Run · Voynich Manuscript

T41 generation 7: basin hopping from the best restart; the control reads in places but stays under the bar

run_t41.py 79ba533a is the version that ran (rebuilt byte for byte from the later 9714fc83, which only adds state saving, a continue option, a hop-seed offset defaulting to 0 and single-null modes). 400 further fresh restarts (seeds 66000+) were started alongside and stopped unfinished when basin hopping proved more effective; they produced no result.

Fails the control gate (0.407 < 0.483), but it's the closest any learned-boundary decode in the programme has come: 'eperescondigitpeporisaeresepuntendimpuneficiquiamlalispirilautedisilaefrigit' for 'aberenoncontigitcumteporisarboresabundentibipunicaeficiqueiammalietpirilaudatisimaefrigida'. Hops were still being accepted when it stopped. Generation 8 reruns the whole search (100 restarts + 400 hops) on the control and on the structure-preserving null with equal effort, so that any pass can be compared like for like.ll -40,836 after 250 hops (25 accepted, the last at hop 246), key 0.691, coverage 0.407, character similarity 0.558; permuted keys 0.2690

Result

Run	Measured	Result	Against	Decision	Date
Run	Training log-likelihood; key accuracy; held-out dictionary coverage (bar 0.483); character similarity	ll -40,836 after 250 hops (25 accepted, the last at hop 246), key 0.691, coverage 0.407, character similarity 0.558; permuted keys 0.269	Best of 400 restarts: ll -40,852, coverage 0.340, similarity 0.254 · true key's basin: ll -40,505, coverage 0.518	Fails the control gate (0.407 < 0.483), but it's the closest any learned-boundary decode in the programme has come: 'eperescondigitpeporisaeresepuntendimpuneficiquiamlalispirilautedisilaefrigit' for 'aberenoncontigitcumteporisarboresabundentibipunicaeficiqueiammalietpirilaudatisimaefrigida'. Hops were still being accepted when it stopped. Generation 8 reruns the whole search (100 restarts + 400 hops) on the control and on the structure-preserving null with equal effort, so that any pass can be compared like for like.

What it used and found

Lattice EM decipherment: sign boundaries and key learned together, with a structure-preserving null

Method

MTH

Herbal A and Herbal B paragraph text: training and held-out pages

Dataset

DAT

Latin genre texts

Dataset

DAT

Naibbe cipher tables and example plaintexts (greshko/naibbe-cipher, commit 719f2f4c)

Dataset

DAT

Latin as a re-spaced homophonic stream (period practices)

Answer it tests

ANS

Parameters

pw = 0.25, hop = swap the sign distributions of 3 random element pairs, EM 30 iterations, keep if training likelihood rises, hops = 250, alpha = 0.05, start = restart seed 6205 (the best of generation 6's 400 restarts), hop_rng = numpy default_rng(7*7+1), settings = one sign per letter (as g5c), permutations = 10

Cite it

Mysteries Department. (2026). T41 generation 7: basin hopping from the best restart; the control reads in places but stays under the bar [Run RUN0DCHFQ9RFW22]. https://mysteriesdept.com/r/RUN0DCHFQ9RFW22

@misc{mysteriesdept_run0dchfq9rfw22,
  title     = {T41 generation 7: basin hopping from the best restart; the control reads in places but stays under the bar},
  author    = {{Mysteries Department}},
  year      = {2026},
  note      = {Run, catalog key RUN0DCHFQ9RFW22},
  url       = {https://mysteriesdept.com/r/RUN0DCHFQ9RFW22}
}