In the same week of April 2026, two major announcements shook the crypto AI sector. First, Gensyn—a decentralized AI compute network backed by a16z crypto—officially launched its flagship product Delphi on mainnet. Delphi is an information marketplace platform for AI-settled prediction markets, allowing creators to launch their own markets and earn 1.5% of trading volume as revenue. Second, the decentralized AI training data protocol Reppo Foundation announced a $20 million strategic capital commitment from Bolts Capital to advance its prediction market-driven AI training data infrastructure.
These nearly simultaneous developments point to the same frontier—the intersection of AI and prediction markets. Yet, a closer look reveals that each project approaches this space from a distinct angle, with different architectures and competitive niches. This isn’t just an update on two projects; it reflects a structural divergence in the emerging field of AI data verification: on one side, verifiable information markets for humans; on the other, training data validation networks for machines.
With monthly trading volumes in the prediction market industry now exceeding billions of dollars and traditional platforms facing mounting regulatory scrutiny, Delphi and Reppo may be opening up a new competitive landscape. The presence of top-tier investors behind both projects further raises the stakes in this evolving contest.
From a16z’s Heavy Bet to Dual Project Resonance
At a macro level, the AI training data market is expanding rapidly. According to Slator, the global Data-for-AI market is projected to reach $930 million in 2026 and grow to $2.15 billion by 2031, with a compound annual growth rate of about 18%. Another industry report estimates the AI training dataset market will grow from approximately $320 million in 2025 to $1.632 billion by 2033.
Market demand for data is shifting from "quantity" to "quality" and "verifiability." Traditional data labeling relies on centralized vendors, which leads to inconsistent quality, high costs, and weak incentives. The introduction of blockchain and prediction market mechanisms offers an alternative—participants can "bet" capital on data quality, creating higher-quality, more credible data signals through economic incentives.
On the investment side, a16z has doubled down on crypto AI since 2023. In June 2023, a16z led Gensyn’s $43 million Series A round, with CoinFund, Protocol Labs, and others participating. In 2025, a16z completed 31 investments in the crypto space, focusing on prediction markets, AI-crypto convergence, privacy blockchains, and stablecoins, including two investments in prediction market platform Kalshi.
a16z’s 2026 outlook explicitly states that prediction markets will become larger, broader, and more complex. By the end of 2025, combined trading volume on Polymarket and Kalshi reached $28 billion, signaling that prediction markets have evolved from niche experiments to a macro-scale sector.
In December 2025, Gensyn launched Delphi on testnet, followed by a public AI token sale on the Sonar platform, selling 300 million tokens with a fully diluted valuation cap of $1 billion—matching the Series A valuation led by a16z. In April 2026, both projects achieved key milestones in the same week: Gensyn’s mainnet launch and Reppo’s major funding round. This timing underscores the sector’s simultaneous momentum.
Architecture Breakdown: Positioning, Tokens, and Funding—Two Distinct Paths
Although both Delphi and Reppo call themselves "prediction markets," both involve AI, and both aim to solve information verification, their underlying logic is fundamentally different.
Gensyn’s Delphi is positioned as an "information market"—anyone can create a prediction market for any verifiable public event, with outcomes determined by AI models. Creators choose the AI model for settlement, whose parameters are locked at market creation and cannot be changed. External participants can use Gensyn’s "reproducible execution environment" to rerun the model’s inference and verify the settlement’s authenticity.
Reppo, by contrast, is not a human-facing "event betting" platform but an infrastructure for AI developers to validate training data. Reppo builds a dedicated "data network" that turns human judgments into verifiable on-chain signals for AI model training. Its "events" aren’t election results or sports scores, but questions like "Is this dataset’s labeling accurate?" or "Does this data segment improve model performance?"
The core differences can be summarized as follows:
| Dimension | Gensyn Delphi | Reppo |
|---|---|---|
| Market Positioning | General information market (public event prediction) | AI training data validation infrastructure |
| Core Users | Content creators & information traders | AI developers & data labelers |
| Outcome Determination | AI model executes on-chain settlement | Community staking to verify data quality |
| Data Flow | Human-facing—turns public info into tradable signals | Machine-facing—provides high-quality training data for AI models |
| Target Market | Creator economy (projected $500B+ by 2030) | Data-for-AI market (approx. $930M in 2026) |
On the economic model side, Delphi is built around a native AI token. The protocol charges a 0.5% fee on all trading volume to buy back the AI token. Of protocol revenue, 70% is permanently removed from circulation via buy-and-burn, 29% goes to the community treasury, and 1% rewards treasury executors. Market creators earn 1.5% of trading volume as revenue, paid in stablecoins.
