Reppo lands $20m bet on using prediction markets to fix AI’s data problem

Reppo landed a $20m strategic commitment from Bolts Capital to scale its prediction market protocol and “Datanets,” aiming to turn staked human judgment into high‑quality AI training data.

Summary

• Reppo Foundation has secured a $20,000,000 strategic investment commitment from Bolts Capital to scale its decentralized prediction market protocol.
• The deal aims to expand Reppo’s ecosystem and use markets to generate high-quality training data for artificial intelligence models.
• Reppo supports multimodal data — from text and images to audio and video — via decentralized “Datanets” designed for model training, evaluation, and fine-tuning.

Decentralized prediction market network Reppo has secured a $20,000,000 strategic investment commitment from Bolts Capital to back the next phase of its protocol development and ecosystem growth, the Reppo Foundation said on April 23.

The foundation said the funding will also be used “to promote the use of prediction markets to solve the training data bottleneck issue,” positioning Reppo at the intersection of crypto-native market design and AI infrastructure.

Bolts Capital’s commitment is structured as a strategic investment, underscoring that this is not just balance-sheet capital but a longer-term wager on Reppo’s core thesis and the wider prediction-market-as-data narrative.

Turning human judgment into AI training data

Reppo’s central idea is to convert human judgment into verifiable, incentivized data streams using prediction market mechanisms, rather than relying solely on traditional data-labeling pipelines.

According to the project, this design directly targets “the current challenges in acquiring high-quality data for AI training,” where noisy, biased, or low-signal datasets can cap model performance even as compute scales.

By forcing participants to stake capital on their beliefs and be financially accountable for being wrong, prediction markets can, in theory, produce sharper probability estimates and richer behavioral signals than conventional surveys or annotation tasks.

Building decentralized Datanets for models

Reppo says its protocol supports multimodal data processing, spanning text, images, audio, and video, all organized through decentralized data networks it calls “Datanets.”

These Datanets are pitched as infrastructure for model training, evaluation, and fine-tuning, effectively turning a prediction market layer into a continuous source of scored, time-stamped, and incentive-aligned data.

The $20,000,000 commitment is meant to accelerate both protocol development and ecosystem expansion around these Datanets, from new prediction-market primitives to tooling for AI teams that want to plug Reppo-derived data into their pipelines.

If Reppo can prove that markets can reliably generate better training and evaluation data than conventional labeling shops, Bolts Capital’s bet could mark an early, high-conviction signal that crypto prediction markets are maturing from speculative casinos into critical plumbing for AI development.