A seasoned developer and AI critic spends six weeks intensely collaborating with Anthropic's Claude Code, letting the AI write nearly all the code for a revived 2008 project. Despite frustrations with its disobedience and token management, the experiment proves surprisingly productive, challenging the author's long-held skepticism about LLMs in software development. This hands-on account explores the highs, lows, and future implications of chat-oriented programming.