We Built 6 Games with Google AI Studio in Under 37 Minutes 🎮

EDITOR’S NOTE

Dear Future-Proof Humans,

What if you could build with Google's most advanced AI models without writing a single line of machine learning code?

This week, I've been exploring Google AI Studio, the browser-based playground for Google's powerful Gemini models. After connecting AI to Gmail, WhatsApp, Figma, and Reddit in previous newsletters, I wanted to see what's possible when working directly with these sophisticated models.

I've been testing the platform extensively and found that it lets you:

✔️ Create working prototypes with just a few prompts—no complex setup required
✔️ Process images, text, and other media types in a single interface
✔️ Export functional code that you can integrate directly into your applications.

In this edition of Nanobits, I cover:

⚡️ A complete walkthrough of Google AI Studio's capabilities
🔗 Practical examples you can try today with the generous free tier
⚠️ Pricing details and limitations to consider before building

Let's explore how these powerful models can transform your next project and be fun at the same time!

GOOGLE AI STUDIO: HANDS-ON WITH GEMINI MODELS

Google AI Studio offers developers a straightforward way to work with Google's Gemini models. This browser-based tool lets both experienced developers and AI beginners prototype and test AI applications without deep machine learning knowledge.

Here's what I found after testing it.

What is Google AI Studio?

Google AI Studio is a browser-based development environment for testing and building with generative AI. Previously called MakerSuite, it helps developers quickly test different models and prompts, making it easier to create AI-powered applications.

The platform bridges the gap between idea and implementation. You can experiment with different approaches and then export your work as functional code using the Gemini API, saving significant time when bringing AI projects to life.

The tool makes advanced AI accessible to developers at all skill levels through a clean, user-friendly interface.

MY (FUN) EXPERIMENTS WITH GOOGLE STUDIO AI

After exploring Google AI Studio's capabilities, I couldn't resist trying something more playful. Could Gemini generate complete, playable games in a single HTML file? Turns out, it absolutely can!

Here are three mini-games I created with simple prompts.

1. Classic 2D Space Shooter

I wanted to recreate that nostalgic arcade feeling, so I asked Gemini to build a space shooter with the basics: player ship, enemies, lasers, and score tracking.

My prompt:

Create a single HTML file with JavaScript and CSS for a classic 2D top-down space shooter game. Include:
- Player controls (arrow keys to move, spacebar to shoot)
- Enemy ships that move down the screen
- Collision detection for lasers hitting enemies
- Score counter
- Life Counter [3 lives per game]
- Game over state when enemies hit the player
Make the game visually appealing with simple shapes and colors.

The result was impressive - a fully functional game with smooth controls and increasing difficulty. The player ship fires lasers that destroy enemy ships, with each hit adding to your score. The game gets progressively harder as your score increases, with enemies moving faster.

All I had to do was copy the HTML file into an online compiler like OneComplier, CodePen, and I had an instant arcade game running in my browser!

2. Virtual Pet (Tamagotchi Style)

Feeling nostalgic for those little egg-shaped virtual pets from the 90s, I asked Gemini to create a browser-based version.

My prompt:

Create a single HTML file with a Tamagotchi-style virtual pet game using HTML, CSS, and JavaScript. Include:
- A cute pet character with simple animations
- Hunger, happiness, and cleanliness stats that decrease over time. Show the numbers. Stats should be out of 100
- Buttons to feed, play with, and clean the pet
- Visual indicators of the pet's mood based on stats
- Along with visual indicators add text indicators as well
- Day/night cycle
- Game over if any stat reaches zero
Keep the graphics simple but charming with CSS.

The generated game featured an adorable pixel pet that responds to care through three main buttons: feed, play, and clean. The pet's expression changes based on its stats, looking happy when well-cared for and sad when neglected.

What surprised me was the attention to detail - the pet actually "sleeps" during the night cycle, and its needs decrease at different rates. The stats are displayed as simple bars that change color when getting low. I spent way too much time making sure my virtual friend stayed happy!

