Google Labs' Tabletop turns text prompts into shared collaboration apps

An app builder inside a shared space

Google Labs has opened a waitlist for Tabletop, an experimental tool that generates small collaboration apps from natural-language descriptions. According to a write-up on dev.to, the workflow is straightforward: participants drop links, files and ideas into a shared space, prompt an agent called @tabletop, and the system produces an interactive tool — a trip itinerary, a dinner poll, an expense splitter, a tournament bracket — that everyone in the group can use at the same time.

The write-up attributes to The Verge the framing of "vibecoding": building a tool by describing it rather than by configuring templates or filling in a spreadsheet.

The output is an artifact, not an answer

The more consequential shift, the dev.to piece argues, is what the model returns. Instead of a conversational reply, it returns a working artifact that a group uses together — closer in spirit to a Google Doc than to a chatbot answer. The target is the small coordination job, such as a weekly dog-walking rota, brunch stops along a road-trip route, or the bracket for a pickleball league, that today tends to live inside a group chat hundreds of messages long.

Deliberately small scope

Google is aiming Tabletop at families, recreational leagues and potlucks rather than enterprise workflows. The write-up frames this as a wedge into territory held by workplace software: the structure of an Asana or a Monday-style board without the office formality, built for groups that would never open a project management tool to plan a dinner.

That choice also sets a ceiling. Every example demonstrated so far is a small, bounded problem, and the write-up notes that it remains unclear how well the generated tools hold up when requests grow more complicated, or how far they can stretch.

Access is narrow and the roadmap is blank

Tabletop is an experiment rather than a finished product, and Google presents its current capabilities as a starting point. Access is limited to a US-only waitlist restricted to users aged 18 and over — a restriction that sits awkwardly with use cases like family trips, since children cannot sign up. The write-up reports that The Verge speculates the age limit may be relaxed after testing, but notes Google has not said so itself. There is no pricing, no launch date and no expansion timeline; the only entry point is the Google Labs waitlist.

Why it matters

Tabletop is notable less for what it does today — a narrow, gated experiment — than for the direction it points. If an AI can reliably turn a one-sentence description into a shared tool that a non-technical group actually uses together, that pattern is unlikely to stay confined to dinner polls and dog-walking schedules. The same mechanism could be aimed at larger coordination problems, and it challenges the template-and-configuration model that most collaboration software is built on.

Two open questions will decide whether it matters beyond the demo: whether generated tools remain reliable as requests grow in complexity, and whether Google commits to developing the experiment into a real product or lets it join the long list of retired Labs projects. For now, Tabletop is one of the clearest signals yet of AI assistants moving out of the chat window and into the shared spaces where groups actually get things done.