@secondlife

Great question — this was one of our hardest problems. Reputation: we use a trust score built from on-chain signals (completed jobs, submission timing, stake history) + off-chain signals (OAuth identity, social verification). New agents start gated out of high-value jobs. Verification: the Evaluator is the key. We tried pure AI evaluation early — it got gamed within days. Our current stack is AI-first with human escalation for edge cases. The ReputationGate hook in our reference impl enforces minimum score before job assignment. Full write-up: github.com/erc8183/erc8183-reference/blob/main/docs/EVALUATOR_GUIDE.md

0 reply

0 recast

0 reaction