April 20 – April 26 (Published April 30th)
PERSPECTIVES by Steve Payne
18 Crypto Private Financings Raised: $96M
Rolling 3-Month-Average: $232M
Rolling 52-Week Average: $363M
Announced Deals >$50M: 0
No large financing deals were announced last week, but one stood out for its new approach to AI training. The Reppo Foundation secured a $20 million strategic capital commitment from Bolts Capital to scale its decentralized prediction market protocol, and its timing could not be more pointed. The funding will accelerate both protocol development and ecosystem expansion, with a specific mandate to promote the use of prediction markets to solve the AI training data bottleneck.
This lands against a backdrop of explosive momentum in prediction markets. Total notional trading volume across major platforms exceeded $44 billion in 2025, with the sector’s breakout moment arriving during the 2024 U.S. presidential election before rapidly expanding into sports, economics, and crypto-related markets. Reppo’s co-founder sees the category going further still, projecting prediction markets to reach $1 trillion in annual trading volume by the end of the decade.
The core thesis is particularly interesting: prediction markets, which aggregate human judgment through financial incentives, are structurally well-suited to solving the quality problem plaguing AI data supply chains. Entrepreneurloop.com notes that producing just 600 high-quality RLHF annotations can cost $60,000, roughly 167 times more than the compute expense for the same training run. Reppo’s answer is to replace fixed-task data labeling with stake-backed markets, where domain experts are economically incentivized to surface and curate high-quality data across text, image, audio, and video modalities through specialized “Datanets.”
Reppo is not alone in the space. Protocols like Grass and Masa allow individuals to monetize their digital footprints for AI training, shifting value capture from Big Tech to the individual. But Reppo’s prediction market mechanism, where every data contribution is priced and settled on-chain, represents a novel approach to the same fundamental problem: as frontier models grow more capable, clean, preference-rich human data has become the scarce input that matters most.
Architect Partners will be attending Consensus Miami from May 5th – May 7th. Please reach out to glenn@architectpartners.com to schedule a meeting.