3. Hangman with Animated Stick Figures

For my third experiment, I wanted something classic but with visual flair, so I requested a Hangman game with actual animations.

My prompt:

Create a single HTML file Hangman game with:
- An animated stick figure that builds piece by piece with wrong guesses
- A customizable word list that randomly selects words
- On-screen keyboard for letter selection
- Win/lose states with animations (happy stick figure for win, complete hangman for loss)
- Option to add custom words to the game
- Score tracking across multiple games
Use CSS for animations and make it visually appealing.

The result was both fun and slightly morbid (as Hangman tends to be). With each incorrect guess, another piece of the stick figure appears through smooth animations. Correct guesses reveal letters in the mystery word.

The on-screen keyboard grays out letters you've already tried, and the game included a surprisingly large word list. The winning animation shows a happy stick figure jumping, while losing completes the traditional hangman drawing.

I added a few custom words related to AI and technology to personalize my version, which was as simple as editing the word array in the HTML file.

What I Learned

These experiments showed me how powerful Gemini's code generation has become. Each game required just one prompt and minimal tweaking. The HTML files Gemini created included:

• Clean, organized code with comments

• Responsive designs that work on mobile and desktop

• Proper event handling and game state management

• Simple but effective visuals using CSS

The most surprising part was how complete these games were - not just bare-bones demos but fully playable experiences with proper game loops, scoring, and even difficulty progression.

If you want to try creating your own games, the free tier of Google AI Studio gives you plenty of requests to experiment with. Just copy the generated HTML files into an online compiler like CodePen, JSFiddle, or CodeSandbox to instantly play your creations.

What would you build? A puzzle game? A visual novel? The possibilities seem endless with just a well-crafted prompt.

Copy the generated HTML from each prompt into an online compiler and enjoy!

KEY FEATURES AND CAPABILITIES

Access to Cutting-Edge Models

As of May 2025, Google AI Studio supports a range of flagship Gemini models, including:

• Gemini 1.0 Pro with 30k context tokens

• Gemini 1.5 Pro, offering an impressive 2 million context tokens

• Gemini 1.5 Flash, capable of processing 1 million context tokens

• Gemma 2, the open 27B parameter model with 8k context tokens

Legacy model access for PaLM 2 was removed in early 2024, however fine-tuned Bison models are still accessible.

Intuitive Interface and Rapid Prototyping

In April 2025, Google rolled out significant updates to Google AI Studio, featuring a cleaner look and a more intuitive prompting experience. The streamlined layout makes testing models and navigating the platform easier while keeping essential developer tools at the forefront.

The platform offers different workflows for creating freeform, structured, and chat prompts, each optimized for particular use cases. This flexibility allows developers to explore various approaches to interacting with AI models.

Starter Apps and Native Code Editing

The revamped Starter Apps experience now features native code editing capabilities directly within Google AI Studio. These Starter Apps are fully functional samples demonstrating model capabilities like multimodal understanding, function calling, media generation, and audio streaming.

Developers can customize each app using the built-in code editor, save and share their work, and integrate the code into their applications. All provided code is available for use in developers' own projects, accelerating the path from prototype to production.

Model Tuning Capabilities

Google AI Studio allows developers to tune (customize) Gemini models with their own data to perform better for niche tasks. Using the Parameter Efficient Tuning (PET) technique, developers can produce higher-quality customized models with lower latency than few-shot prompting.

This approach produces high-quality models with as little as a few hundred data points, reducing the burden of data collection for developers. Tuning enables customization for various use cases, including content classification, entity extraction, and summarization.

Multimodal Understanding

A standout feature of Google AI Studio is its support for multimodal AI through the Gemini models. These models can process various inputs, including text, images, and audio, and convert those prompts into multiple outputs, not just the source type.

Multimodal AI expands the capabilities of traditional generative AI, allowing users to prompt with virtually any input to generate almost every content type. For example, developers can build applications that extract text from images, convert image text to JSON, or generate answers about uploaded images.

