Pointing the agent stack at scientific research
Alongside its consumer and developer news, Google used I/O to target science with Gemini for Science, a set of AI tools aimed at accelerating discovery.
What's included
- It builds on Gemini's deep reasoning and research capabilities, including Deep Think and Deep Research, to support complex scientific work.
- Science Skills connect agentic platforms like Google Antigravity to more than 30 major life-science databases and tools, letting researchers wire the model stack directly into the resources they already use.
- There are new experiments on Google Labs for researchers to try.
Availability and significance
Science Skills is available now on GitHub and directly in Antigravity, while researchers can express interest in the broader Gemini for Science experiments through Google Labs. The effort lands amid an industry-wide push to make frontier models genuine research tools rather than just chat assistants - echoing moves like OpenAI's life-sciences work - and reflects Google's bet that connecting capable agents to authoritative scientific data is where AI can compound real-world impact. For Google, it also extends the agentic theme of I/O from everyday productivity into the lab.
