Google introduces Gemini 3.8 Live to handle tasks while you talk
On September 15, Google announced Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking. They can continue talking with a user while requested tasks run in the background.
Visual input can provide context alongside spoken questions and instructions. Extended Thinking is designed to reason through multi-step tasks while describing their progress aloud.
Google said rollout would begin that day through Search Live, Gemini Live and developer services, among other channels. Model and feature availability depends on the service and plan.
The model card notes possible incorrect answers and delays. A smooth conversation and a correctly completed task still need to be assessed separately.
Our analysis
If AI chooses the shop, advertising’s role could change
If people can delegate research and tasks during a conversation, they may browse fewer shops and services themselves. The question we find most interesting is who decides which products enter the shortlist. If assistants also help with shopping, prominent advertising alone may no longer be enough to get chosen.
A small shop could become a useful option for a customer in a hurry if an assistant can check its stock and pickup times. But if assistants repeatedly favor particular sellers, control of customer access could concentrate in the companies running those conversations. Voice AI could reshape who connects shops with customers, beyond simply making conversation more natural.
How could everyday life change?
The following is a possible future based on this news.
Getting things done by talking, without hunting through screens
Imagine a shopkeeper arranging stock while discussing next week’s sale. As the AI researches options, the owner adds a budget and names the products to promote. They then review the draft and schedule before approving publication. Work need not stop while the assistant gathers information.
At home, people might plan a weekend while cooking and revise it as family members join in. As connected services and agreed permissions expand, fewer chores would require opening separate apps. The interesting shift is from searching by voice to getting real tasks done during a conversation.
