I modeled the Voynich Manuscript with SBERT to test for structure
I built this project as a way to learn more about NLP by applying it to something weird and unsolved. The Voynich Manuscript is a 15th-century book written in an unknown script. No one’s been able to translate it, and many think it’s a hoax, a cipher, or a constructed language. I wasn’t trying to decode it — I just wanted to see: does it behave like a structured language? I stripped a handful of common suffix-like endings (aiin, dy, etc.) to isolate what looked like root forms. I know that’s a strong assumption — I call it out directly in the repo — but it helped clarify the clustering. From…
In plain words
This project applies natural language processing to analyze the Voynich Manuscript, a 15th-century text written in an unknown script. Rather than attempting translation or decryption, it uses SBERT embeddings and clustering to examine whether the manuscript exhibits structural properties of a language. The creator extracted potential root forms by removing suffix-like patterns, then used KMeans clustering and Markov transition matrices to model word relationships and flow. Built as a learning exercise, it explores linguistic patterns in a historically mysterious document without claiming to decode it.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
I built this project as a way to learn more about NLP by applying it to something weird and unsolved. The Voynich Manuscript is a 15th-century book written in an unknown script. No one’s been able to translate it, and many think it’s a hoax, a cipher, or a constructed language. I wasn’t trying to decode it — I just wanted to see: does it behave like a structured language? I stripped a handful of common suffix-like endings (aiin, dy, etc.) to isolate what looked like root forms. I know that’s a strong assumption — I call it out directly in the repo — but it helped clarify the clustering. From there, I used SBERT embeddings and KMeans to group similar roots, inferred POS-like roles based on position and frequency, and built a Markov transition matrix to visualize cluster-to-cluster flow. It’s not translation. It’s not decryption. It’s structural modeling — and it revealed some surprisingly consistent syntax across the manuscript, especially when broken out by section (Botanical, Biological, etc.). GitHub repo: https://github.com/brianmg/voynich-nlp-analysis Write-up: https://brig90.substack.com/p/modeling-the-voynich-manuscrip... I’m new to the NLP space, so I’m sure there are things I got wrong — but I’d love feedback from people who’ve worked with structured language modeling or weird edge cases like this.
Does the same job
all alternatives →- AKA Karpathy-style LLM wiki your agents maintain (Markdown and Git)Apr 2026 · github.com · ▲260
I shipped a wiki layer for AI agents that uses markdown + git as the source of truth, with a bleve (BM25) + SQLite index on top. No vector or graph db yet. It runs locally in ~/.wuphf/wiki/ and you can git clone it out if you want to take your knowledge with you. The shape is the one Karpathy has been circling for a while: an LLM-native knowledge substrate that agents both read from and write into, so context compounds across sessions rather than getting re-pasted every morning. Most implementations of that idea land on Postgres, pgvector, Neo4j, Kafka, and a dashboard. I…
- IWI wrote a technical history book on LispFeb 2026 · berksoft.ca · ▲252
The book page links to a blog post that explains how I got about it (and has a link to sample content), but the TL&DR is that I could not find a lot of books that were on "our" history _and_ were larded with technical details. So I set about writing one, and some five years later I'm happy to share the result. I think it's one of the few "computer history" books that has tons of code, but correct me if I'm wrong (I wrote this both to tell a story and to learn :-)). My favorite languages are Smalltalk and Lisp, but as an Emacs user, I've been using the latter for much longer and for my…
- CAChonky – a neural approach for text semantic chunking2025 · github.com · ▲169
TLDR: I’ve made a transformer model and a wrapper library that segments text into meaningful semantic chunks. The current text splitting approaches rely on heuristics (although one can use neural embedder to group semantically related sentences). I propose a fully neural approach to semantic chunking. I took the base distilbert model and trained it on a bookcorpus to split concatenated text paragraphs into original paragraphs. Basically it’s a token classification task. Model fine-tuning took day and a half on a 2x1080ti. The library could be used as a text splitter module in a RAG system or…
- TNTurn native language audio into flashcards and shadowing practiceJun 2026 · lingochunk.com · ▲92
Here is a tool I built initially for myself to help with my German and Greek language studies. It started as a hack for creating Anki cards from native language audio. It extracts the words, finds their base forms (lemmas) and groups the examples by the lemma. At some point I realised that I have a transcription with word level timestamps that opens a lot of other opportunities. So I added a mode to click the first and last word in the transcript and it starts looping with the right gap and repeat count. Another feature I use a lot is selecting an audio fragment, sending a predefined prompt…
- HLHow LLMs Work – Interactive visual guide based on Karpathy's lectureApr 2026 · ynarwal.github.io · ▲245
All content is based on Andrej Karpathy's "Intro to Large Language Models" lecture (youtube.com/watch?v=7xTGNNLPyMI). I downloaded the transcript and used Claude Code to generate the entire interactive site from it — single HTML file. I find it useful to revisit this content time to time.
- WFWe fingerprinted 178 AI models' writing styles and similarity clustersApr 2026 · rival.tips · ▲78
We have a dataset of 3,095 standardized AI responses across 43 prompts. From each response, we extract a 32-dimension stylometric fingerprint (lexical richness, sentence structure, punctuation habits, formatting patterns, discourse markers). Some findings: - 9 clone clusters (>90% cosine similarity on z-normalized feature vectors) - Mistral Large 2 and Large 3 2512 score 84.8% on a composite metric combining 5 independent signals - Gemini 2.5 Flash Lite writes 78% like Claude 3 Opus. Costs 185x less - Meta has the strongest provider "house style" (37.5x distinctiveness ratio) - "Satirical…
More work this month
the category →
The app store for voice native apps that lives in your notch
Work · 28d ago · voiceos.com
The New Calendly▲211Handle all of the work before, during, and after meetings
Work · 17d ago · calendly.com
Launched alongside, May 2025
the whole month →
- C9
Life & fun · 2025 · felixrieseberg.github.io
