- Google is working to make the Gemini app, a universal AI assistant that can perform multiple actions on your behalf.
- It will be powered by the Gemini 2.5 Pro model, which is being dubbed as a "world model".
- In the future, the Gemini app will be able to look up information, buy things for you, make bookings, do research, and more.
At I/O 2025, Google demonstrated its vision of building a universal AI assistant, powered by the Gemini 2.5 Pro model. Google is working to make the Gemini 2.5 Pro a “world model” that can make plans, imagine new experiences, understand and simulate different aspects of the world, and take actions, just like the human brain.
Since Gemini 2.5 Pro is natively multimodal, Google says it can power the universal AI assistant in the Gemini app that can multitask for you. The idea is to extend the capability of Project Astra aka Gemini Live in the Gemini app by making it more action-driven.
Currently, you can only share your screen and camera with the Gemini app and interact with the model in real-time. But Google is now testing intelligent actions as part of the Gemini Live experience. While connected to Gemini Live, it can find information from the web, make bookings by calling the vendor, buy things for you, do research, and more.
For example, you can ask Gemini Live to open YouTube and play a specific video. It can even download user manuals from the web, scroll through pages, and find the correct information for you. In addition, you can share your camera, and Gemini can highlight the exact part on your screen.
Google says all these capabilities will be integrated into the Gemini app, making it a universal AI assistant. Basically, it will understand the context, plan what you want, and execute the action on your behalf. The early research prototype of Project Astra is being tested with trusted testers, and it will be integrated into the Gemini app in the future.
