ Stefan

...

(09-06-2026, 01:18 PM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.@ Stefan
...

Yes, you give very qualified replies…

You started your thread under the flag „why and how the text could be Bavarian“ here — are you still on track…?

.....

I've have possibly "cracked" a part of the structure of the Cores (formerly C2) — don't worry, not a solution, no plaintext yet, just the structure. But before I lay this out, I need to bring you up to speed.

Something occurred to me. We all know the relationships between ched and shed with y and qo. Less well-known are the relationships of the "e" family in Slot 3 to the "o" after y, and some others. A few things just didn't seem to make sense. 

An example:
Why does schedy get followed by q 51% of the time?

[attachment=16079]

And why only 39% for chedy?

[attachment=16078]

Why doesn't the value in front of it change when the run gets shorter?

[attachment=16082]

Why does kedy then sit at only 28%?

[attachment=16083]

Okay, part of the effect has to do with the number of hits—the “n” in the graph.

But the difference between Kedy and shedy—n = 621 / 542—isn't affected by that.

So how do you explain the fact that Slot 2 has an obvious influence across the Space?

And: Isn’t that an indication that the spaces aren’t really spaces, since they all continue running smoothly toward K? We don't know, but it.... 

Or let’s take another example: daiin and aiin

[attachment=16084][attachment=16085]

They run almost synchronously after the space (the possible divergences are plotted here as percentages)—amazing, isn’t it? What are Spaces???

My solution - Part 1

Well, that puzzled me. And then something struck me: we all know that "qo" is actually two characters (because they aren't fused like ch and sh) and yet they're tied together nearly 100% of the time. If you approach Voynichese from the back, reading line by line in reverse, then q is simply a normal successor to "o". Sure, o is exclusive to q, but not the other way around. And once you approach those chains above from behind, the whole thing suddenly makes sense. The high connections between qy and the E-family suddenly reveal a clear structure — the relationships develop from left to right, against the writing direction.

With this view, a few other things tipped over too:

1. The aiin / air family lines up beautifully with the endings in MHD/FNHD — read in reverse, they show up exactly where they should: in C1, i.e. at the word/morpheme end. Previously I had to somehow reverse-engineer them into C2 as endings, which was very unsatisfying and led to downstream errors in the stream.
   
2. The vowel bridges, read in reverse, strongly conditions the following E-family — meaning they influence it. This is where the actual code of the Cores sits. Running it in reverse now, it works exactly as it should. Previously you had to build the existing connection backwards from the end to the front to construct a matrix — but that's not how humans work, and I didn't like it. Now it runs cleanly in sequence: first the vowel, then the switch, then the E-family. Exactly the way you'd build it up.
   
3. The LEMs become the actual line-start markers, and you need so few of them because the only thing that has to be clarified is whether the line starts with a vowel or a consonant. What the former Line Start Markers are actually doing now — I haven't looked into that yet, but they probably terminate the stream.
   

The whole thing was a bit like the planet effect: as long as you think the Earth is at the center, you can barely make sense of those strange planetary motions, patching them up with little epicycles. The moment you see that the Sun sits at the center, the orbits suddenly become strictly, logically explainable. All at once, everything in the VBM fell into place.

These extreme oddities disappear quite easily when we look at the sequence in reverse:

Here, the sequence is viewed from the back, starting at k (where the symbol ends at the top)—the most possible branches are indicated here. You can see in the last few branches that everything is distributed fairly evenly.

[attachment=16086]

Some of the most awkward dependencies become much more natural in mirror direction. K does not influence ch. It is a simple sequence of dependencies.

When the sequence is read normally from left to right, oddities arise that make sense if the current were actually flowing in the opposite direction.

But that’s not all—the structur will become clearer in the next post.

Part 2

To put the directional pattern into a broader context, I converted the structural data into a Sankey-style diagram.

[attachment=16087]
Edges below 50 occurrences are omitted for readability, so node totals are not meant to be strictly conserved.

The diagram shows the bridge structure from a mirror/reverse-reading perspective:

VL -> VR -> following family

Here, VL and VR are the two parts that, across a word boundary, form a bigram that is read as a vowel. In principle, the bigram can convey two possible pieces of information.

What I find interesting is not just one particular chain such as kedy qo..., but the broader pattern. Many different VL inputs on the left do not scatter randomly. Instead, they converge through a small number of VR carriers into a few distinct output families. "Chaos" turns into order, as you can clearly see in the picture

This is the point I was trying to make above: in normal EVA direction, these patterns appear as strange dependencies across word spaces. In mirror direction, they become local routing structures. The visible EVA space lies at the hinge where the stream is routed into a family.

From a cipher perspective, this is where it becomes interesting.

One possible interpretation is that the bridge is not just a vowel-like bigram. It may encode both the carrier family and a selector for how the following family form is to be interpreted. The VL glyph does not carry a fixed value on its own - it's may function as a selector. VR is the vowel. The combination of VL + VR then determines which route of the following family is active.

For example, using the E-family only schematically:

VL = q + VR = y -> E-family route A
VL = o + VR = y -> E-family route B
VL = ch + VR = y -> E-family route C

In such a system, a form like "ed" would not have one fixed value everywhere. Its value would depend on the bridge leading into it.

So, schematically, as an example (no plain text) reverse reading:

Assumption:
VR = y = "e"

q + y + ed could represent one value qy=e+ed = "en"
o + y + ed another, oy = e + ed = "em"
ch + y + ed another. chy = a+ ed ="den"

The point is not a specific plaintext value here. The point is the mechanism: the repeated family form may remain visually similar, while its actual function is selected by the bridge that leads into it.

The same principle may apply to other families as well: the AIIN-family, the A/AR-family, the O/OR-family, and so on. Some of those links are partly definitional, so I would not use them as proof by themselves. The more important point is the overall routing pattern: broad input on the VL side, a smaller number of VR carriers, and then a much clearer family structure.

Of course, this is still only a structural hypothesis. But it offers a possible cipher-based explanation for two otherwise very odd features of the VMS:

First, why the text is so repetitive on the surface.

Second, why these repetitions are still so strongly constrained by their surroundings.

In other words: the repeated forms may not be repeated plaintext words. They may be repeated family-forms whose value is selected by the bridge that leads into them.

And:

The visible EVA spaces do not behave like ordinary word boundaries. They appear to lie inside a larger structural unit, possibly at the hinge of a routing mechanism.


The third post will follow a little later...

The reason I called the centre or right justified lines ‘Titles’ was because they felt like the text might actually be RTL and bottom to top.