PRICING & USAGE LIMITS

Free Tier

Google AI Studio offers a generous free tier, making it accessible for developers to experiment before committing to paid plans. Each model has its own specific rate limits:

• Gemini 1.0 Pro: 15 requests per minute (RPM), 32k tokens per minute (TPM), and 1,500 requests per day (RPD)

• Gemini 1.5 Pro: 2 RPM, 32k TPM, and 50 RPD

• Gemini 2.5 Flash Preview: Free tier with unspecified limits

It's important to note that for users of the free tier, Google uses the data (inputs and outputs) to improve their products. However, this data is de-identified from the user's Google Account and API key.

Paid Plans

For developers who outgrow the free tier limits, Google offers pay-as-you-go pricing:

• Gemini 1.5 Pro: $7 per million input tokens and $21 per million output tokens

• Gemini 1.0 Pro: $0.5 per million input tokens and $1.5 per million output tokens

• Gemini 2.5 Flash Preview: $0.15 per million input tokens (text/image/video) and $0.60 per million output tokens (without thinking)

Additional features like grounding with Google Search have their own pricing structure. The free tier allows up to 500 RPD at no charge, while the paid tier offers 1,500 RPD for free, with additional requests priced at $35 per 1,000 requests.

IDEAL USE CASES AND APPLICATIONS

Google AI Studio is particularly valuable for:

Rapid Prototyping and Experimentation

The platform excels at allowing developers to quickly test ideas and iterate on AI-powered features without significant technical overhead. This makes it ideal for proof-of-concept development and exploring the capabilities of different Gemini models.

Chatbot and Virtual Assistant Development

With its support for chat-based prompting and conversation flows, Google AI Studio provides an excellent environment for developing and refining chatbots and virtual assistants.

Content Generation and Enhancement

Developers can leverage the platform to create applications that automatically generate or enhance content, such as summarizing articles, creating marketing copy, or transforming content between different formats.

Multimodal Applications

Google AI Studio's support for multimodal AI enables the development of applications that process and generate content across different modalities, such as image recognition with text output, or text prompts that generate visual analysis.

Low-Code AI Integration

For businesses looking to incorporate AI capabilities without extensive machine learning expertise, Google AI Studio offers an accessible entry point with its intuitive interface and export options.

LIMITATIONS & CONSIDERATIONS

While Google AI Studio offers powerful capabilities, it's important to understand its limitations:

• It is not a full-scale software development environment and cannot replace traditional coding platforms for non-AI-related projects.

• The platform operates primarily in the cloud, making it unsuitable for on-device AI processing that requires local execution.

• Google AI Studio doesn't support deep custom model training beyond the provided Gemini models. For training AI models from scratch using custom datasets, more specialized platforms would be required.

• There are rate limits on the free tier that might restrict large-scale applications or high-volume development projects.

• It's distinct from Vertex AI Studio, which is Google's more enterprise-focused AI platform with additional features for businesses requiring deeper model customization.

THE FUTURE OF GOOGLE STUDIO AI

As AI continues to evolve, we can expect Google AI Studio to incorporate more sophisticated capabilities and integrations. Recent updates focusing on improved UI and native code editing suggest a direction toward making the platform even more developer-friendly.

The introduction of Gemini 2.5 Flash Preview, described as Google's first hybrid inference model with a thinking budget, points to ongoing innovation in the platform's underlying models. This continuous improvement cycle ensures that developers always have access to cutting-edge AI capabilities.

END NOTE

Google AI Studio opens doors for creators of all backgrounds to experiment with advanced AI. Its simple interface lets you test and deploy Gemini models for everything from chatbots to games without specialized knowledge. The free tier gives you room to play, while paid options scale when you're ready for production. This tool is part of a larger shift making AI accessible to everyone, not just experts.

A tweet by Deedy (@deedydas) on May 8, 2025 caught my attention:

This hints at where we're headed - AI that can watch video demonstrations and build what you show it. Instead of writing prompts, you might soon just record your screen and say, "make this." The jump from text prompts to video understanding could change how digital products are created.

Which Google AI Studio project will you try first? Let me know what you build!