This week's cover of *Nature* features an illustration of AI research-assistance systems, whose development has recently been accelerating. The journal introduces two multi-agent AI systems: 'Coscientist' by Google DeepMind and 'Robin' by Future House. A research team at Google Cloud AI Research in Zurich, Switzerland, unveiled Coscientist, built on Google's LLM 'Gemini', in *Nature* on May 19. When researchers input their research goals and existing evidence, the system proposes new hypotheses through a 'tournament-style evolutionary process'. Its performance was validated mainly in biomedicine, applied to drug repurposing, discovery of new therapeutic targets, and explanation of antibiotic resistance mechanisms. Coscientist identified candidate drugs and combination therapies for acute myeloid leukemia and confirmed efficacy in some cell-based experiments. A Future House research team in San Francisco, USA, reported the multi-agent system 'Robin' in *Nature* on May 19, focusing on experimental biology workflows. Robin combines a literature-search agent with a data-analysis agent to carry out the entire process from hypothesis to experiment.