Google Labs tests GenTabs to turn messy tabs into interactive apps

We have all been there: dozens of browser tabs open for a single trip, a research topic, or a home project, turning your screen into complete visual chaos.

Google Labs is now testing a radically different way to clean up that mess. The team just unveiled Disco, an experimental workspace designed around modern browsing, alongside its headlining feature called GenTabs.

Powered by Gemini 3, which Google calls its most intelligent model yet, GenTabs does not just summarize what is on your screen. Instead, it reads through your open tabs and recent chat history to figure out what you are actually trying to achieve, then writes an interactive web application on the fly to help you finish the job.

Browsing turns into building

You do not need to write a single line of code to use it. If you are comparing hotels, flights, and train tickets, you can simply tell GenTabs what you need in plain English, and it will spin up a tailored dashboard to organize your plans.

If the generated tool is not quite right, you refine it using natural conversation. The system can even proactively suggest interactive mini-apps based on the context of your open tabs before you explicitly ask for one.

Crucially, Google made sure every generative block stays anchored to the web. Every piece of dynamic data links back directly to the original web sources, cutting down on hallucination worries.

According to Google, early testers are already spinning up bespoke applications for weekly meal prepping, organizing cherry blossom trips to Japan, and building interactive planetary study guides for elementary school students.

Disco is currently running an invite waitlist, starting exclusively on macOS with a small cohort of testers.

Why it matters

We think GenTabs represents a fundamental shift in what a browser is supposed to do.

Until now, browser AI has mostly meant sidebars that summarize text or answer basic questions. Google is pushing the concept a step further by turning browsing into instant software creation. You stop jumping between twenty tabs to copy and paste notes, and instead get an interactive tool built just for your session.

For everyday researchers and power users, the potential time savings are huge, assuming the apps are reliable. The main risk here is reliability: if a generated interface glitches or misunderstands your intent, fixing it in natural language could quickly become more annoying than manual research.

For the broader industry, this is a clear shot across the bow for standalone productivity apps. If your browser can synthesize arbitrary web pages into customized planners or comparison dashboards on demand, dozens of lightweight web utilities suddenly become redundant.

Putting this in Google Labs first is a smart play. It lets Google iron out the rough edges with power users before deciding whether to bake this generative capability into Chrome, where it could reshape how billions of people interact with the open web.