Google has introduced Android Studio Panda 2, a new feature drop that adds an AI agent capable of generating apps from scratch. The update also includes an AI-based version upgrade assistant to help developers manage project dependencies more smoothly.
Android Studio is built on the community edition of IntelliJ IDEA. By default, the integrated AI runs on Gemini, Google’s large language model family. Developers can switch to other AI providers through the settings if they prefer.
Gemini inside Android Studio offers a free tier that provides access to a lightweight version of Gemini 2.5 Pro with a smaller context window. Google says this version is suitable for most development tasks. However, professional teams are expected to subscribe to a paid business plan for expanded capabilities.
According to senior product manager Matt Dyor, developers can now create a working prototype using a single prompt. The AI agent first drafts a project plan, then follows an automated loop that writes code, builds the project, reviews errors, and corrects itself until the application is complete.
To use the AI features, developers must sign into Gemini and enable AI integration within Android Studio. A privacy notice explains that Google may collect prompts, chat text, related source code, and generated outputs. Human reviewers may also analyze this data.
Developers are advised not to include confidential information in prompts. At the same time, full functionality requires enabling project context access. One suggested safeguard is adding sensitive files to a .aiexclude file so they remain outside the AI’s reach.
Get a Custom Strategy for Successful App Creation in Android.
In a hands-on test, the AI was asked to create a bridge deal analyzer for the card game, bridge. The app needed to accept deals in PBN format and perform a double-dummy analysis, which calculates the maximum number of tricks possible for each side.
The AI generated a clear project plan and appeared to understand specialized bridge terminology. After about 15 minutes of automated iterations, it delivered a working prototype that ran in the Android emulator.
However, closer inspection revealed errors. One test deal contained 14 cards instead of 13, and a card was duplicated. Attempts to fix the issue triggered additional input errors. Reviewing the code showed that the double dummy analysis relied on a simplified heuristic rather than a fully accurate algorithm.
Google itself notes in the settings that Gemini is still experimental and may produce inaccurate or misleading results, even when presented confidently.
The concept behind this workflow, often called “vibe coding,” encourages developers to refine prompts repeatedly until the AI produces the desired output. More detailed instructions, such as referencing established open source bridge analysis libraries, would likely improve results.
While promising, the tool still requires careful oversight and manual validation from developers.
Beyond AI integration, Android Studio Panda 2 continues to evolve as a mature development environment, though performance issues remain. During testing, common errors included Gradle plugin exceptions and slow dependency synchronization.
Google has also deprecated certain features. The Custom View preview tool, once used to create controls without Jetpack Compose, is being phased out in favor of modern UI frameworks. Additionally, the 3D mode in the Layout Inspector has been deprecated due to low usage.
New to App Creation in Android? Let Our Team Guide You Step by Step.
User Retention Hacks 2026: Reduce App Churn & Boost Engagement
[…] notifications can increase app downloads and encourage users to return frequently, supporting overall engagement and […]".