Browsing the web often feels like a juggling act, multiplying windows until you get lost. Faced with this fragmentation we've all experienced, a new approach is emerging: it no longer seeks to better manage tabs, but to make them obsolete. That is the bet behind Disco, a project from Google Labs that reimagines web exploration as a dialogue with artificial intelligence.
This project is not a simple update to Chrome, but a distinct browsing environment, built on Chromium and powered by the Gemini AI model. Its goal is to move beyond the logic of clickable links to offer a dynamic and interactive synthesis. The user describes a task, and the system takes care of gathering, analyzing, and presenting the information in a completely new form.
GenTabs: moving from tab-based browsing to a custom application
The core feature of Disco is called GenTabs. It analyzes open tabs and chat history in real time to understand the user's context and goal. Rather than maintaining a collection of static pages, the AI then proposes generating an ephemeral web application dedicated to that specific task. For example, researching a trip could trigger the creation of a unique tool integrating an interactive map, calendar, price comparators, and weather forecasts.
This generation requires no programming skills. The user guides the process with natural language instructions, refining the result according to their needs. They can ask to add a feature, change the presentation, or integrate new information sources. The created application is temporary, but fully functional and interactive, centralizing in a single space data that was previously scattered.
Google assures that links are always provided to trace back to the sites the AI consulted in the background. This transparency aims to preserve access to source content, while offering a layer of reorganization and synthesis designed to increase efficiency and clarity.
An experiment with profound implications
The development of Disco and GenTabs relies on advanced AI capabilities for code and interface generation. The Gemini model is trained to translate an intention, described in words, into a coherent interactive structure, assembling components like maps or filters. This form of "intuitive programming" potentially opens up the creation of digital tools to a much wider audience, without requiring prior technical know-how.
Such an evolution, however, questions the traditional web ecosystem. If the user gets a synthesized and complete answer within the Disco interface, their motivation to directly visit the source sites might diminish. Although advertising revenue or publisher engagement are impacted, Google presents this experiment as a natural evolution of information access, where the value lies in intelligent aggregation and personalization.
For now, Disco is only a prototype accessible by invitation via a waiting list, and only on macOS. Its future is not guaranteed; it is a testing ground whose lessons could, in time, influence more widely distributed products. This controlled testing phase allows for gathering feedback on perceived usefulness, friction points, and user expectations regarding this new form of navigation.