Google introduces Vibe coding Gemini that lets users create apps using AI

Google vibe coding

Google has announced a redesigned development flow in Google AI Studio called “Vibe coding” powered by Gemini models. The experience is built to compress the traditional multi-step app-building workflow into a single conversational starting point, allowing developers and non-developers to go from a plain-language prompt to a functional AI app without handling APIs, SDKs or service wiring manually.

How Vibe coding works

The new interface allows users to describe what they want to build — for example, a storytelling app that uses video, image editing or real-time search — and Gemini figures out the needed components and connects them without human intervention. Google said the goal is to eliminate blockers like model orchestration, key management and boilerplate wiring. An “I’m Feeling Lucky” button is available to auto-suggest project ideas when users want to explore possibilities without a defined brief.

Gemini’s multimodal stack inside AI Studio can wire image generation, video tools such as Veo, and search-verified text workflows without switching environments. The app scaffold appears with working code instantly, after which users can iterate.

New interface for iteration and inspiration

Google has redesigned two surfaces alongside Vibe coding. The App Gallery has been rebuilt as a visual inspiration board with tappable previews, starter code and remix options. A new Brainstorming Loading Screen shows Gemini-generated ideas while an app is compiling or provisioning, turning waiting time into ideation time.

For edits, a new Annotation Mode lets users highlight UI areas and speak changes such as “make this button blue” or “animate this card from left” instead of manually describing edits in text or digging into code. Gemini translates these visual annotations into code changes.