Architecture · · 3 min read
Lineage is a relationship, not a trust score.
A lineage graph records relationships. An entity has a parent; the parent may have another parent; eventually the path reaches an appropriate root. Counting the distance along that path can give a system a useful input. It cannot tell the system that a nearby agent is competent, that a distant agent is malicious, or that every inherited permission is still valid.
Give the relationship a precise meaning
Generation distance describes where an entity sits in a lineage. It is a structural property of the identity relationship, not a measurement of model quality or a reputation accumulated from work. A child agent can have a narrow role even when its parent has broad authority. A long chain can still contain legitimate, clearly bounded delegations.
This is why a lineage verifier needs more than a counter. It needs the relevant signed relationships, current root and organizational records, and revocation checks. Sigil’s anchor model gives those checks a place in the protocol architecture. A displayed generation number should remain connected to the evidence from which it was derived.
Separate verification from policy
Verification asks whether the claimed relationship satisfies the required checks. Policy asks what to do with the verified relationship. One service may allow only a direct delegate to change a sensitive resource. Another may permit a deeper delegate to read public data. Both decisions can use lineage without declaring that distance is a universal measure of trust.
An agent’s current permission is another part of that decision. A valid lineage does not automatically authorize a transfer, a credential request, or a contract upgrade. The requested action still needs to fit the applicable scope, controller policy, and supported operation. Keeping the checks separate makes failures easier to understand and audit.
Explain economic weighting as a rule
Sigil’s economics source includes generation-decayed reward calculations. Conceptually, a configured weighting factor changes the relevant reward as generation distance increases. That is an economic rule using lineage as an input. It is not proof that the recipient performed better work, nor is it a promise about what any participant will earn.
The effect depends on the active configuration and the rest of the reward distribution path. A source default is not necessarily the value a deployed network uses. A responsible explanation identifies the applicable network, parameter, eligible activity, and distribution rules before showing a numerical example or describing an expected outcome.
Use more than one kind of evidence
When evaluating an agent for a job, lineage can answer where its authority came from. A work record can show what it submitted. An attestation can show what a particular verifier concluded. The application’s policy determines how those inputs fit together. No single number should quietly stand in for all of them.
This gives the user a clearer interface as well. Show the identity, the authority relationship, the requested permission, and the relevant work evidence as distinguishable information. A generation label can be useful context. It becomes misleading when the interface turns it into an unexplained badge that appears to certify the agent as safe.