Google Research launched Gemini for Science at I/O 2026, a suite of AI tools designed to accelerate scientific discovery through autonomous research agents.

The platform builds on two foundational systems published in Nature: Empirical Research Assistance (ERA) and Co-Scientist. ERA acts as a code-optimizing research engine that proposes concepts, writes expert-level empirical software, and iterates through thousands of code variants using tree search optimization.

Co-Scientist operates as a multi-agent system based on Gemini that functions as a collaborative AI partner for researchers. The system uses specialized agents that generate, evaluate, and refine hypotheses across scientific domains.

Real-world applications emerge

ERA has already demonstrated practical impact, helping predict hospital admissions for respiratory illnesses and forecasting seasonal runoff across California's river basins. The system has accelerated discoveries spanning neuroscience to cosmology, with results available in Google's new GitHub directory.

Researchers have deployed Co-Scientist to tackle antimicrobial resistance, plant immunity, and liver fibrosis challenges, according to validation papers and researcher testimonials published alongside the Nature research.

Gemini for Science includes several specialized tools. Computational Discovery generates and scores thousands of code variations in parallel, enabling rapid hypothesis testing that would typically require months of manual work. The system combines ERA with AlphaEvolve for enhanced research capabilities.

Hypothesis Generation runs multi-agent "idea tournaments" to generate, debate, and evaluate research hypotheses, complete with clickable citations for scientific rigor. Literature Insights, built with NotebookLM, synthesizes findings across scientific literature.

The platform also offers Science Skills for agentic coding on platforms like Google Antigravity, automating complex workflows including structural bioinformatics and genomic analyses in minutes rather than hours.

Google developed the suite through collaboration with Google Cloud, Google DeepMind, and Google Labs teams. The company is gradually opening access through partnerships with the global scientific community.

Interested researchers can register at labs.google/science. Google is also piloting complementary tools for agentic peer review and scientific validation as part of broader ecosystem efforts.