How Crypto and AI Agents Will Interact With Exchanges and Each Other
Speakers
Laura Shin — Host
— Unchained podcast host
Robert Leshner — Guest
— Superstate CEO, Compound founder
Summary
- Bitget recently launched an "Agent OS" AI Hub featuring 58 tools across 9 modules, integrating open Claude and MCP to assist users with trading decisions, Know Your Transaction (KYT) fraud detection, and customer service.
- Decentralized AI inference and training are currently underperforming their centralized big tech counterparts, showing only early "signs of life."
- Contrarian view: AI agents will not exclusively rely on permissionless crypto networks for microtransactions. Traditional payment rails like Visa will remain highly useful for agentic payments, utilizing stablecoins purely as the backend settlement infrastructure.
- High volatility in both equity and crypto markets is largely driven by the market's inability to price in what AI will actually do over the next 12 months.
Trade Ideas
"Decentralized inference and training. Some a lot of these things have not worked... So far, it's definitely net worse than we're seeing call it and, you know, our non-decentralized counterparts." The market is currently assigning high speculative premiums to decentralized AI compute and inference tokens based on narrative alone. However, the actual technological performance and reliability lag significantly behind centralized big tech solutions (like OpenAI or Anthropic). Until the fundamental tech catches up to the narrative, these tokens carry high downside risk as users and developers default to superior centralized options. AVOID. The narrative outpaces the actual utility and performance of decentralized AI infrastructure, making them dangerous speculative holds in the near term. A sudden breakthrough in decentralized compute architecture or a major censorship event by centralized AI providers could rapidly drive developers to decentralized alternatives, causing these tokens to reprice higher.
"A Visa network will still actually be quite useful in things like agentic payments." The crypto industry broadly assumes that autonomous AI agents will exclusively use native, permissionless crypto rails for payments. However, traditional payment networks are already deeply integrated into global commerce. Agents will likely leverage these established networks (Visa/Mastercard) for front-end payments while utilizing stablecoins for back-end settlement, driving massive new transaction volume to legacy payment processors rather than disintermediating them. LONG. Traditional payment rails will capture significant value from the emerging AI agent economy, contrary to the purist belief that crypto will entirely replace them. Pure permissionless crypto networks (like Solana or Layer 2s) could successfully scale microtransactions with near-zero fees and instant finality, rendering traditional networks obsolete for AI agents.
"Stable coins replace call it the settlement infrastructure for what still exists underneath like a Visa network." If stablecoins become the dominant backend settlement layer for global payment networks handling AI agent transactions, the issuers and major stakeholders of compliant stablecoins will see exponential growth in float and volume-based revenue. Coinbase, through its partnership with Circle and heavy integration with USDC, is positioned to be a primary beneficiary of stablecoins transitioning from pure crypto trading tools into global B2B and agentic settlement infrastructure. LONG. Coinbase serves as the premier public equity proxy for the mass adoption of regulated stablecoins in traditional finance and AI payment routing. Regulatory crackdowns on stablecoins, or traditional finance institutions (like JPMorgan) launching their own proprietary stablecoins that bypass crypto-native issuers.
Up Next
This Unchained (Chopping Block) video, published March 11, 2026, features Robert Leshner discussing RNDR, AKT, V, MA, COIN. 3 trade ideas extracted by AI with direction and confidence scoring.