Gemini is evolving from chatbot to cross-device AI assistant
Google Gemini is an AI system that combines conversational AI assistant features, reusable Skills, and multi-step Spark automation across web, desktop, and browser integrations so it can handle real-time conversations, organize information from your tabs, and run background tasks for everyday work and personal use.
The headline shift is that Gemini is no longer confined to a mobile app or a single web page. Google is preparing Gemini Live for the web, with a button for real-time voice mode now appearing in recent builds. At the same time, testing is underway to bring Live to the desktop app through the company’s Trusted Tester program, with a likely simultaneous release across web and desktop even though Live is still missing in the current desktop build. This is not a cosmetic tweak; it is Google turning Gemini into something you talk to wherever you are. Together with new chat-level Skills and expanded Spark access, the message is clear: Gemini is meant to sit on top of your devices and apps, not inside a single product.
Gemini Live: real-time conversations move beyond your phone
Gemini Live features are finally stepping out of mobile confinement and into the web and desktop world. In current builds, a Gemini Live button has started to appear in the web interface, signaling that real-time voice conversations are on their way to browsers. Live is still not present in the desktop app, but Google is testing it with Trusted Testers and a joint web-and-desktop launch now looks plausible.
This matters because voice is how an assistant feels alive. Google previously promised new voice capabilities for the macOS app at its I/O event, including voice selection options and a screen-sharing overlay, groundwork that has been visible for months. If those pieces land together, Gemini becomes something closer to a system-level helper: you can speak to it while sharing your screen, get live feedback, and keep working instead of bouncing between tabs and prompts.
The timing is messy—Gemini 3.6 Flash is in preparation while Gemini 3.5 Pro has slipped past a mid-July target—but Google seems willing to ship the assistant experience upgrades even as the model roadmap wobbles.
Google Gemini Skills: reusable brains for everyday chats
If Gemini Live is the voice, Google Gemini Skills are the habits. Skills have so far lived only inside Gemini Spark, where they act as reusable instruction packages the Spark agent can apply to recurring tasks under the AI Ultra subscription tier. Now, Google is wiring Skills directly into regular chats: traces in recent builds show that ordinary chat users may soon upload their own Skills, pick from predefined ones, build new ones from scratch, or ask Gemini to generate a Skill for them even without Spark access.
"Skills currently live exclusively inside Gemini Spark, the autonomous agent Google gates behind its AI Ultra subscription, where they act as reusable instruction packages the agent applies to recurring tasks". Google is also working on a native Skills menu on Gemini desktop, including the ability to upload, create, edit Skills and organize them into folders. This would put Gemini closer to competitors that already support custom instructions and workflows, and it would make everyday Gemini chats more capable rather than forcing users into a separate “power user” mode.
In other words, Google Gemini Skills turn “remember how I like this” from a one-off instruction into a reusable, shareable tool, which is exactly what a serious AI assistant needs.
Chrome tab groups AI: Gemini inside your browsing, not beside it
The most consequential change for everyday browsing may be Chrome tab groups AI integration. Google is testing an “Ask Gemini about this group” option inside Chrome. When you right-click a tab group in Canary builds, this option gives Gemini access to every tab in that group. Once shared, Gemini summarizes and details the contents of each tab on a new screen named Glic (Gemini Live in Chrome) Tab Mode.
On the Glic Tab Mode interface, you can ask Gemini questions about your tabs, turning a pile of open pages into a single conversational AI assistant session. The integration goes beyond summaries: Gemini can pin or unpin tabs inside the group and tracks changes in real time, so adding or removing tabs does not require re-sharing the group. A flag currently controls access, and it mentions support for Windows, Mac, Linux, ChromeOS, and Android.
This is the kind of integration that makes AI feel useful. Instead of pasting URLs into a chat box, your browser becomes a live data source that Gemini can reason over and help you manage.
Gemini Spark access widens—and hints at where this is heading
While Live and Skills reshape how you talk to Gemini, Spark reshapes what it can do when you walk away. Google is expanding Gemini Spark access after an initially narrow rollout. Google AI Pro subscribers in the US are now getting Spark, and AI Ultra subscribers worldwide are also gaining access with local language support. However, Spark still excludes users in the EEA, the UK, Switzerland, and Nigeria.
Spark is designed to offload longer, more time-consuming tasks to AI; instead of only answering prompts like a chatbot, it can perform multi-step tasks, help with research, planning, and complex workflows while you focus on something else. It is expected to land on macOS “in a much bigger way” later in the summer, working inside the Gemini desktop app to interact with local files and automate desktop workflows.
The broader rollout marks a notable expansion from Spark’s initial availability, which was limited to AI Ultra subscribers in the US. Combined with chat-level Skills and Gemini Live, this signals Google’s intention: Gemini should not be one chatbot among many; it should be an ambient system that shares context across devices, applications, and tasks—even if model versions like Gemini 3.6 Flash and the delayed 3.5 Pro show that the backend is still catching up.
