What’s wrong (and right) with AI coding agents
Summary
AI coding agents can produce large volumes of code quickly, but experts warn that trust, testing, architecture, and supply-chain risk have become the real bottlenecks. Leaders from Chainguard, Sonatype, Memgraph, and Sweep urge embedding guardrails—automated tests, policy enforcement, supply-chain intelligence, and AI-assisted review—to prevent speed from turning into chaos. They recommend focusing on architecture and data models over raw lines-of-code to avoid slower, memory-hungry, hard-to-maintain outputs. The piece highlights the risk that easier code generation will increase total software volume and maintenance burden unless governance and visibility improve. Adopt CI-like engineering practices and machine-speed validation to make agentic coding both useful and safe.
