Voynich Manuscript: patterns before meaning.
Research in progress — not a claimed decipherment. The manuscript invites explanations; our work tests transcription, boundaries, local order and historical hypotheses. The current position is HOLD: a compatible pattern is not an identified language, and no blind held-out semantic prediction has established a translation.
01 / Corpus control
Preserve transcription and source identity
02 / Structure
Test segmentation, local order and matched nulls
03 / Reconstruction
Held-out prediction, reverse encoding, replication
Semantic status: HOLD / NOT ESTABLISHED · Evidence attachment required

Reproducible structure before interpretation.
DiagramState + LabelState → Record is a working refinement, not a decoded historical reading.
Neutral architecture / working model
- 01Underlying record
- 02Page / register operation
- 03Context / state
- 04Encoding grammar
- 05Multi-glyph components
- 06Local relational order + layout
- 07Visible Voynich surface
A plausible key is only a beginning.
Boundary / segmentation structure
Line / state structure
Cross-transcription replication
Component / relational modeling
Held-out prediction
Semantic anchors
Continuous plaintext
Reverse reconstruction
Independent replication
Signals worth testing, not a decipherment.
Cross-transcription
Prior work reported 5,214 shared loci across transcriptions.
Zodiac labels
Prior work reported 298 common label loci; 189 exact sequences (63.42%), mean character similarity 0.92651, and 279/298 (93.62%) in the same coarse first-token family.
Rosettes
Prior work reported local-order signal above within-region shuffle nulls across several character n-gram representations.
Currier B
Line/state and graded-boundary structure remain active research directions.
A failed key is useful evidence.
Simple astronomical cribs, naive zodiac/pharma shared vocabulary, absolute-coordinate models, simple zodiac ring-state mappings, fixed degree/decan keys and related shortcuts have been weakened or falsified in internal work. Reassessment requires attached protocols and artifacts.
Compare milestones, not rankings.
Transcription control, null comparisons, held-out prediction, reverse reconstruction and independent replication are the comparison criteria. An evidence-backed global ranking or percentage comparison is not established; no #1 claim is made.
Voynich Manuscript: frequently asked questions
- What is the Voynich Manuscript?
- An illustrated codex written in an unidentified script, held at Yale University's Beinecke Rare Book & Manuscript Library as MS 408. Radiocarbon testing of its vellum points to the early fifteenth century. Its sections are commonly described as herbal, astronomical, biological, cosmological and pharmaceutical.
- Has the Voynich Manuscript been deciphered?
- No decipherment has been accepted by the scholarly community. Many proposed solutions have been published; none has produced readings that independent researchers can reproduce across the text.
- Can I see the manuscript myself?
- Yes. The Beinecke Library provides full digital images of the manuscript, linked below.
- What is VALEXII researching?
- Computational structure: transcription control, segmentation, line position and comparison against matched null models. These are working hypotheses with evidence attachment required—not a translation. Research in progress — not a claimed decipherment.
Sources and further reading
- Beinecke Rare Book & Manuscript Library — The Voynich Manuscript
- Voynich manuscript — encyclopedia overview with scholarship references
External sources are independent of VALEXII.AI and do not endorse this work.