Agent Q, The AI Scientist, OpenResearcher, rStar - mutual reasoning, scaling test-time compute, RAG Foundry, improving RAG-augmented LLMs with self-reasoning.