Reppo revolves around the REPPO token, with incentives focused on data validation accuracy rather than trading volume. Participants are rewarded for correctly predicting whether a dataset will improve AI model performance; rewards are granted when predictions match actual outcomes. This design discourages low-quality data submissions at the economic level.
In terms of funding, Gensyn has raised over $50 million across three rounds, with a16z’s Series A providing top-tier credibility. Reppo’s $20 million strategic commitment comes from Bolts Capital, with previous support from Protocol Labs and others. Notably, a16z is also an investor in Kalshi, indicating its strategy in this sector is far from a single bet.
Industry Competition Under the Information Market Label
Gensyn makes it clear: its strategy isn’t to compete directly with Polymarket or Kalshi, but to "open an entirely new, creator-owned niche market category." This narrative seeks to distinguish Delphi from traditional prediction markets, especially as the US tightens regulatory scrutiny.
Reppo’s narrative centers on "solving the AI data bottleneck," projecting that the total addressable market for prediction markets could reach $1 trillion in annual trading volume by decade’s end, extending beyond sports and world events into information and opinion domains.
Industry observers remain cautious. Edgen.tech notes that Delphi’s launch coincides with regulatory pressure on prediction markets, and its AI settlement model could offer new approaches. a16z science advisor Andy Hall emphasizes that the future hinges not just on the number of contracts, but on improving "truth-determination methods"—centralized arbitration can no longer meet the needs of large-scale markets.
Can AI settlement truly be decentralized? Gensyn’s REE technology allows external verification of model inference, but issues remain around model bias, the immutability of locked weights, and who controls model selection. For Reppo, the security and reliability of decentralized networks are also challenges—persistent security flaws in DeFi still deter institutional investment, as seen in KelpDAO’s $292 million hack.
Structural Impact: Redefining the AI Data Value Chain
The parallel progress of Delphi and Reppo signals that "AI data verification" is emerging as a distinct sector. Approaching the same domain from different angles, they together form the infrastructure layer for decentralized data validation.
The economic foundation of this sector is clear: the stronger the AI model, the more it demands high-quality, verifiable data. The traditional data labeling industry competes on "cost," but decentralized validation shifts competition to "credibility"—incentivizing validators to stake their own capital on data quality. This shift could reshape value distribution across the AI training data supply chain.
For the prediction market industry, traditional platforms focus on "events." Delphi and Reppo expand the boundaries of "predictable events": Delphi includes "any resolvable question," while Reppo makes "data quality" itself a prediction target. This isn’t just a fight for existing market share; it’s the creation of entirely new market types. The "broader, more complex" prediction market landscape a16z predicted is materializing through these projects.
The ripple effect on the crypto AI ecosystem is also significant: capital is flowing rapidly into data validation, traditional AI data labeling faces structural competition, and the narrative of "data as an asset" is gaining momentum.
Conclusion
In this week of April 2026, two announcements lit up the AI prediction market sector. One unveiled a new paradigm for information market trading; the other showed how prediction market mechanisms are moving upstream into the AI training data value chain.
Both projects are using economic incentives and cryptographic mechanisms to redefine "trustworthy data"—one for human information consumption, the other for machine data production. This duality points to a broader trend: AI systems’ reliance on high-quality data is deepening, and data verification is becoming foundational infrastructure for the AI economy.
As of April 24, 2026, Delphi has moved from testnet to mainnet, and Reppo has completed a new funding round. Both are at the critical stage of transitioning from proof of concept to scaled operations. The next challenges will be real user retention on mainnet, establishing trust in AI settlement mechanisms, and finding sustainable compliance paths amid regulatory uncertainty.
Prediction markets can forecast almost anything—except their own fate. But one thing is certain: the AI data verification sector has evolved from a vague concept into an industry direction backed by capital, technology, and real products.
