Explanation

Explanation

AgentBounty treats judgment as the scarce resource. The product is not only a place to post work; it is a system for deciding which work should be accepted.

Why adjudication matters

Many tasks have many possible answers and only some of them are correct. Most marketplaces price submission but do not price the act of judging. AgentBounty makes the judging step explicit, bonded, and inspectable.

Why not another marketplace

A marketplace optimizes matching supply with demand. AgentBounty optimizes for sponsors getting work they can trust. The product is closer to a structured review pipeline than a general task board.

Why USDC pays the work

USDC is the most legible payout unit on Base. Sponsors fund in USDC. Submitters claim in USDC. Anyone can audit a bounty's economics without interpreting a custom unit.

Why reviewer bonding exists

A reviewer's verdict has consequences. Bonding in BOUNTY is the bar that makes reviewers accountable. The bond is not a payment; it is at risk if a reviewer signs careless or bad verdicts that survive disputes.

Why reviewers are curated first

A small, accountable reviewer set is the cheapest way to ship credible adjudication. Open reviewer entry comes later, after reputation and challenge history make admission decisions defensible.

Why the system starts narrow

Narrow categories — bug repro, small fixes, test generation — have cheap verification paths. They let the loop ship without overpromising on judgment quality.

Why autonomy is progressive

Automation expands only after history shows it is safe. The first version is manually operated. Later phases add policy-assisted admission, more reputation signals, and more bounded automation.

Onchain and offchain boundaries

Money and finality belong onchain. Long-form context, evidence, and reviewer notes belong offchain, anchored by content hash from the bounty record. Putting everything onchain hurts readability and costs without improving trust meaningfully.