Google quietly opened a new door into what browsing might feel like a few years from now. The company’s Labs team introduced Disco — an experimental, Chromium-based browser — and its headline feature, GenTabs, which uses Gemini 3 to stitch your open tabs and chat history into one-off interactive web apps.
An app born from your tabs
Imagine researching a trip, juggling a half-dozen guides, hotel pages and flight results, then asking the browser to “make a planner.” Instead of a list of links, GenTabs builds a small app: calendar widgets, timelines, maps, crowd‑level tips and direct links back to the original sources. Ask for study help and you might get flashcards or an interactive model. Ask for a meal plan and GenTabs can aggregate recipes and turn them into a shopping list.
That capability isn’t magic so much as a different output for a large model. Gemini 3 can generate interactive interfaces on the fly; Disco packages that ability into a browsing experience that proactively looks across your tabs and past Gemini chats to figure out what kind of tool would help you most. You can refine the generated interface with natural‑language prompts — no coding required — and each generative element is supposed to be grounded with links back to the pages it used.
What Google says (and what it isn’t)
Disco is explicitly experimental. Google Labs frames it as a "discovery vehicle" for ideas that might later make their way into mainstream products like Chrome. The team built Disco partly as a hackathon curiosity that caught on internally, not as a replacement for Chrome. It’s currently gated behind a waitlist and launching on macOS first.
Because Disco runs on macOS at the start, you’ll need a MacBook or other macOS device if you want to try the early testers’ route — for example, a new MacBook is the obvious machine for those who want in. The rollout will be small while Google collects feedback and irons out the rough edges.
Why this matters (and why it might worry you)
This is another step in the long trend of folding large models into everyday products. Google has already been threading Gemini into services and tools — features like an agentic booking flow and deeper Workspace integrations show the company’s intent to make models more proactive and tied to user tasks. See how Google is expanding AI in Chrome and apps with features like AI Mode and Gemini-based document search to understand the broader push toward assistant-driven workflows AI Mode experiments in Chrome and Gemini’s deeper search into Drive and Gmail.
GenTabs’ appeal is practical: it can turn messy, multi‑tab research into a focused, interactive surface that helps you act on information quickly. But it also raises predictable questions — privacy, data handling, and transparency. Disco explicitly uses your open tabs and Gemini chat history to generate experiences, which means the system is ingesting browsing activity. Google says generated elements will link back to sources, which helps with traceability, but users and privacy advocates will be watching how history, cookies, and personal data are used and stored.
What Disco looks like in use
Early previews show a left-hand chat column you can type into, a vertical rail for switching prompts, and a GenTab canvas that updates as you click options like “Historical Bloom Trends” or “Book Nearby Stays.” The UI still has a conventional tab strip for opening original pages. Examples Google has shown — from trip planners to 3D science models — are useful illustrations of how generative interfaces might augment the web rather than replace it.
Limits and plausibility
Disco and GenTabs are experiments, and Google acknowledges there will be wonkiness. Generative models can hallucinate, misattribute, or oversimplify; grounding outputs to sources helps, but it isn’t a cure-all. The Labs approach makes sense here: test with a focused group, see which generated tools are genuinely useful, and then decide whether to fold them into more widely used products.
There’s also the engineering angle: Disco is Chromium-based, so it’s not reinventing the browser stack; it’s redefining how an assistant can reshape a collection of pages into a task-specific interface. That makes the experiment lower-friction for Google while still offering a clear proof-of-concept for “browsing as building.”
If you follow Google’s AI moves in mapping and navigation, you’ll recognize the pattern — models powering more contextual, task-oriented features across Google apps. For instance, Gemini’s influence is already spreading into maps and other surface-level copilots in Google’s ecosystem Gemini arriving in Maps and navigation tools.
This is an early test of a different idea: the web as a toolkit you can remix with natural language. Disco may never be more than a Labs curiosity, or it might seed features we use every day. Either way, it’s another view of a future where your browser doesn’t just surface links — it assembles small apps to help you finish work faster, learn more clearly, and act on what you find online.