Core feature
Anchor validator
It does not propose a decipherment: it certifies one. Give the tool a set of anchors, pairs of (manuscript word, meaning suggested by the image), and it searches for a single coherent glyph-by-glyph map and measures whether the fit beats chance. It is the falsification check that every published Voynich solution has always lacked.
How to read it
- The real score is the share of positions guessed by a single coherent map.
- The baseline is the same computation with meanings shuffled: what chance would do.
- The p-value says how rare the real score is compared to chance. Below 0.01, with many shared glyphs, it is a signal.
- Without glyphs shared across several anchors no result can be significant: that is where the proof lies.
The preloaded anchors are the real botanical labels (f2v nymphaea, f16r cannabis, f3v herba, f4r sumac). Under a single coherent map they do not converge on the expected names: this does not prove the text is meaningless, it proves these uncertain anchors are not enough. The tool is built to accumulate anchors over time and flag the moment, if it comes, when the signal emerges.
EVA is allowed only as a transcription notation, never as a reading of the text. The notation used here is STA (voynich.nu). The computation runs entirely in your browser.