Slopification and Its Discontents
Summary
Charles Leifer chronicles his six-month Claude experiment, applying Claude to review Peewee's asyncio extension, optimize cysqlite, and restructure Peewee's docs. The piece highlights Claude's strengths in bug detection, test coverage improvement, and planning, but reveals limits in producing novel implementations and refactoring, prompting a rethinking of prompting strategies and human oversight in AI-assisted coding. It also touches on broader debates about AI in professional work, including a healthcare example about the risks of over-reliance on AI.