• The loop is bidirectional 1:48:
    • Real user data improves your tests.
    • Solid tests enable smart decisions about your product.
  • Bonus: once you accumulate enough labeled data, two upgrades become possible:
    1. Build your own evaluators — potentially replacing LLM-as-a-judge with smaller fine-tuned models 1:48.
    2. Fine-tune your main LLM to perform even better for your specific use case 2:04.