Codex Mondragonis

Beinecke MS 408 · Voynich Manuscript research

The research on the origin of Beinecke MS 408

After years of statistical, lexical and iconographic analysis, the manuscript's origin has been identified in the cluster of Germanic language islands of the eastern Alps (Sprachinseln): Carnia (Tischlbong/Timau, Sauris/Saurano), Cadore (Sappada/Sappadino) and the Venetian Prealps (Cimbrian) — Alpine High German dialects, 13th–15th centuries. The Timau/Tischlbong lexicon shows the highest density of correspondences (638 attested entries). It is a medieval herbal-recipe book, to be read by the healer's visual protocol: image plus a reduced lexicon.

Browse the manuscriptThe hypotheses testedThe transposition coming soon

The parchment is radiocarbon-dated to between 1404 and 1438. The book's structure — a herbal, an astrological-medical calendar, a balneological section, a pharmacy with albarello jars — is the knowledge of an early-fifteenth-century physician-apothecary. The text's statistical signature, low conditional entropy and a short, redundant lexicon, is that of a specialised dialect lexicon: not a full European language, not a cipher.

1404–1438
Radiocarbon dating
Parchment, Italo-Germanic Alpine area
~2.38 bits
Conditional entropy h2
European languages sit at 3–4
184
Navigable pages
Of 213 Yale scans
30
Hypotheses tested
4 in progress or convergent; the others excluded against the baseline

Our method

Five levels of analysis, one direction

The research combines five original, documented levels: statistical analysis with a baseline control (p-value < 0.05), multi-dialect phonological reconstruction from EVA graphemes to Old High German, phonetic clustering on IBM quantum hardware, semantic analysis of plants and iconography, and reading by the healer's visual protocol. It is not traditional philology: it is a statistical-computational approach integrated with historical linguistics and medieval iconography.

The key measure is the Voynich's conditional entropy (h₂ ≈ 2.38 bits): it rules out any standard European language (Latin 3.50; Greek 3.42; modern German ~3.3) and any simple cipher, which would preserve the source's entropy. This signature is the proof that the text is a specialised, short and redundant lexicon, not enciphered prose. Every strong claim carries its data; every hypothesis its random control.

The statistical analysis, the phonological reconstruction from the EVA graphemes and the phoneme clustering are carried out with the proprietary artificial-intelligence technologies of YurekAI, the research's technological engine.

How the validator worksThe hypotheses tested

The method, level by level

1 · Statistics with a baseline

Every hypothesis goes through the validator: a match counts only if it beats a random control (p-value < 0.05). Conditional entropy h₂ ≈ 2.38 bits is the measure that rules out standard European languages and simple ciphers.

2 · Phonological reconstruction

EVA graphemes are mapped to Alpine Germanic phonemes following the rules of Old High German (consonant shift, Grimm's law) and compared with over ten lexica: Tischlbong, Cimbrian, Sauris, Sappada, Ladin, Romansh, Middle High German.

3 · Quantum clustering

The thirty most frequent EVA sequences are analysed with a quantum kernel on real IBM Quantum hardware, to find dialect clusters: phonetic variants traceable to different Alpine areas within the same codex.

4 · Plants and iconography

Every illustrated folio is studied for botanical identification (Gart der Gesundheit 1485, Hildegard's Physica, Cod. Pal. germ. 539) and for the doctrine of signatures: the plant's shape signals its medicinal use.

5 · Visual protocol

The text is not prose but image plus five keywords: quality, parts, method, organ, virtue. A medieval professional dialect naturally has few recurring words: not a riddle, but the normal feature of any compressed specialised lexicon.

Quantum analysis

What the physics of the text says

We measured the Voynich with tools from quantum information: entanglement entropy (how much a letter stays correlated with distant ones) and lexical richness section by section. Two charts, one outcome: very rich local structure, but no long-range structure like real languages.

Entanglement entropy by distance

Real LatinVoynich (EVA)Voynich (Currier)Voynich (Cappelli)Shuffled (baseline)
0123124812distance k (symbols)entropy S(k)Real Latin · k=1 · S=2.86Real Latin · k=2 · S=2.64Real Latin · k=4 · S=2.28Real Latin · k=8 · S=2.42Real Latin · k=12 · S=2.44Voynich (EVA) · k=1 · S=3.18Voynich (EVA) · k=2 · S=2.86Voynich (EVA) · k=4 · S=2.04Voynich (EVA) · k=8 · S=0.79Voynich (EVA) · k=12 · S=0.48Voynich (Currier) · k=1 · S=3.09Voynich (Currier) · k=2 · S=2.8Voynich (Currier) · k=4 · S=1.93Voynich (Currier) · k=8 · S=0.67Voynich (Currier) · k=12 · S=0.37Voynich (Cappelli) · k=1 · S=3.01Voynich (Cappelli) · k=2 · S=3Voynich (Cappelli) · k=4 · S=2.13Voynich (Cappelli) · k=8 · S=0.87Voynich (Cappelli) · k=12 · S=0.33Shuffled (baseline) · k=1 · S=0.33Shuffled (baseline) · k=2 · S=0.38Shuffled (baseline) · k=4 · S=0.34Shuffled (baseline) · k=8 · S=0.33Shuffled (baseline) · k=12 · S=0.34

S(k) = how much structure survives at a distance of k symbols. The Voynich (in every alphabet) starts very high and collapses toward the baseline within ~12 symbols; Latin stays high and flat. Cappelli expansion does not bring it closer to Latin.

Lexical richness by section

Real Latin ~0,80AstronomicalAstronomical · 0.5810,581HerbalHerbal · 0.510,510PharmaceuticalPharmaceutical · 0.4950,495Recipes / starsRecipes / stars · 0.4640,464BiologicalBiological · 0.350,350

Types over occurrences, at equal sample size (2,900 words). There is a real gradient — astronomical and herbal richest, biological poorest — but every section stays below a real language.

The quantum measurements give the text's signature: very rich local structure, no long-range correlation as in a full European language. It is consistent with a specialised, redundant lexicon, and with the dialect clustering that guides the phonological reconstruction.

View the dossiers and scripts

How this research began

Two different paths, the same mountain

From a reel to a challenge

It all started in early June 2026, with an Instagram video about the Voynich Manuscript, the «world's most mysterious book», undeciphered for over six hundred years. For Diego «DMUX» De Maio and Simona «51m0» Fenoglio, founders of ART AG (YurekAI), it became an immediate challenge: text-entropy analysis, a phonological pipeline on the EVA graphemes, quantum clustering on IBM Quantum, comparison with over ten medieval Alpine Germanic lexica.

The convergence

Over the following weeks the picture took shape: the origin points to the Germanic language islands of the eastern Alps — Carnia, Cadore, the Venetian Prealps — from the 13th century, with the Tischlbong dialect of Timau showing the highest density of correspondences. On 4 July 2026, from a podcast, the name of Eleonora Matarrese arrived; buying and reading her volumes on Beinecke 408 (2023–2024) confirmed a convergence: the same area, by philological means and in full independence.

Two different paths. The same mountain. The convergence strengthens the direction, it does not found it: our identification was reached independently, before we knew her work.

Coming soon

The full transposition into a modern language

We are completing the transposition of Beinecke MS 408 into a clear modern language — in Italian and English — applying the Codex Mondragonis method folio by folio. The publication will include the full methodology, sources, statistical data and an annotated lexicon.

Write to be notified

The research

Diego «DMUX» De Maio and Simona «51m0» Fenoglio

The Codex Mondragonis method is developed independently, with no institutional academic affiliation. The method is documented, the data are public or originally produced, the sources are cited.

About us