Primary technical disclosure
Secondary technical
The Valuation-Free Ordinal Standing Quantity An ordinal standing multiplier for governed semantic agents, computed from direction of initiation, latency to reciprocate, and regularity of recorded exchange, with no cardinal magnitude of benefit computed, solicited, or stored.Determination-Invariance Under an Agent's Own Policy Succession How a persistent semantic agent earns trust in a counterparty's recorded determinations by re-resolving them across its own signed policy object succession, producing a policy-invariant set and an invariance count without any rating, score, or ground truth.Elective Exposure Enrollment Without Deference Elective exposure enrollment lets a semantic agent enroll another agent's conformance-filtered corrector classes at its own authorization-budget cost, without deference, without adopting any determination, and without decrementing the second agent's budget.Interventional Attribution by Receipt-Path Control Interventional attribution by receipt-path control separates counterparty-caused confirmation latency from domain-caused latency by routing matched interactions through counterparty-controlled or non-counterparty-controlled receipts under a pre-committed, verifiable closure-path assignment.The Portable Exoneration Record The portable exoneration record is a signed, time-bounded governed observation that lets a semantic agent carry a verifiable record that its confirmation latency was attributed to its domain, not to it, into first-encounter determinations with new agents.Attribution, Exoneration, and the Domain Baseline: An Ordered Branch Over the Latency Differential The attribution-versus-exoneration branch evaluates a latency differential in fixed order: a per-path minimum volume gate returning a not-determinable outcome, a threshold test attributing latency to the counterparty, and a second-path domain-baseline elevation test that attributes latency to the domain and appends a portable exoneration record.Bounded Count of Dispatches Per Deference Record: Capping How Many Executions One Counter-Signed Adopted Determination Authorizes A bounded count of dispatches declared in a deference record caps how many executions one counter-signed adopted determination authorizes: each dispatch is appended with a reference to that record, exhaustion of the count ends the record's satisfaction of the dispatch-authority predicate, and further deference requires a fresh counter-signature and a fresh decrement of the followed agent's authorization budget.Coarse Lineage-Segment Granularity: Verifying a Tested-Entry Enumeration Without Disclosing Which Lineage Entries Were Tested How a followed semantic agent presents its tested-entry enumeration at a declared lineage-segment granularity, disclosing that a segment of its append-only lineage field was tested against a conduct evaluation artifact without revealing which entries were tested, while the retrievability-form verification runs against segment identifiers and the deference record records the granularity applied.Correction-Exposure Decay: An Authorization Quantity That Depletes With Agent Operation and Renews Only on Recorded External Evaluation How correction-exposure decay meters an autonomous agent's per-action-class authorization quantity against recorded external evaluation, depleting it on executions and successor epochs, setting the depletion rate from complied corrective encounters, and withholding the authorization gate at a policy-declared floor.Cause-Keyed Gate Restoration: A Distinct Return Path for Decay-Induced Withholding of Agent Authorization How a governed autonomous agent restores a decayed authorization gate through exposure to a previously uncounted origin-equivalence class, while harm-sourced withholding stays locked to its own repair path.Edge Severance Events and the Three-Valued Survival State: Resolving Whether a Counterparty Relationship Outlived a Payment-Class or Obligation-Class Termination How a governed semantic agent records payment-class and obligation-class severance events against a counterparty edge and computes a three-valued severance-survival state of not-tested, survived, or not-survived over an observed interval, structurally and never from message content or a party's declaration.Emitting the Deference Correction Linkage Record to the Followed Agent: Reporting a Downstream Correction Without Producing a Rating How a deferring agent emits a deference correction linkage record to the agent it followed, so the followed agent records that an adopter was later corrected on the adopted matter while a conversion bar forecloses any rating, score, scalar, or threshold operand.The No-Address Omission Branch: What a Deferring Agent Records When an Exposure Dispatch Reaches Nobody How a deferring semantic agent records an exposure dispatch it cannot address: when a class-defining reference enumerates no identity primitive for which a dispatch address is held, the omission is appended to the append-only lineage field and the origin-equivalence class stays enrolled as an admissible-but-unexercised entry.Channel-Baseline Normalization: Taking Channel Overhead Out of the Latency Differential Channel-baseline normalization derives a latency differential from two closure paths by subtracting a pooled channel baseline built from qualifying first-path closures in a semantic agent's own records, so confirmation latency caused by a counterparty is separated from latency caused by its domain without consulting any counterparty-supplied score.Reason-Type Resolution From Severance-Survival States: Typing an Agent's Relationship Edges Without Asking the Counterparty How a persistent semantic agent resolves an edge's reason-type from a pair of severance-survival states, by an ordered procedure yielding payment-sustained, obligation-sustained, want-sustained, or not-typeable, computed structurally from recorded severance events and a declared continuation interval rather than from any standing declared by the asserting party.Reason-Type Retyping by a Superseding-Type Entry: Recomputing an Edge's Reason-Type on Later Severance Evidence Without Disturbing Modifications Already Applied How a governed semantic agent revises the reason-type of a counterparty edge after a later-recorded severance event, appending a superseding-type entry that reweights only modifications asserted afterward while leaving every applied scoped integrity vector and self-esteem