Robin: The AI System That Discovered a New Drug Candidate on Its Own

Scientific discovery has always followed a familiar rhythm: observe, hypothesize, experiment, analyze, repeat. Each step usually requires a human researcher in the loop. But that rhythm may be about to change. In May 2026, researchers from FutureHouse published a landmark paper in Nature introducing Robin, a multi-agent AI system capable of automating the entire discovery cycle from start to finish.

Robin is built around a team of specialized AI agents, each with a distinct role. One agent reads the scientific literature and proposes novel hypotheses. Another designs experiments to test those hypotheses. A third analyzes the resulting data and refines the next round of questions. Working together, these agents replicate the collaborative structure of a research lab — but they never sleep, and they think at machine speed.

The team put Robin to the test on a genuine medical challenge: dry age-related macular degeneration (dAMD), a progressive eye disease that is one of the leading causes of blindness worldwide and currently has no effective treatment. Robin reviewed the literature, proposed a hypothesis involving a known glaucoma drug called ripasudil, designed cell-based experiments, analyzed the results, and concluded that ripasudil could be repurposed to treat dAMD. The entire loop — hypothesis to validated result — was completed autonomously, without a human writing a single line of the experimental protocol or interpreting the outcomes.

What makes this milestone significant is not just that an AI found a drug candidate. It is that Robin performed every intellectual step of the scientific method on its own. Earlier AI tools had helped with individual stages — generating hypotheses or analyzing data — but Robin is the first system to close the loop, treating discovery as an end-to-end process rather than a collection of isolated tasks. The system identified ripasudil as a promising candidate for a completely different disease from its original indication, a classic example of drug repurposing done entirely by machine.

Knowledge takeaway: FutureHouse's Robin is the first multi-agent AI system to autonomously complete the full scientific discovery cycle, identifying ripasudil as a potential treatment for dry age-related macular degeneration. Published in Nature, it marks a shift from AI as a tool to AI as a research collaborator that can generate and test its own ideas.