programming model for superintelligence

two objects to program

every other system has one programming model. cyber has two:

token   (balance)  — what you HAVE
neuron  (prog)     — what you DO

semcons = physics (what CAN happen). progs = agency (what DOES happen). cyberlink = action (neuron spends tokens to create knowledge edge). this combination exists nowhere else.

cyberlink: action, not storage

every cyberlink moves value AND conveys meaning. always both. a > 0 always — free links do not exist:

cyberlink(ν, p, q, τ, a, v, t)
  ν = neuron, p = from, q = to, τ = token denomination,
  a = amount, v = valence (+1/-1), t = timestamp

τ is token, not conviction. conviction = derived from age of unspent cyberlink.

two scripts per cyberlink

every cyberlink validated by two nox programs:

  • lock script: WHO can act (neuron identity, multisig, timelock, delegation)
  • type script: WHAT rules (plumb conservation + semcon validation)

no plumb without semcon. no semcon without plumb. inseparable.

proofs compose, calls don't

no reentrancy. no call stack. no gas estimation. a signal can contain cyberlinks governed by different semcons — zheng proof validates all atomically. every application = semcon + optional prog. no contracts. no accounts.

wrote cyber/research/programming model: three levels (semcons validate, progs act, proofs compose), comparison table vs Ethereum/Solana/Cardano/Nervos.

BBG polynomial state

updated bbg everywhere: 13 NMT sub-roots → BBG_poly(index, key, t). one polynomial, 10 dimensions. no NMT, no LogUp. cross-index consistency structural. ~200 bytes per proof, 10-50 μs.

fixes across the stack

  • optica: found stray ``` that swallowed entire page into code block
  • hemera: remaining ~300→~736 in zheng, nox, cyb repos
  • .td→.tri: 77 replacements in bootstrap, 10 in stack, 1 in 16-languages
  • bbg: polynomial state in stack page + subgraph page
  • radio: hemera::hazmat → hemera::tree API rename

111 commits pushed across 15 repos. all working trees clean.