Lesson 3 · 7 min
Evaluator–Optimizer Loop
Use Claude to grade Claude's own output and iterate.
An evaluator-optimizer loop runs the generator, then a separate Claude call grades the output against a rubric, then a third call revises. This works well for outputs that are easy to judge but hard to produce in one shot — copy editing, code review, design critique.
Keep the rubric explicit and the budget bounded (max 3 iterations).
Quick check
Agentic ArchitectureSelect all that apply
Which of the following are true about evaluator-optimizer loops? (Select all that apply)