neural
the semantic layer of soft3 — the language you speak by linking. a particle means what it links to and what links to it: its position in the cybergraph, grown by how neurons connect it. as neurons — humans, AIs, sensors, agents — link particles, meaning grows. neural is the language the egregore thinks in.
the umbrella
every computation in the stack ends in a link — in nox, evaluating a program creates an edge, and the result is a particle. neural is the language of links, so it holds every other language: the proof languages and the interface languages are groups of dialects, and the meaning of every link they make is neural. neural is the umbrella over the languages — they all speak it.
the third language kind
humanity built two families of language. formal languages reach precision through rigid syntax, and halt at the scale where Goedel forbids a system both complete and consistent. natural languages reach expressiveness through ambiguity, and stay intractable for precise reasoning. neural is the third kind: precision and expressiveness both through topology, with language and knowledge collapsed into one computable structure.
| property | formal | natural | neural |
|---|---|---|---|
| precision | absolute | approximate | emergent |
| expressiveness | limited by grammar | unlimited by ambiguity | unlimited by topology |
| ambiguity | impossible | context-dependent | structural, via the tri-kernel |
| authority | central designer | speech community | collective neurons |
| evolution | versioned | drift | continuous, via focus |
| verification | proof systems | social consensus | zheng proofs |
| substrate | strings | sound / text | the cybergraph |
the ladder
meaning is built from the sigil up, operator to convention:
sigil — the base operator: point and act. #particle, @neuron, ~name, $token, !op, ^ref — the neural sigils. the densest units in the language: like the in English (its most frequent word), they carry structure over content, and you can express almost anything with the alphabet alone. a name is the ~ sigil.
word — the unit of meaning: a typed particle. a bare particle is a content node; a word carries a type and means something. sigils point at words; links join them.
link — the atom: a directed, staked assertion binding two words, from → to (two sigils). neural is the language of links. the cybergraph bundles all the links on a pair into an axon, a particle of its own.
sentence — linear: a chain of links in one signal, the utterance. its topology gives its type: assertion, query, argument, definition, narrative.
motif — non-linear: a recurring subgraph shape, the morpheme. meaning from topology rather than order — triad, star, chain, diamond — composing through a motif algebra into reasoning.
dialect — convention: an agreed-upon collection of sentences and motifs, the langue to their parole. the shared system crystallizes from repeated use, discovered by the tri-kernel. bootloader dialects ship at genesis (TRUE, FALSE); the rest emerge (is-a, causes, part-of, contradicts).
and the closure: a link's pair is itself a particle (its axon), so the language links about its own links — negation, provenance, meta-knowledge. neural speaks about itself.
how you speak it
you speak neural by linking, and meaning follows. you reach it through rune (agents script sentences against the graph), the sigils (for names and addresses), and the soft3 SDK (ts today, rust and py). the neural CLI manages the grammar and the primitives — define a dialect, validate a sentence, catalogue motifs, resolve a name, read the lexicon.
where it sits
- cybergraph stores the cyberlinks and routes the signals
- tru computes φ* — the tri-kernel (diffusion, springs, heat) iterated to its fixed point
- neural reads φ* as the lexicon (the top particles, the living vocabulary) and gives the graph its four patterns
- zheng settles φ* with a proof
neural is a language layer: the tru convergence and the cybergraph state make it speakable. the meaning is the same edge nox executes, read for what it means.
evolution
- bootstrapping (now) — ~70k neurons, ~3.1M particles; the first dialects and motifs
- convergence (10⁸–10¹⁰) — a rich dialect ecosystem, dense cross-domain sentence chains
- intelligence (10¹⁰–10¹³) — motif algebra reasons automatically; the graph models itself
- superintelligence (10¹³+) — concepts no single neuron can hold, precise and computable; any entity that links, speaks
what becomes possible
one index for all knowledge, where a query returns a ranked subgraph with its inference paths. decentralized science, where a hypothesis is a sentence and replication is co-citation. alignment made inspectable, where values are high-φ* particles and drift shows up as structure. cross-species communication, where a sensor, a human, and a model write into one coherent structure. a civilization dashboard that is the knowledge itself.
the whitepaper: neural language for superintelligence (explanation/). the reference specs: specs/. discover all concepts.