Two new artificial intelligence systems, developed independently by Google DeepMind and FutureHouse, are demonstrating the potential to accelerate scientific discovery by assisting researchers in hypothesis generation, experiment design, and data analysis. Published in Nature, these tools, named Co-Scientist and Robin, are designed to work alongside human scientists rather than replace them, addressing the growing complexity and interdisciplinary nature of modern research. Both systems operate as multi-agent platforms, utilizing multiple specialized AI agents that collaborate to execute different stages of the research workflow. This architecture allows the agents to propose novel hypotheses, design appropriate experiments, interpret results, and refine their theories based on new data. The approach aims to streamline the traditional cycle of scientific inquiry, which often requires deep subject-specific expertise combined with broad knowledge across fields. Co-Scientist, built on the Gemini foundation, is a general-purpose system applicable across various scientific disciplines. Initial validations focused on biomedicine, where the system successfully proposed new drug candidates and combination therapies for acute myeloid leukemia, an aggressive cancer of the white blood cells. Cell line experiments indicated that these suggested treatments were potentially beneficial, though the researchers emphasize that rigorous preclinical and clinical assessments are still required for full therapeutic validation. Beyond oncology, Co-Scientist also identified new drug targets for liver fibrosis and uncovered genetic mechanisms behind antimicrobial resistance. The study was led by Vivek Natarajan and colleagues. Robin, developed by FutureHouse, utilizes models from OpenAI and Anthropic to support research in experimental biology. The system was applied to drug discovery for dry age-related macular degeneration, a leading cause of blindness in developed nations. Robin successfully identified a modifiable process within retinal cells as a potential therapeutic target and suggested a repurposed drug candidate that had not previously been considered for this condition. Additionally, Robin proposed follow-up studies to investigate underlying mechanisms, leading to the discovery of novel drug targets. Like Co-Scientist, the authors note that any treatments suggested by Robin would require extensive preclinical and clinical testing before human application. This work was conducted by Samuel Rodriques and colleagues. A critical aspect of both projects is the "human-in-the-loop" approach. The teams stress that these AI systems serve as collaborative partners, with human scientists maintaining oversight and decision-making authority throughout the research process. These real-world demonstrations offer a glimpse into a future where AI agents handle repetitive or computationally intensive tasks, allowing researchers to focus on high-level strategy and interpretation. By integrating autonomous agents into the discovery pipeline, these systems aim to reduce the time and effort required to validate new scientific ideas, potentially speeding up the path from hypothesis to life-saving therapies.