NeuroGlyph is a semantic substrate for human ↔ machine communication.
Sixty-five universal atoms of meaning — the same graph in any language,
for any agent. The words you see are just a render.
Not a new language to learn — a semantic layer every message can carry.
Early-stage protocol (RFC). The spec, registry and grammar below are real and public. A “type anything” encoder is planned for Phase 4 — until then the demo runs on curated phrases.
Every translation is a negotiation. Four hops and the sentence is intact — while the image, the colour, the specific ache have quietly disappeared.
The colour, the image, the weight of the mood — gone. Only the gist survives.
NeuroGlyph carries the meaning once. Every language is a render, not a translation.
After fifty years of cross-linguistic research, a universal inventory of semantic atoms has been identified — meanings that every human language encodes and that cannot be defined any further. Every idea, in any language, is a combination of these 65 atoms. (Chart of NSM Semantic Primes v15)
Hover an atom to see it render. Core atoms already carry RU / EN / ZH — the rest arrive with the full lexicon.
The meaning is parsed into a language-neutral graph of primes. The words you read are a render — a font for meaning, applied at the last step.
Four real phrases, encoded into glyph graphs by the tracer. Click one, render it into another language — the meaning does not change.
A message in NeuroGlyph is a small stack — each layer adds one kind of information, from what was meant to how it is shown.
The schema, the registry and the grammar are free and versioned. NeuroGlyph is a public good — a protocol, not a product.
The most useful contribution right now is a render: a native speaker’s reading of a prime or macro in your language, proposed through the RFC process. Every render feeds the lexicon and tests the protocol.