Google launches Android Studio Panda 2 with new “Create with AI” feature

Google’s latest Android Studio release, code‑named Panda 2, adds an AI‑driven “Create with AI” assistant that can spin up a full Android project from a single natural‑language prompt. According to Tim Anderson’s report for The Register, the feature is built on Google’s Gemini large‑language model and runs by default on a free tier that offers a “lightweight version of Gemini 2.5 Pro with a smaller context window” suitable for most development tasks. Senior product manager Matt Dyor told The Register that the assistant generates a detailed project plan, then enters an automated loop of code generation, compilation, error analysis, and self‑correction until a working prototype is produced. The workflow can be initiated simply by signing into Gemini and toggling the AI integration flag in Android Studio’s settings.

The assistant’s capabilities extend beyond initial scaffolding. The same Gemini integration also powers an AI‑driven version‑upgrade assistant that can suggest migration steps when a project moves between Android SDK levels. While the free tier limits usage to the lightweight Gemini model, Google’s documentation indicates that professional developers can subscribe to a business tier for access to the full Gemini 2.5 Pro model and larger context windows, which may be required for more complex codebases. The Register notes that developers can also configure alternative LLM providers in the AI settings, though Gemini remains the default.

Privacy and data handling are explicitly called out in the new release. When a developer enables the AI features, a dialog warns that Google will collect “chat text, prompts, related code, generated output” and that human reviewers may inspect this data. The Register advises developers to avoid feeding confidential information into prompts and points to a .aiexclude file mechanism for omitting sensitive files from the AI’s context. This mirrors Google’s broader approach to AI data governance, where user‑generated content is retained for model improvement unless explicitly opted out.

The Register’s hands‑on demo illustrates both the promise and the current limitations of the tool. Prompted to build a bridge‑deal analyzer that performs a “double dummy analysis” of bridge hands in Portable Bridge Notation, Gemini produced a plausible project plan and iterated through code generation for roughly 15 minutes, asking for approval before each file edit. The resulting app compiled and ran in the Android emulator, but a closer inspection revealed logical errors: one generated hand contained 14 cards and the same card appeared twice, triggering an “invalid rank” error during correction. Further analysis showed that the heuristic used for the double‑dummy calculation was simplified and produced inaccurate results on malformed input. The Register attributes these shortcomings to the free‑tier model and the early stage of the feature, noting that “one must not be too hard on Gemini” for its first attempt.

Industry observers see Google’s move as a strategic push to embed generative AI deeper into the developer workflow, competing with Microsoft’s Copilot and GitHub’s AI‑assisted coding tools. By integrating Gemini directly into Android Studio—a product built on JetBrains’ IntelliJ IDEA community edition—Google can offer a seamless, IDE‑native experience that leverages its own LLM stack. However, the mixed results from the bridge‑deal demo underscore the challenge of delivering reliable, production‑grade code from AI alone, especially when developers must remain vigilant about data privacy and the correctness of generated logic.