Gemini for Science: AI experiments and tools for a new era of discovery

Details

Gemini for Science is Google's coordinated AI science toolkit announced at I/O 2026, comprising three experimental prototypes on Google Labs. Hypothesis Generation (built with Co-Scientist) simulates the scientific method: it collaborates with researchers to define a challenge, then runs a multi-agent 'idea tournament' where AI systems generate, debate, and rank hypotheses against the published literature, with clickable citations for verification.\n\nComputational Discovery (built with AlphaEvolve and ERA — Empirical Research Assistance) is an agentic research engine that generates and scores thousands of code variations in parallel, letting researchers test novel computational modeling approaches for fields like solar forecasting and epidemiology in days rather than months. Literature Insights (built with NotebookLM) synthesizes understanding across scientific literature to support all stages of research.\n\nThe initiative also introduces Science Skills within Google Antigravity. The broader vision positions general-purpose AI agents — rather than narrow specialist models — as the tools that will accelerate discovery across all scientific domains. Tools are designed for active researchers and available through Google Labs as prototypes; general availability timelines are not specified.