Google AI Studio generates Android apps in minutes, but quality issues persist
New testing reveals that while Google AI Studio can produce functional native applications from text prompts quickly, the resulting software often contains significant bugs, inaccurate data, and gameplay flaws.
Google AI Studio, powered by the Gemini model, allows users to generate functional native Android applications directly from natural language prompts within minutes. Recent testing by The Verge demonstrated the rapid creation of three distinct applications: a text adventure game titled "MOOD", a calorie counter, and a platformer titled "Super Peach Rescue". The process requires initial device configuration, specifically enabling USB debugging mode and connecting the Android device to a PC, after which the generated apps can be installed and run directly on the phone.
The speed of generation is notable, with one app created from a 148-word prompt in ten minutes. However, the resulting software exhibited significant quality issues, including logical errors, poor narrative design, and crashes. The text adventure game "MOOD" was generated after a prompt referencing a previous Google demonstration of coding a Doom-like game. While the game is playable, it suffers from a single narrative branch and gameplay mechanics that allow players to win by spamming attack buttons. The platformer "Super Peach Rescue" crashes immediately upon contact with power-up blocks and contains gameplay flaws, such as the character being unable to clear the second pipe.
Data accuracy remains a concern for utility applications. The calorie counter app initially used the paid Gemini API for data retrieval, which failed for the user, and subsequently provided significantly inaccurate nutritional data. For instance, the app listed 16 ounces of boba milk tea as 190 calories, incorrectly matching the term "milk" to low-calorie 1 percent milk. Although the AI corrected the error when pointed out, it continued to underestimate calories for other foods, such as Taiwanese popcorn chicken.
Google AI Studio does not present a development plan for user approval before coding; instead, it proceeds automatically, though users can inspect the code. This contrasts with competing tools like Claude Code, which reportedly utilise a planning phase where the user approves the approach before execution. The service enforces daily usage limits, requiring users to pay or wait for additional iterations, which can restrict productivity for those requiring high volumes of code generation.
The emergence of this technology highlights the growing practice of "vibe coding," where code is generated through natural language prompts with minimal traditional programming knowledge. While the ability to create apps quickly is impressive, the current iteration of Google’s tool suggests that generated software may still require significant human oversight to be viable for direct deployment.