aggregate modification undisturbed.Renewal-Side Replenishment of the Authorization Budget: Restoring an Agent's Metered Spending Capacity Only on Exposure to an Uncounted Origin-Equivalence Class Renewal-side replenishment ties an autonomous agent's authorization budget to external exposure: an admitted conduct evaluation artifact from an uncounted origin-equivalence class replenishes it, a declared authorization budget ceiling bounds it above, and elapsed time, further issuance, and self-assessment replenish nothing.Following Does Not Count as Being Tested: Retention of the Short Decay Constant When an Agent Defers, and the Action Class That Stays Untested How a deference record advances no compliance count and records no corrective encounter, so an autonomous agent's action class stays untested and its authorization quantity keeps depleting at the short decay constant declared in policy.Standing-Determination Designation: Substituting a Verified Counterparty Determination for One Conjunct of the Dispatch-Authority Predicate How a persistent semantic agent computes determination invariance over a counterparty's retrieved determinations, writes a standing-determination designation into its counterparty identity record, and thereby makes one conjunct of its dispatch-authority predicate satisfiable by a verified counterparty determination while the policy, lineage, and authorization stamps remain non-substitutable.Uncounted-Class Renewal: Gating Agent Authorization on a Windowed Register of Origin-Equivalence Classes How a windowed register of origin-equivalence classes gates renewal of an autonomous agent's decaying authorization quantity, keeping renewal valence independent, scoping it per action class and per scope partition, and foreclosing renewal manufactured by repeated evaluation from a single source.
Applications · general
Vendor Qualification Without Ratings: Standing Earned From Recorded Conduct How vendor qualification and supplier onboarding can work without transferable risk scores: a deference record counter-signed at the followed party's own cost, enrolling the origin-equivalence classes that corrected it, so exposure is earned rather than presented.Credentialing Autonomous Practitioners: Deference That Survives Across Agents How autonomous practitioner agents can earn professional standing through recorded deference and corrective exposure rather than presented credentials, using counter-signature metering, conformance-filtered corrector enrollment, and no transferable score.Coalition Operations: Standing Without a Shared Authority How coalition and joint operations can let automated participants rely on determinations made in other organizations without a shared trust registry, using recorded deference, counter-signature, and metered exposure earned through conduct.When Standing Is Bought Instead of Earned How a newly deployed autonomous agent obtains admissible corrective feedback without buying reputation: origin-disjointness, counter-signed deference, merit-independent budget decrement, and corrector-class enrollment in the disclosed architecture.
Applications · specific
Marketplace Seller Reputation Systems: Presented Scores Versus Earned Admissibility How marketplace seller reputation systems and an earned-exposure conduct architecture answer the same stranger problem differently: operator-computed seller scores versus deference records, merit-independent counter-signature metering, and admissibility earned from recorded correction.Letta: Agent Memory and Standing Earned From Conduct How Letta's publicly described stateful agent memory compares with a filed deference and earned-exposure architecture in which admissibility rather than reputation is earned through recorded conduct, and a counter-signature costs the followed agent its own dispatch budget.Trustpilot Signals and Conduct-Derived Admissibility How Trustpilot's publicly described consumer review and rating model compares with a filed architecture in which agents earn admissibility from recorded conduct, with no rating, score, or standing quantity ever passing between them.Copilot Coding Agents: Review Gates and Earned Exposure How the GitHub Copilot coding agent's pull request review gates compare with an earned exposure architecture in which deference costs the followed agent authorization budget and no rating, score, or standing quantity is ever transferred between agents.Supplier Risk Scores and Recorded Counterparty Conduct How supplier risk scoring services such as Dun & Bradstreet compare with a filed architecture for earned exposure, deference records, counter-signature, and recorded counterparty conduct in third-party risk, vendor verification, and agent trust.Verified Profiles and Standing Built From Recorded Conduct How LinkedIn verification, as publicly described, and an earned-exposure conduct architecture address different halves of the same trust problem: presented professional identity signals alongside deference records, merit-independent counter-signature metering, and admissibility earned from recorded correction.
Terminology
counterparty identity record a substrate-resident record of an encountered party, comprising at least one identity primitive..deference record an entry appended by a deferring semantic agent binding at least the identity primitive of the followed agent and the followed determination relied upon, and transmitted to that agent for counter-signature..edge a persisted dyadic relation between the semantic agent and one counterparty, constituted by that counterparty's counterparty identity record together with each lineage entry recording an interaction with that counterparty..reason-type a classification of an edge as obligation-sustained, payment-sustained, want-sustained, or not-typeable, resolved by a severance-survival test over recorded severance events..severance event an entry of the append-only lineage field recording a structural change terminating the capacity of the semantic agent and a counterparty to interact, comprising at least a severance class and a recorded time..standing quantity an ordinal quantity computed from a direction of initiation, a latency to reciprocate, and a regularity of recorded interactions.untested class as applied to an origin-equivalence class, a class every constituent edge of which resolves not-typeable, and, as applied to an action class, a class for which no corrective encounter with subsequent recorded compliance is recorded..
