Evaluator-Optimizer

Implement an evaluator-optimizer agent loop with the Vercel AI SDK where a critic LLM scores outputs and a generator LLM iterates until quality thresholds are met. This self-improving pattern produces significantly better outputs than single-pass generation for complex tasks.

• High-quality writing that iterates until tone, length, and clarity pass review
• Code generation with automated test and fix cycles
• Data extraction with validation-retry loops for structured accuracy
• Creative content that self-evaluates against scoring rubrics