# Voynich Evidence Atlas — public v1.0

Selected public Atlas release authorized by Jon on 2026-10-19. Publication does not change scientific dispositions or restart research.

## How to check it

Open the evidence files linked from each story to inspect its cited text and numbers. public-evidence.json identifies exact copies, path-normalized copies, excerpts and references that remain private. Both source and copy hashes are retained. This provides selected inspectable evidence, not a complete reproduction package: the code, inputs and environment needed to rerun every experiment are not uniformly supplied.

Use records.json as the single record source. records.csv carries the same records with nested objects encoded as JSON cells. connections.csv derives from the record relationships. coverage.json/csv account for all 51 indexed findings, not all Voynich knowledge. Stable IDs and scientific dispositions are preserved. The manifest hashes the downloadable release files. Previous versions remain preserved separately.

## Changes from v1.1-evidence

Evidence-access instructions now reflect the accessible copies. Unsupported typed edges F-19 to F-01, P-87 to F-04 and c-586 to ATLAS-BRIDGES are removed. The distinction between the two repetition measures remains in explanatory text. P-87 corrects only part of F-40/F-41; F-20 withdraws a particular c-377 evidence line, not an entire finding. c-586 numbers and correction now reference its accessible claim excerpt, with the private audit retained separately. Prior-work credits name Currier, Smith and Ponzi, Zandbergen and Stolfi where relevant. F-01 explicitly includes its short-verse limitation. The frozen s132 threshold of 0.95 on each line is shown with a minimal protocol excerpt. A Yale composite scan link supplies folio-102 recto context, not scored regions or evidence for the verso panel.

## Corrections and limits

D-086 controls F-20: CORROBORATED; erroneous text-side s53 no longer counts, while separate image s53 is unaffected. P-87 changes the cross-line repetition comparison; F-04 within-line length rhythm remains distinct. Seven selected catalogue correspondences share physical folio 102, and the grouped proposal closed without scoring. Fontana s132 failed; Egerton s133 is process-invalid and unscored. db-S141 calibrated a Penn/Sloane graph rule, not Voynich identification: q99=3 and maximum=5. Historical evidence can contain superseded prose; the current story's correction and linked notice govern its use.

## Reuse and release

Third-party publications are credited and linked, not redistributed. Context images do not define a finding's sample scope. Names and links do not establish permission to redistribute source texts or images. Only the selected files in this manifest are published; the complete laboratory and private candidate remain private. Reading a file, loading it into a site, reproducing a study and interpreting its meaning are distinct accomplishments.

## Public copy transformations

Some result copies omit session clock times but retain calendar dates, run identifiers and measured values. These files are marked normalized in public-evidence.json, which retains each original source digest and public-copy digest. They are not byte-identical reproductions. No third-party full text, private conversation or blind-study response is included. No general redistribution license is granted for laboratory material.