The field of artificial intelligence is moving towards the development of multi-agent systems that can autonomously conduct scientific inquiry and identify previously unknown scientific principles. These systems, often powered by large language models (LLMs), are being applied to various domains such as drug discovery, materials science, and protein design. The use of LLMs as core agents in these systems enables the integration of various functionalities, including data analysis, knowledge graph construction, and physics-aware property modeling. Noteworthy papers in this area include: PharmaSwarm, which introduces a unified multi-agent framework for hypothesis-driven drug discovery. Sparks, a multi-modal multi-agent AI model that executes the entire discovery cycle without human intervention and uncovered two previously unknown phenomena in protein science. m-KAILIN, a knowledge-driven agentic scientific corpus distillation framework for biomedical large language models training, which achieves notable improvements in biomedical question-answering tasks. UserCentrix, an agentic memory-augmented AI framework designed to enhance smart spaces through dynamic, context-aware decision-making. These systems have shown promising results, including improved efficiency, accuracy, and discovery of new scientific principles, and are expected to have a significant impact on various fields of research.