Uploaded February 2026 | Updated September 2026, 1 week ago
At Duke University, the Wang Lab utilized Gemini 3 Deep Think to solve a complex challenge in materials science: optimizing fabrication methods for crystal growth.
By applying expert-level scientific knowledge to research-level data, Deep Think successfully designed a precise recipe for growing thin films larger than 100 μm—hitting a specific target that previous methods had struggled to reach.
Learn more at https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think
At Duke University, the Wang Lab utilized Gemini 3 Deep Think to solve a complex challenge in materials science: optimizing fabrication methods for crystal growth.
By applying expert-level scientific knowledge to research-level data, Deep Think successfully designed a precise recipe for growing thin films larger than 100 μm—hitting a specific target that previous methods had struggled to reach.
Learn more at https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think



![Project Mariner | Solving complex tasks with an AI agent in the Chrome browser [full length]
This video has not been edited or sped up for the purposes of the demo. See the shorter, sped up version here: https://youtu.be/2XJqLPqHtyo
Project Mariner is an early research prototype built from Gemini 2.0 that explores the future of human-agent interaction within the Chrome browser. Though in the early stages, it can understand and reason across various web elements, using that information to complete tasks via an experimental Chrome extension.
Learn more about Project Mariner: deepmind.google/project-mariner
—
Subscribe to our channel https://www.youtube.com/@UCP7jMXSY2xbc3KCAE0MHQ-A
Find us on X https://twitter.com/GoogleDeepMind
Follow us on Instagram https://instagram.com/googledeepmind
Add us on Linkedin https://www.linkedin.com/company/deepmind/ Project Mariner | Solving complex tasks with an AI agent in the Chrome browser [full length]](https://i.ytimg.com/vi/KDnij5Q6j1U/mqdefault.jpg)






