Circle has unveiled Arc Studio, an AI coding agent built for its blockchain-based Arc network. Designed to generate onchain applications, smart contracts and AI agents from a prompt, the tool creates full-stack apps within minutes based on natural-language descriptions. It can be used in a web browser and connected as a sub-agent to tools such as Claude Code, Codex and Cursor. It follows Arc’s mainnet launch and supports testing across nine blockchains.

Andrew Ng, a Coursera co-founder and deep learning expert, said software engineering fundamentals remain essential in the era of agentic coding. He said developers must understand core principles to tell coding agents what to do and what to forgo, and to manage trade-offs such as latency, availability, reliability and cost. Citing his “AI Engineering Skills” research, Ng listed five needed capabilities: building full-stack applications, data management, system architecture design, making systems secure and stable, and scaling and operating in production.