TDD inside the agent loop - theater or actual value?
Summary
The article investigates whether running a full test-driven development (TDD) workflow inside an AI agent loop provides real value. It describes three possible TDD patterns for AI-assisted coding and reports an exploratory evaluation where non-TDD approaches occasionally produced better design and test quality, with no clear mutation-score advantage for TDD. It also highlights trade-offs such as higher token costs and prompt maintenance, suggesting monitoring outcomes and exploring alternative approaches.