cyber/cybergraph contract.md

the cybergraph contract

what cyber demands of its one data structure — the commitments every implementation must honor, stated at protocol altitude. the idea lives at cybergraph; the full mathematics at cybergraph spec; the reference machine at soft3/cybergraph; the query surface at inf cybergraph. this page is the interface between them

the object

a cybergraph is a triple $\mathbb{G} = (P, N, L)$: content-addressed particles, authenticated neurons, and a multiset of staked cyberlinks — with tokens and karma derived from $L$, never primitive. everything else the protocol does is a reading of this triple

six commitments

the axioms are the protocol's promises. break one and the thing running is not a cybergraph:

commitment what it buys
A1 content-addressing — identity equals content, $H$ collision-resistant no one can quietly edit the record; the same knowledge is the same particle everywhere
A2 authentication — every signal carries its neuron's valid signature every claim has an author; there is no anonymous write path to bypass staking
A3 append-only — $L$ grows monotonically; only economic weight decays history cannot be rewritten, only outweighed — forgetting is economic, not destructive
A4 entry — a particle exists iff it is linked no orphan data: the graph contains exactly what someone paid to connect
A5 conservation — $\sum_p \phi^*_p = 1$ at every block focus is a budget, not a score: attention gained here is attention lost elsewhere
A6 homoiconicity — every edge hashes into a particle (axon) the graph can rank, link, and reason about its own structure

A5 is the bridge clause: the graph promises a conserved $\phi^*$ exists, but computing it is not the graph's job — that boundary is next

sharp boundaries

the contract works because each organ owns exactly one thing and the seams are explicit:

organ owns must not
cybergraph the triple, adjacency, the signal lifecycle compute focus, store bytes, decide anything
tru the dynamics — tri-kernel to $\phi^*$, syntropy, impulse define the graph it reads
bbg authenticated storage and commitments interpret what it stores
foculus finality — $\phi^*_i > \tau$ rank or vote
zheng proofs gating every commit trust anyone
soma deciding and executing — the smart half write unproven results

adjacency is where economics enters the mathematics: raw weight from stake, effective weight $A^{\text{eff}}$ after karma and market inhibition multiply it. the graph hands tru a matrix that already encodes who is trusted and what the market believes

one structure, the whole protocol

identity, key exchange, consensus, fork choice, finality, privacy, incentives, version control, file system, type system, computation, data availability, sybil resistance — fifteen protocol functions run through these five primitives with no second data structure anywhere. the full table with per-function mechanisms is in cybergraph spec §the graph is the protocol

this is the deepest design bet in cyber: a protocol with one structure cannot disagree with itself. there is no state that is not graph, so there is nothing to reconcile

see the whitepaper §3 for the narrative form · cybergraph spec for axioms with full mathematics · soft3/cybergraph for the machine that enforces them · inf cybergraph for reading the result

discover all concepts

Graph