Structured outputs reduce errors in recursive AI reasoning

Original: RLM Agents live healthier when they talk via Structured Outputs

Source: x.com ↗