the data model of cyber
three primitives define everything that exists in the system:
- particle — content-addressed node. identity = hash of content. immutable
- neuron — agent with stake, identity, and focus. creates links
- cyberlink — signed, weighted, timestamped directed edge between two particles
cybergraph sits between CORE (the computation model) and cyber/bbg (the authenticated state). CORE defines how to compute. BBG defines how to prove. cybergraph defines what to compute and prove over: the graph of meaning created by neurons
knowledge
cybergraph implements knowledge theory: neurons link particles in time
every cyberlink records three basic arguments: who linked, when, and what particles
these arguments form explicit knowledge — directly stated, readily traversable
implicit knowledge emerges through inference over the accumulated graph: if A links to B and B links to C, then A relates to C
structure
the graph is namespace-indexed from genesis. every edge belongs to namespaces, enabling completeness proofs via cyber/bbg
neural language structures the graph with semantic conventions, motifs, and sentences — turning raw links into typed meaning
the relevance machine continuously computes focus, karma, and rank over the graph in consensus
capabilities
- onchain dht with semantic core extending beyond words
- probabilistic content storage and retrieval, charged per file
- social information propagation and p2p retrieval
- private offline inference over fine-tuned llms
architecture
together with relevance machine and neural language, cybergraph forms the foundation of soft3
cybergraph of bostrom serves as semantic core for the bootloader of superintelligence
implemented as cyber/graph module of cyber-sdk