I Read The Clankers' Code
Summary
Charles Féval reviews the code produced by large language models (the “clankers”) and argues for maintaining human review and accountability in software engineering. He acknowledges the speed and usefulness of clanker-generated code but warns that blind trust leads to garbage, hidden issues, and misaligned incentives; he advocates formal processes, human scrutiny, and ownership of code by engineers. The piece emphasizes that clankers don’t learn in the same way humans do, and stresses the need for governance and careful integration into toolchains.