I built a multi-turn agent-vs-agent blind eval in n8n

*Why single-prompt evals miss the failure modes that matter in production, and how to build multi-turn tests anyone can import and run.* Continue reading on Medium »