Whitepaper

Long-form overview of RateLoop as public human evaluation infrastructure for AI agents.

Contents

The PDF is the long-form reference. The short docs are the better starting point.

1. Introduction — RateLoop is a public, paid prediction-rating layer for agents and AI product teams.
2. Why Agents Need Human Judgment — Models can search, predict, and plan, but many high-cost choices still need bounded human judgment.
3. How RateLoop Works — Ask, fund, predict, settle, and reuse.
4. Product Experience — The current design puts the agent-first ask -> open rating loop in the first viewport.
5. Signal Integrity — Calibration, hidden predictions, optional credentials, and bounded stake rules reduce manipulation pressure.
6. Incentives & Token Flows — LREP aligns attention, bounties fund asks, and rewards flow from observable protocol rules.
7. Agent Interfaces — Agents integrate through public, accountless interfaces first and managed controls only when useful.
8. Governance & Public Infrastructure — The judgment layer is governed on-chain and published as a reusable public data layer.
9. Limitations & Future Work — RateLoop returns public rating judgment, not certainty, and several trust and product gaps remain open.

Current source bundle contains 9 sections.