About Brain-V

What is this?

Brain-V (VoynichMind) is a fork of Project Brain, a persistent cognitive architecture. It runs a continuous loop of perception, prediction, and scoring — the same loop that originally tracked AI research trends on arXiv, now retargeted at deciphering the Voynich Manuscript.

The Voynich Manuscript (Beinecke MS 408) is a 15th-century codex written in an unknown script that has resisted all decipherment attempts for over 600 years. Its vellum has been radiocarbon dated to 1404-1438.

Methodology

Brain-V does not claim to solve the manuscript. It runs a systematic, transparent process:

1. Perceive— parse the complete EVA transliteration (Zandbergen ZL3b, 38,053 words across 226 folios) and compute a statistical profile: glyph frequencies, entropy, Zipf's law fit, positional constraints, Currier A/B distribution.
2. Predict— generate testable hypotheses about the manuscript's cipher system, underlying language, or text structure. Each hypothesis must specify the exact statistical test that would confirm or deny it.
3. Score — run each test against the corpus statistics. Update hypothesis confidence. Eliminate those that fail. Promote those that pass. Log every result.
4. Learn — update the belief state based on scoring results. Generate new hypotheses informed by what was learned. Repeat.

Every cycle is logged. Every hypothesis — including failures — is published. Nothing is hidden. The on-chain provenance via AgentProof means the reasoning trail is verifiable and timestamped from day one.

What Brain-V is NOT

• Not a claim to have solved the Voynich. It is a tool for systematic exploration.
• Not a replacement for domain expertise. Codicological, paleographic, and historical analysis are outside its scope.
• Not infallible. It uses a local 8B parameter LLM for hypothesis generation. The quality of hypotheses depends on the model.
• Not closed. The community can submit hypotheses and Brain-V will test them.

Infrastructure

Contributing

Submit a hypothesis via GitHub Issues. Use the hypothesis template. Brain-V will test it against the corpus statistics and publish the results — pass or fail.

The most valuable contributions are hypotheses that would narrow the solution spaceif confirmed or denied. Vague theories ("it might be Latin") are less useful than precise, testable claims ("if the text is simple-substitution Latin, glyph entropy should be ~4.0 bits and word-final glyphs should match Latin suffixes").