soft3/cybergraph/specs/intent.md

intent

an intent is an unsealed signal — a neuron's public declaration of a directed action before it is finalized and proven

structure

$$i \;=\; (\nu,\; h_0,\; \Sigma,\; \pi_\text{id})$$

field name type semantics
$\nu$ subject $N$ declaring neuron
$h_0$ inception $\mathbb{Z}_{\geq 0}$ block height at which the intent was declared
$\Sigma$ scope $S$ structured description of the intended action: target particle, predicate, deadline, constraints
$\pi_\text{id}$ identity proof $\text{Sig}$ neuron's signature over $(\nu \;\|\; h_0 \;\|\; \Sigma)$

an intent carries an identity proof (the neuron's signature over the declared scope) but no content STARK. the STARK is produced at sealing — it proves the neuron actually ran the scope. the intent commits to what will be computed; the signal proves it was.

lifecycle

declared
   │
   ├──→ sealed     — neuron finalizes: adds sealing height + zheng proof → signal
   │
   ├──→ abandoned  — intent never sealed; record persists in the graph
   │
   └──→ cascaded   — intent triggers coordinated sub-signals → cascade

sealed

a sealed intent is a signal: the neuron ran the scope, produced a zheng proof $\sigma$ over the execution (its cyberlinks and cyber/impulse), and committed it. seal is accepted only if $\sigma$ attests the declared scope — see completion below.

abandoned

if the neuron never seals, the intent record stays in the cybergraph at inception height. other neurons can observe that the action was declared but not followed through. abandonment is on record — there is no silent cancel

cascaded

a neuron may declare an intent with scope that invites participation: subscribers observe it, self-organize into coordinated sub-signals, and the lead neuron seals a parent signal with a recursive zheng proof over the entire cascade. see cascade for the full multiparty protocol

completion — how an intent becomes a complete signal

an intent is a deferred computation. its scope is not a passive description — it is an executable specification of what to compute. the intent commits to it; the signal proves it was run. completion fills the three things an intent lacks — the cyberlinks, the impulse $\Delta\phi^*$, and the proof $\sigma$ — by executing the scope and proving the execution:

declare    intend(scope)         scope_hash committed + signed — nothing run yet
execute    run the scope on nox  reading committed state via cybergraph query;
                                  iterate until it converges → cyberlinks + Δφ* (+ a trace)
prove      zheng proves the trace → σ
seal       seal(intent, signal)  commit the proven run

the orchestration — decide, collect, run, judge what is left, iterate — is the soma control loop, not cybergraph's. cybergraph stores the intent, serves the reads (query), and gates the seal. nox executes; zheng proves; soma drives.

the seal binding

seal is valid only if $\sigma$ proves an execution of the intent's declared scope:

$$\text{seal}(i, s)\ \text{accepted} \iff \sigma(s) \vdash \text{scope\_hash}(i)$$

this binding is what makes intent and signal one transaction rather than two unrelated records. the intent is a promise (committed scope); the signal is its proven fulfillment. a sealed signal is cryptographic proof the neuron did exactly what it declared — the alignment property, enforced at the boundary. without the binding, any signal could seal any intent.

relationship to signal

signal and intent share the same substrate. signal is what intent becomes:

intent   = (neuron, inception, scope, identity_proof)
signal   = (neuron, links, impulse, proof, inception, sealing)

the difference is completeness. an intent asserts what will happen; a signal proves it happened

identity proof

$\pi_\text{id} = \text{Sign}_\nu(\nu \;\|\; h_0 \;\|\; \text{hash}(\Sigma))$

this binds the neuron to the declared scope at the stated height. it is not a content proof — it does not cover computation correctness — but it is unforgeable and non-repudiable. any observer can verify that $\nu$ declared $\Sigma$ at $h_0$

scope

a scope describes the intended action in structured terms. minimum fields:

field semantics
target one or more particles this intent addresses
predicate the declared relationship or transformation
deadline latest acceptable sealing height (optional)
constraints any additional conditions the neuron self-imposes

scope encoding is dialect-specific; the intent mechanism is dialect-agnostic

discovery

intents at a given height are readable from the cybergraph by any neuron. this enables coordination without a central scheduler: subscribers discover pending intents, self-assign, and begin producing their sub-signals. the lead neuron observes incoming sub-signals and seals when ready

Homonyms

intent
an intent is a signal with inception but no sealing — a neuron's public declaration of a directed action before proof is produced | component | role | |-----------|------| | neuron | the declaring agent | | inception | block height when declared | | scope | structured description: target particles,…

Graph