Google has officially introduced "Gemini for Science," a comprehensive suite of experimental tools and science skills debuted at I/O 2026 to accelerate core stages of the scientific method. Moving beyond narrow, specialized tools, the initiative shifts toward general AI agents designed to handle dense data synthesis, complex computational modeling, and automated workflows across diverse scientific disciplines.
The ecosystem introduces three prototype platforms on Google Labs: Hypothesis Generation (built with Co-Scientist), which runs a multi-agent "idea tournament" to debate and evaluate peer-verified theories; Computational Discovery (built with AlphaEvolve), an agentic engine that simultaneously codes and tests thousands of parallel modeling variations; and Literature Insights, which structures vast research corpuses into searchable side-by-side matrices. Additionally, Google launched "Science Skills," a specialized bundle drawing data from over 30 primary life science databases like AlphaFold and UniProt. Integrated into agentic platforms like Google Antigravity, the framework compresses tedious multi-hour bioinformatics tasks down to mere minutes. Early internal testing has already produced novel insights into rare genetic mutations of the AK2 gene, while global industry partners simultaneously pilot the tools to optimize commercial supply chains and accelerate structural immunotherapies.