Hacker News Reader

TDD seems to be a good strategy if you trust the AI not to cheat by writing tests that always pass.

You need to keep the TDD LLM and code LLM as a "ping pong pair" with you as the curator / moderator.