Recently, discussions about AI Agents in the market have become more specific. These autonomous code units are seeking the underlying network that best supports their existence. Interestingly, a significant number of models consider Bitcoin as their preferred store of value during reasoning. This logic is actually easy to understand: Bitcoin's fixed supply and self-custody features make it the cleanest and most trustworthy monetary foundation for Agents that need to operate independently over the long term.

However, problems also arise. Bitcoin is an extremely conservative chain, designed as a value settlement layer rather than an execution environment. The needs of AI Agents are entirely different—they require high-frequency micro-payments to invoke machine services, programmable identity systems to verify the credibility of interacting parties, and proof on the network that certain tasks have been correctly completed after transactions. These requirements point to high throughput, complex programmability, and on-chain coordination capabilities—features that Bitcoin natively cannot provide.

This creates an interesting mismatch: Agents want the security level of Bitcoin but also need the execution flexibility of Ethereum or Solana. The industry’s current solution to this contradiction is quite clear: separate these two layers. Let Bitcoin L1 continue to serve as its “digital gold,” responsible for basic asset issuance and settlement, while delegating execution layer tasks to L2. But the key challenge is, as L2 pursues high performance, how to avoid losing the core features of L1.