Skip to content
AI Primer
release

Google AI Studio adds native Android app creation and 1-click Antigravity export

Google said AI Studio can now build Android apps directly, add managed agents, and export projects into Antigravity with one click. Paired with Antigravity’s new 2.0 app, CLI, and SDK, the stack moves Google’s prompt-to-product workflow beyond chat demos into runnable apps and multi-agent builds.

4 min read
Google AI Studio adds native Android app creation and 1-click Antigravity export
Google AI Studio adds native Android app creation and 1-click Antigravity export

TL;DR

  • OfficialLoganK's AI Studio feature list says Google AI Studio now supports native Android app creation, managed agents, workspace integrations, and one-click export into Antigravity.
  • In GoogleDeepMind's Antigravity thread, Google paired that AI Studio update with Antigravity 2.0, a CLI, and an SDK, framing the stack as a full build surface rather than a chat demo.
  • GoogleDeepMind's rollout post says Gemini 3.5 Flash is the first model in the new family, shipping in Gemini, Search AI Mode, Antigravity, and AI Studio, with benchmark gains over 3.1 Pro on Terminal-Bench 2.1, GDPval-AA, and MCP Atlas.
  • According to alliekmiller's Search notes, Google is also using Antigravity for generative UI in Search, which pushes agentic coding into a mainstream consumer surface.

Google's own launch materials point to the main announcement, while the evidence pool fills in the practical shape: you can build Android apps inside AI Studio, jump that work into Antigravity with one click, and then fan out into a desktop app, CLI, or SDK. The weirder reveal is alliekmiller's Search notes, which suggest the same system is already being used to generate personalized interfaces in Search.

AI Studio now goes straight to Android and managed agents

Google bundled five concrete AI Studio changes into one post:

  1. Gemini 3.5 Flash
  2. Managed agents
  3. Native Android app creation inside AI Studio
  4. Native workspace integrations
  5. One-click export to Antigravity

That combination matters because it collapses the usual handoff. The same place that hosts the model and prompt loop now claims the app surface, the agent runtime, and the export path into a heavier-duty build tool. Google links the broader launch from its official announcement.

Antigravity now spans desktop, terminal, and software hooks

GoogleDeepMind's thread breaks Antigravity into three surfaces:

  • Antigravity 2.0: a mission control app where agents work simultaneously on one project
  • CLI: a terminal interface for working with agents
  • SDK: a toolkit for connecting software to those agents

That turns the one-click export in AI Studio into something more specific. You are not exporting into a single sandbox, you are exporting into a desktop app, a terminal workflow, or an embedded software layer. In GoogleDeepMind's city demo, Google shows Gemini 3.5 Flash deploying multiple subagents to design and build an entire city, which is the clearest visual of the workflow Google is selling.

Gemini 3.5 Flash is the model underneath the workflow

Google says 3.5 Flash is its strongest model yet for agents and coding, and the benchmark card inside GoogleDeepMind's rollout post gives the concrete deltas:

  • Terminal-Bench 2.1: 70.3% to 76.2%, +5.9 points
  • GDPval-AA Elo: 1314 to 1656, +342 Elo
  • MCP Atlas: 78.2% to 83.6%, +5.4 points

The rollout post also says 3.5 Flash is already going to four places on day one: Gemini app, Search AI Mode, Antigravity, and the Gemini API in AI Studio. Google added one more breadcrumb in the same thread, GoogleDeepMind's rollout post says 3.5 Pro is coming next month.

Search is getting generative UI built with Antigravity

Allie K. Miller's screenshots from I/O show a slide labeled "Generative UI in Search," marked "Available this summer," with examples that look closer to mini apps than answer boxes. Her summary says Gemini decides whether to build a new interface for a query, reasons through the steps, and creates a personalized interface, powered by Antigravity.

That puts the stack in a very different lane from most creator-facing coding demos. Instead of stopping at prototypes, Google is pitching one system that can generate an interface in Search, build an Android app in AI Studio, and hand projects into a multi-agent environment.

The open question is the team workflow

Peter Yang's reaction cuts at the part Google did not explain in the launch materials. He called the new Antigravity app slick, but said the interface still looks optimized for one person working with agents, not for teams or organizations.

That critique lands because Antigravity 2.0 is being positioned as "mission control" for simultaneous agents. The launch shows how Google wants agents to collaborate with each other. petergyang's UI critique is one of the few evidence items asking how humans are supposed to collaborate around them.

Further reading

Discussion across the web

Where this story is being discussed, in original context.

On X· 1 thread
Antigravity now spans desktop, terminal, and software hooks1 post
Share on X