Reppo scores $20M from Bolts Capital to turn prediction markets into AI training data
Decentralized prediction-market protocol Reppo has secured a $20 million strategic commitment from crypto investor Bolts Capital to scale its platform and expand “Datanets,” the project announced on April 23. The funding is explicitly framed as a long-term bet on Reppo’s vision: using staked human judgment in prediction markets to produce higher-quality, incentive-aligned data for AI training and evaluation.
Reppo’s pitch addresses a persistent bottleneck in AI development: poor or biased training data that limits model performance even as compute increases. Rather than relying on traditional labeling pipelines, Reppo aims to convert financially backed market bets into verifiable streams of labeled, time-stamped signals. By forcing participants to stake capital and be accountable for incorrect forecasts, the protocol seeks sharper probability estimates and richer behavioral data than standard surveys or annotation gigs can deliver.
The protocol supports multimodal inputs — text, images, audio, and video — and organizes them into decentralized “Datanets” that serve as continuous infrastructure for model training, evaluation, and fine-tuning. Reppo positions these Datanets as a bridge between crypto-native market design and AI infrastructure: a prediction-market layer that continuously emits scored, incentive-aligned datasets that AI teams can plug into their pipelines.
Bolts Capital’s commitment is described as strategic rather than pure balance-sheet financing, signaling confidence in the broader “prediction-market-as-data” thesis. The $20 million will be used to accelerate protocol development, roll out new prediction-market primitives, and build tooling to help AI teams integrate Reppo-generated data.
If Reppo can demonstrate that market-driven signals consistently outperform conventional labeling shops, Bolts’ wager could be an early sign that crypto prediction markets are evolving from speculative venues into foundational tooling for AI development. The coming months will show whether staked human judgment can truly become a reliable alternative for high-quality training and evaluation data.
Read more AI-generated news on: undefined/news
