Nature published two papers describing AI systems — Google's Co-Scientist and FutureHouse's Robin — designed to help scientists identify drug-repurposing opportunities by synthesising vast amounts of scientific literature far faster than humans could. Both systems use an "agentic" approach, autonomously searching and evaluating research to generate and rank hypotheses, while keeping human experts involved in key decisions. Although the tools showed promising results in cell-culture experiments targeting leukemia and macular degeneration, their successes remain limited to relatively straightforward biological hypotheses, and significant challenges remain before AI can tackle harder scientific problems.