Crypto projects have long proposed markets for dispute resolution, data, computing power and real-world assets. In an article published on November 11, 2026, at 09:52 (UTC+8), Rex St. John argues that many earlier attempts ran into three practical barriers: judging ambiguous real-world claims was expensive, volatile tokens made poor units of account, and settlement was too slow or costly for small transactions.
The proposed remedy is a combination of AI-assisted judgment, AI agents as potential customers, stablecoins and faster, cheaper chains. That is a case for testing old ideas again—not evidence that the resulting businesses will work. An AI model can assess documents quickly, but it cannot guarantee that the documents are genuine or that its conclusion is correct.
Where the argument is strongest—and where it stretches
The article ranks 15 opportunities by the author’s estimates of revenue potential and readiness, not by forecast returns. They fall into several groups:
- Decisions about transactions: Online commercial courts, prediction-market settlement, insurance claims and escrow-backed work marketplaces all depend on resolving disputes or checking whether conditions were met. The author proposes panels of models to review stated criteria and evidence. The economic test is whether customers will pay for decisions that are accurate, appealable and cheaper than existing processes. Insurance also requires appropriately regulated partners.
- Rights and records: Data and media licensing, content provenance, tokenized assets with legal terms, and intellectual-property licensing could use AI to compare contracts with evidence of use or performance. But a timestamp does not establish authorship, and a token does not itself confer legal rights. The article says complex tokenized assets would still need licensed issuers and lawyers; its proposed binding arbitration depends on the underlying agreement.
- Agent-facing markets: Machine payments, reputation-backed guarantees, identity systems, and shared computing or other infrastructure could serve AI agents buying small services. Here, available payment rails or token-funded supply should not be confused with proven paying demand. Citing TRM Labs, the article says x402 had processed about 198.9 million transactions worth $52.7 million by September 2026, while TRM’s screening found that most volume was not yet genuine agent commerce. Citing Messari via Decrypt, it puts the shared-infrastructure sector’s 2025 on-chain revenue at about $72 million, against roughly $10 billion in token market capitalization.
- Governance and verification: Rules-based organizations, curated lists and supply-chain tracking might benefit from document review against written standards. Yet a model checking invoices and certificates cannot verify what is inside a physical box. Human accountability and reliable evidence remain central.
The author’s most useful filter is narrower than the list of opportunities: Was costly human judgment the old bottleneck, and is someone already paying for the imperfect human version? That points first to disputes, market settlement and claims handling. It also demands a product that works without token incentives propping up demand.
The failure modes are substantial. Models can share blind spots; submitted evidence can be fabricated; and insurance, securities and lending rules do not disappear when judgment gets cheaper. Public reasons, an appeal route and comparisons with human review are proposed safeguards, not proof that AI adjudication is fair or dependable. Before reviving an old crypto narrative, the question is still who pays for the service—and what evidence shows it works.
Original article at PANews — written by Rex St. John; translated into Chinese by AIdidiaoJP for Foresight News.