Google DeepMind announced on February 12, 2026 that its specialized reasoning mode Gemini 3 Deep Think is now being used to solve real-world scientific challenges, with Duke University's Wang Lab actively employing the AI to design new semiconductor materials. The announcement marks a shift from theoretical AI capabilities to practical laboratory applications in materials science.
Duke University Lab Designs 2D Semiconductor Materials With AI
Dr. Haozhe "Harry" Wang's lab at Duke University is using Gemini 3 Deep Think to design next-generation 2D semiconductor materials. Deep Think successfully designed a recipe for growing thin films larger than 100 micrometers, a precise target that previous methods had challenges to hit. The AI system generates synthesis recipes, assists with laboratory experiments, iterates on code, and creates molecular models to optimize properties like crystal structure. A video demonstration released by Google DeepMind shows the Wang Lab's active work, emphasizing practical scientific application rather than theoretical research.
Gemini Deep Think Uses Agentic Workflows for Research Problems
Google DeepMind released two papers with Google Research showing how Gemini Deep Think employs agentic workflows to solve research-level problems in mathematics, physics, and computer science. According to Winbuzzer's analysis, the AI solved 18 previously unsolved research problems and disproved a decade-old mathematical conjecture from 2015. Deep Think achieved 48.4% on Humanity's Last Exam, 84.6% on ARC-AGI-2, and Legendary Grandmaster status on Codeforces.
Semiconductor Design Represents Critical Real-World Application
The focus on semiconductor materials design addresses a critical need in electronics and computing. By assisting with the creation of new 2D semiconductor materials, Gemini 3 Deep Think contributes to the development of next-generation electronic components. The Wang Lab's work demonstrates how advanced AI reasoning can accelerate materials discovery and optimization processes that traditionally require extensive trial-and-error experimentation.
Key Takeaways
- Google DeepMind's Gemini 3 Deep Think is being used by Duke University's Wang Lab to design new 2D semiconductor materials for electronics applications
- The AI generates synthesis recipes, creates molecular models, and assists with laboratory experiments to optimize material properties
- Deep Think solved 18 previously unsolved research problems and achieved Legendary Grandmaster status on Codeforces
- Two new papers show Gemini Deep Think uses agentic workflows to solve research-level problems in mathematics, physics, and computer science
- The announcement represents a shift from theoretical AI capabilities to practical scientific laboratory applications in materials science