cyber/truth/inhibition.md

why the cybergraph without markets is not a functional model — and what markets provide that raw cyberlinks cannot

the missing half

every neural network has two kinds of weights: positive (excitatory) and negative (inhibitory). this is not an optimization detail. it is a structural requirement for discrimination.

a network with only positive weights can cluster — it can group similar things together. it cannot discriminate — it cannot say "this pattern excludes that one." without inhibition, a neural network cannot learn a boundary. it can only learn a blob.

the current cybergraph without market pricing is excitation-only. every cyberlink has a positive weight (stake amount). focus flows toward heavily-linked particles. nothing pushes back. the tri-kernel converges to π* — but π* is shaped only by positive association. it cannot represent "this edge actively misleads."

what the market provides

the market assigns each edge a price p(e) ∈ (0,1) — the ICBS market's consensus probability that the link is true/useful.

this price enters the tri-kernel as the effective edge weight:

$$w_{\text{eff}}(e) = \text{price}(e) \times \text{stake}(e)$$

now consider what different price regimes do:

at p → 0, the edge exists structurally but contributes nothing to π*. it is deactivated. this is the inhibitory signal that raw cyberlinks cannot provide.

the transformer parallel

from focus flow computation and graph-native-transformer: a transformer layer is one step of tri-kernel diffusion. attention weights are Boltzmann distributions over keys — they can suppress as well as amplify.

in a trained transformer, the compiled weights $W_Q, W_K$ encode both attraction (query-key alignment → high attention) and repulsion (misalignment → near-zero attention). the softmax normalizes across all keys, so amplifying some necessarily suppresses others.

in the cybergraph compiled transformer:

• without market weights: all edges compete equally weighted by raw stake. the softmax distributes attention proportional to structural connectivity only
• with market weights: edges with low market price are pre-suppressed before the softmax. the compiled transformer inherits the market's collective epistemic assessment as a prior on which edges deserve attention

the market provides what negative weights provide in a standard neural network: the signal that certain paths should not be followed, certain connections should not propagate focus.

the functional threshold

this means the cybergraph has two operational modes:

the transition from the first to the second is not a quantitative improvement. it is a qualitative one — the same transition as going from a network with only positive weights to one with both positive and negative weights.

a cybergraph without market prices can be a useful index. it produces cyberank proportional to structural prominence. this is valuable. but it cannot distinguish between a prominently-linked true claim and a prominently-linked false claim.

a cybergraph with market prices produces cyberank proportional to epistemic quality — structural prominence weighted by collective belief. it can suppress misleading links regardless of how many neurons created them.

social networks: the economic protection problem

social networks removed dislike buttons for a precise reason: coordinated attacks were free. a mob could suppress any content at zero cost. without skin in the game, negative signals are weapons, not information.

in the market, buying FALSE costs stake. attacking a link = injecting liquidity into the FALSE side of the ICBS market. two consequences:

1. the attacker takes on financial risk — if the market converges to TRUE, they lose stake
2. the attack improves the market's price accuracy — more liquidity = tighter spread = better signal

this inverts the social network dynamic entirely. attacking a true claim makes the true signal stronger. attacking a false claim makes the false signal stronger. either way, the market becomes more informative. the economic protection is not a feature — it is the mechanism by which the inhibitory signal remains honest.

two kinds of knowledge, one system

from two kinds of knowledge:

• structural knowledge (cyberlinks): "A relates to B" — permanent, individual, binary
• epistemic knowledge (market prices): "the network believes A→B with probability p" — dynamic, collective, continuous

market inhibition is the mechanism by which epistemic knowledge reshapes the structural layer's contribution to focus. the structure persists — the cyberlink is never deleted. but its weight in the tri-kernel reflects collective belief, not just individual assertion.

this is how mycelium operates: the hypha exists (binary). what flows through it depends on concentration gradients set by the whole network (continuous). the structural fact and the economic signal are separate and both necessary.

implication for the formal model

the tri-kernel operator $\mathcal{R} = \lambda_d D + \lambda_s S + \lambda_h H_\tau$ operates over the adjacency matrix A. the collective focus theorem proves convergence under ergodicity.

when market prices are incorporated:

$$A_{pq}^{\text{eff}} = \sum_{\ell: \text{src}(\ell)=p,\, \text{tgt}(\ell)=q} \text{price}(\ell) \cdot \text{stake}(\ell)$$

the convergence theorem still holds — $A^{\text{eff}}$ remains non-negative, satisfying all conditions. but the fixed point π* now reflects epistemic quality, not merely structural prominence. the market-weighted cybergraph and the raw cybergraph converge to different fixed points. only the former tracks truth.

see market for the market design. see coupling for the ICBS mechanism. see focus flow computation for how π* is computed. see two kinds of knowledge for the structural/epistemic distinction. see binary topology ternary economics for the architectural principle.