Google MakerSuite leak reveals Gemini rollout and secret app builder 'Stubbs'

Everyone has been waiting for Google to drop Gemini, and most people assumed the multimodal flagship would land inside Bard first. But a fresh developer leak suggests Google has a completely different playground in mind.

Developer Bedros Pamboukian published leaked screenshots of an unreleased interface inside MakerSuite, Google's developer prototyping environment. The leak confirms that Gemini is on the verge of replacing PaLM 2, but it also unearthed a completely unannounced, secret tool dubbed "Stubbs."

Turn a single prompt into a working app prototype

While Gemini is the marquee name, Stubbs might actually be the bigger surprise here.

Basically, Stubbs lets users generate functional app prototypes straight from MakerSuite with just a single prompt. If you have an image or a screenshot of an app you want to clone or draw inspiration from, you can attach that as well.

To be clear, it does not spit out production-ready backend code. Instead, think of it like an AI-generated, interactive Figma prototype that actually runs. The real kicker is the workflow: you generate the app, deploy it, and publish a shareable link entirely inside one streamlined interface.

Google also built a community hub called the Stubbs Gallery. Your creations stay private by default, but if you choose to publish them, anyone can view your build or hit "Remix" to branch off your idea with their own twist. Oddly enough, the prompt suggestion engine inside Stubbs is currently powered by Google's older, lower-tier text-bison model rather than next-gen silicon.

Multimodal Gemini comes out of hiding

The leak also gives us a clear look at how Gemini—codenamed Jetway—will operate inside MakerSuite, which itself carries the internal codename Alkali.

Sample images inside the leaked UI show Gemini handling optical character recognition, object identification, image captioning, and deeper visual scene understanding. The output settings also hint at support for returning external images and rich HTML content alongside text.

There are still a few quirks worth noting. According to the leak, multimodal inputs currently work inside Text and Data prompts, but the Chat interface does not support them yet. On the quality-of-life front, MakerSuite is finally getting prompt autosave—a lifesaver if your battery dies mid-session—and native, unfiltered translation support that no longer blocks non-English prompts.

Why it matters

We think Stubbs shows Google finally understanding how developers and creators actually want to build with AI. Until now, AI developer suites have felt like glorified parameter playgrounds: you tweak system prompts and temperature settings, then write custom boilerplate code just to test the concept with a colleague.

Collapsing the entire process into a single prompt that yields a deployed, interactive link changes the game for rapid prototyping. For solo founders, product designers, and non-technical builders, being able to pitch an idea with a live web prototype in two minutes is huge.

The Stubbs Gallery with a built-in Remix button is an especially smart move. Social remixing creates sticky platforms, much like what GitHub did for open-source code and what Figma did for interface design.

The real test will be whether the prototypes hold up beyond toy examples once Gemini rolls out broadly. But by turning MakerSuite from a barebones API testing box into an end-to-end app incubator, Google is quietly building a serious moat.