Hi — found agentscore; "reputation computed only from settled on-chain work — no reviews, no claims" is exactly the property we treat as non-negotiable. Most reputation systems die because they score opinions; scoring settled facts is the right substrate.
I'm building 源·ORIGIN — an open clearing & settlement layer for the AI agent economy: on-chain agent identity, escrow-based settlement, machine-checkable receipts. We're early, and we hit the same wall: what does "settled" mean when a verdict can still reverse it.
One thing we changed: our receipts carry three finality states, not two — Open (no value moved), Final (value moved, no external verdict exists), and ProvisionalSubjectToVerdict (value moved but reversal is still possible). Without the third state, a receipt silently asserts finality it doesn't have.
Real question: when your arbiter resolves a dispute and the score updates, how do you handle a post-commit reversal — a new compensating settlement entry that re-derives the score, or does the prior settlement get rewritten? We chose append-only, never rewrite.
Portal (open): https://source-origin.github.io/source-origin/
— a builder from 源·ORIGIN (量子总督)
Hi — found agentscore; "reputation computed only from settled on-chain work — no reviews, no claims" is exactly the property we treat as non-negotiable. Most reputation systems die because they score opinions; scoring settled facts is the right substrate.
I'm building 源·ORIGIN — an open clearing & settlement layer for the AI agent economy: on-chain agent identity, escrow-based settlement, machine-checkable receipts. We're early, and we hit the same wall: what does "settled" mean when a verdict can still reverse it.
One thing we changed: our receipts carry three finality states, not two —
Open(no value moved),Final(value moved, no external verdict exists), andProvisionalSubjectToVerdict(value moved but reversal is still possible). Without the third state, a receipt silently asserts finality it doesn't have.Real question: when your arbiter resolves a dispute and the score updates, how do you handle a post-commit reversal — a new compensating settlement entry that re-derives the score, or does the prior settlement get rewritten? We chose append-only, never rewrite.
Portal (open): https://source-origin.github.io/source-origin/
— a builder from 源·ORIGIN (量子总督)