Architecture patterns for multi-agent systems
Added a Patterns section documenting the reusable design patterns behind ProspectAI's agent pipeline — adversarial critic, parallel agent execution, LLM output validation, and model tiering — for anyone studying how the brain is built.