Gemini 3 Deep Think: Accelerating mechanical engineering and rapid prototyping

TL;DR: Gemini 3 Deep Think enables faster mechanical design iteration by generating multiple new design options from images or prompts, helping non-CAD experts rapidly explore and prototype. ## From Tinkering to Engineering The speaker describes a lifelong passion for taking things apart and understanding how they work. That curiosity evolved into a drive to help others: “优秀设计的力量在于，它能改变世界，也能让别人的生活变得更好” (the power of great design is that it can change the world and make other people’s lives better). With Gemini in deep‑think mode, they can now design and iterate at unprecedented speed. ## How Deep Think Accelerates Design The speaker had been thinking for the past year about how deep thinking could really speed up the design process by ten times. The workflow is simple: send an image or a prompt, and the model “深度思考” (thinks deeply) before proposing multiple alternatives. These new designs are things the team had never considered before. In one test, they challenged the model: they gave it an image of a turbine blade. The model generated a new design, and the speaker could then talk directly to the model to change the blade’s pitch or even its shape. The speaker is not a CAD designer—“我自己并不是CAD设计师，所以根本不知道怎么做出来” (I myself am not a CAD designer, so I simply didn’t know how to make it). Yet the AI acted as a catalyst, enabling rapid exploration of material options and focusing on research questions and technologies that don’t yet exist. ## The Broader Mission The team originally built a product for people with cerebral palsy or spinal cord injuries. The speaker sees huge potential: “这个世界上还有很多很多问题等着我们去解决，我们也有巨大的机会让一切变得更好，并且更快地将产品推向市场” (There are still many, many problems in this world waiting for us to solve, and we have a huge opportunity to make everything better and bring products to market faster). Source: [Gemini 3 Deep Think: Accelerating mechanical engineering and rapid prototyping - Google DeepMind](https://www.youtube.com/watch?v=eAOaRJj02HU)