Panel Session: The Next Phase for Digital Money: From Pilots to Production in Korea and…

### Overview This panel session, hosted at a Linux Foundation Decentralized Trust (LFDT) conference, explores the shift in digital money from experimental proofs of concept to real-world production in South Korea and internationally. Featuring perspectives from the Bank of Korea, infrastructure providers, and industry veterans, the discussion focuses on what financial institutions now require to operationalize blockchain technology, the critical role of open source, and the convergence of blockchain infrastructure with AI agents.

### Key Topics - **Evolution from PoC to Production:** Panelists noted that conversations 2–3 years ago centered on whether blockchain technical experiments were possible. Today, focus has shifted to operational viability, 24/7 reliability, ownership, accountability during failures, existing financial system integration, and regulatory compliance. - **Open Source Blockchain Infrastructure & Besu:** Hyperledger Besu has emerged as a widely reviewed and adopted blockchain platform among Korean institutions like the Bank of Korea, Korea Securities Depository, and Koscom. However, production readiness demands resolving challenges in transaction finality speed, on-chain privacy, and compliance. - **Privacy and Layer-2 Innovations:** Ongoing attempts to solve privacy include off-chain projects like Paladin (Kaleido/Project Agorá), MPC/ZKP approaches (e.g., Zama, RCU/Nite), and ZK-rollup Layer-2 designs like LNX/Lynx (compared to zkSync's Prividium, used in US regional bank deposit token initiatives like the Carat Network). Panelists highlighted Rust-based execution environments (RUST EVM / REVM) as critical for performance. - **AI Agents, Programmability, and State Bloat:** The intersection of generative AI and smart contracts enables hyper-personalized financial products where AI negotiates and executes user intent on programmable settlement rails. However, autonomous coding tools and AI agents risk causing massive state bloat on permissionless chains (evidenced by Base chain experiencing 170 million code injections in Q2 2025). Decentralized Identity (DID), via frameworks like Hyperledger Indy, is proposed to authenticate and issue credentials for both human and AI agents.

### Technical Details - **Decentralized Identity (DID):** Applying Hyperledger Indy not as a general-purpose chain, but as a specialized DID registry to issue verifiable credentials for agent authentication without exposing unnecessary personal data. - **State Bloat Management:** Discussion of how permissionless chains must permanently maintain state from AI-generated contracts, contrasting potential pruning or Sol-style state rent mechanisms with current unmanaged state bloat. - **Open Source Security & Supply Chain Vulnerabilities:** Emphasis on the growing need for coordinated security intelligence, aggregating fragmented exploit reports from social platforms like X alongside standard security bulletins to defend against software supply chain attacks exacerbated by AI automated code contributions.

### Takeaways - **Prepare for Worst-Case Scenarios:** Financial institutions prioritize operational resilience and fallback mechanisms over sunny-day demos; developers must build solutions addressing system failures and risk management. - **Market Segmentation by Regulation:** Developers targeting B2B financial services must design strictly around regulatory compliance, while B2C applications can prioritize P2P conversational flows and user technical experience. - **Modernizing Stack & Language:** Panelists advocated moving toward Rust-based blockchain networks for native multithreading without complex multi-chain overhead, and urged global open-source collaboration on privacy, agent identity, and security telemetry.