Gensyn Delphi: AI Settlement Information Market Fee Buyback and Burn

📌 Background

The Gensyn mainnet launched in April 2026 based on Ethereum OP Stack L2 with the release of the $AI token, though the decentralized training mainnet is not yet fully commercialized, with most models at the 1B parameter scale and training task fees not yet becoming the primary cash flow. Its first production-grade application, Delphi—a decentralized information/prediction market with results automatically settled by AI models—has become the protocol's current actual revenue gateway, making 2026 a critical window to observe whether the model of 'feeding back application fees into the compute network' can succeed.

👤 Target Customers

Institutions and individuals creating markets (earning a share based on trading volume), market trading participants, and future AI model developers and enterprises paying for training tasks with $AI

💰 Revenue Streams

The protocol collects a 0.5% protocol fee from Delphi trading volume, which flows into the BuyBack Vault, while market creators receive an additional 1.5% share; the 0.5% fee is used to buy back $AI on the secondary market, with about 70% permanently burned, 29% entering the community treasury, and about 1% going to the executor. Once the training network matures, collecting protocol/service fees from task providers will become the second revenue curve

🧮 Cost Structure

Protocol R&D and security audits, computing power required for AI settlement models, validator and executor incentives, community treasury and ecosystem expenses, team operations

🛡️ Moat

Technical barriers formed by RepOps reproducible execution environment (cross-hardware bit consistency) and Verde binary dispute verification protocol; a deflationary mechanism where the total supply of $AI is fixed at 10 billion and bound to usage volume via protocol fee buybacks and burns; 150,000 users accumulated on the testnet and endorsements from investors such as a16z crypto

🔑 Keys to Success

• Promptly have protocol/service fees from training tasks replace Delphi as the primary cash flow
• Maintain and expand Delphi trading volume to support the buyback-and-burn flywheel
• Continuously prove the reliability, cost advantage, and low barrier to entry of verifiable training

⚠️ Risks

• Current protocol revenue scale is extremely small, and the model has not yet been fully validated by trading volume
• Training mainnet commercialization progress falling short of expectations, potentially failing the long-term revenue curve
• Compliance risks arising from token price volatility and changes in cross-regional regulatory policies

🏢 Cases

• Delphi launched in April 2026 alongside the mainnet, adopting a trading fee structure of 1.5% for market creators and 0.5% for the protocol, with protocol fees entering the BuyBack Vault to execute buybacks and burns
• Gensyn released the open-1b model in September 2026 with complete training audit proofs, demonstrating the actual output of verifiable training
• The Gensyn testnet has cumulatively attracted about 150,000 users, with co-founder Harry Grieve calling it a path to break the ceiling of AI compute scale

📊 SWOT Analysis