Stake-Assured Human Feedback

Reppo uses a custom veToken mechanism that lets researchers and AI/ML teams capture the strength of preferences in training data, backed by real economic stake. It extends traditional vote-escrowed tokenomics (veTokenomics) beyond governance and applies it directly to AI training data.

How It Works

1. Lock REPPO → Receive veREPPO

   • Voters, who are also data annotators, lock $REPPO tokens for a chosen length.

   • In return, they receive veREPPO, i.e. voting power.

   • The amount of veREPPO is proportional to both the quantity of tokens locked and the lock duration. Longer locks grant disproportionately higher voting power.

2. Vote Each Epoch

   • veREPPO holders vote each epoch to predict which collabs, or AI content published by creators, will receive the most support.

   • Voting power decays linearly within the epoch, so earlier votes carry more weight than later ones.

   • At the end of each epoch, net new $REPPO emissions are split 50/50 between creators who received votes and the voters who backed them.

   • This creates a prediction market around AI content while also crowdsourcing AI training data.

3. Adjust Votes, Not the Lock

   • Voters can adjust allocations every epoch, but their locked REPPO remains illiquid for the chosen duration.

   • This balances long-term capital stability with short-term governance flexibility.

Key Properties

• Alignment of incentives: Token holders are rewarded for surfacing the highest-quality content and publishers.

• Sticky economics: veREPPO creates persistent relationships between voters and creators, reducing short-term churn.

• Anti-farming mechanisms: Locked stake and epoch-based participation make short-term reward farming harder to sustain.