top of page
Search

From Banking to Butterlions: How I Built an AI-Powered Game in 15 Minutes (And You Can Too)

  • Jul 6, 2025
  • 8 min read

It's 6:47 AM on a Saturday. I should be in the gym but I’ve gone down another AI rabbit hole… This blog post was going to be about the impact of AI on employment prospects.  That will have to wait until next week.


I’m staring at a screen full of banking risk frameworks. FMEA. Process hierarchies. Control matrices.


Important stuff but so far, so “day job”.


That’s how I found myself building an AI-powered failure modes and effects analysis (FMEA) explorer application and iterating it into a sophisticated fully-featured tool that could support any bank non-financial risk manager.


Having ‘scratched my own itch’ I thought I would share with you the exact process I used and pick a simpler example - a cute animal game - to illustrate just how easy it is to get phenomenal results with no previous experience with the tools.


If you are an experienced vibe coder you can probably skip this blog.


What follows is a journey that took me from a half-baked idea to a fully functional AI-powered game in under 15 minutes.


Using entirely free tools (or for a few cents for an even better outcome).


And the kicker? As I’ve indicated previously, I’m no coding wizard. If I can do this, trust me, so can you, and with the step-by-step guide below you can follow along to implement your own idea in the time it takes you to drink that morning coffee.


With that, let’s dive in.


When Work Problems Become Weekend Adventures


Let me paint you a picture of where this started. I've been following the explosion of AI-powered development tools for some time, and I started experimenting myself with Replit last year and migrated on to Cursor and recently have been dipping my toes into using Claude Code.


Here's where things get interesting. I had tried the Google AI Studio a while back to build something - it was clunky, created code that didn’t execute and I personally found it frustrating to use.  But I had heard recently that Google’s AI Studio had evolved into an app building rockstar and so when I had my FMEA inspiration I headed over to the AI Studio to try it out.

The FMEA Explorer App
The FMEA Explorer App

Could I use the AI Studio to build my banking app? Could I transform technical operational risk concepts into an engaging user experience? The answer, as it turns out, was a resounding yes.  The AI Studio essentially zero-shotted (where AI is prompted without any examples to work from) a full app, fixing its own bugs before finishing, that exceeded all expectations.


All this in about 90 minutes.  I think it’s pretty slick, but I am biased….


As this was a total success it inspired me to share the process, which if you haven’t built anything with AI yet, will help you get started and show you how easy it is.


So, onto the cute animal game.


How I Built The AI Game in Five Simple Steps


# Step 1: Start with a Terrible Idea (It's Fine, Really)

My first prompt was pretty basic. I opened Anthropic's Workbench (console.anthropic.com it's a few cents to use by adding credits) and typed something like:

"I want to build a fun game that takes a range of cute animals from the real world, including historical creatures such as dinosaurs, and current day vertebrates and invertebrates. Start with maybe a matrix of cards depicting those animals then allow the user to click on two of the animals and create a mythical new card creature with the new synthesis of the two creatures."

Reading it back, there is a lot wrong with it (I’m sure I will blog on prompt and context engineering at some point). But here's the thing - starting badly is better than not starting at all and I knew I had an ace up my sleeve.


# Step 2: Let AI Be Your Coach

This is where the first bit of magic happens. Anthropic's Workbench has this brilliant feature where you can click "Generate Prompt" and it transforms your word salad into something actually useful. Watch what happened to my rambling: The AI restructured my idea into clear sections:


- Initialise the game

- Create the game interface

- Implement user interactions

- Define fusion logic.


It even added placeholders for variables like `{{INITIAL_ANIMAL_LIST}}` and `{{NEW_ANIMAL_DESCRIPTION}}`. Suddenly, my vague idea had structure, logic, and implementation details I hadn't even considered.


# Step 3: Port to Google AI Studio (Where the Real Magic Happens)

Here’s where things get exciting. I copied my improved prompt and headed to aistudio.google.com. If you haven't used this before, it's Google's playground for AI experimentation, and it's completely free. Once there, I clicked on the “Build" icon and then pasted the enhanced prompt and this is important - I didn't overthink it.


# Step 4: Watch Your Creation Come to Life

I hit the button, the screen refreshed and then…. Nothing....  The chat section blinked -“Thinking….”.

The app being built
The app being built

Was this going to be a waste of time?


Then the screen flickered.  Code started cascading down the middle panel like something from The Matrix.  I held my breath as the preview panel on the right began to render…


What happened next still feels like magic. Within seconds, Gemini had generated a complete, working application:


  • A beautiful purple-themed interface (I didn't ask for purple, but I'm not complaining)

  • A grid of adorable animal cards with custom artwork

  • A "Create New Animal" input field

  • A fusion mechanism that actually worked

  • And here's something I hadn’t envisaged: a "Creature Lineage" feature that shows the family tree of each fusion.


That lineage feature? That was Gemini's innovation, not mine. The AI didn't just follow instructions; it improved on my idea.


# Step 5: Test, Play, and Marvel

The first fusion I tried? Butterfly + Lion = Butterlion. The AI generated this description:

"Behold the Butterlion! With a fluffy golden mane like the sun and rainbow shimmer butterfly wings, this friendly fellow flutters softly through the air, landing gently with paws light as petals."

I laughed. Not because it was silly (though it was), but because it worked. In under 30 minutes, I'd gone from a slick - but serious - risk app to creating something genuinely fun.



Why This Changes Everything

"Okay," you might be thinking, "you made a cute game. So what?" Here's what this experience taught me, and why it matters:


# 1. The Barrier to Creation Has Collapsed

As Mustafa Suleyman writes in "The Coming Wave," we're entering an era where AI democratises creation at an unprecedented scale. You don't need to know JavaScript, React, or any programming language. You need an idea and the courage to try.


# 2. AI as Creative Partner, Not Just Tool

Notice how Anthropic helped with the detailed design and Gemini added the lineage feature? This wasn't following orders it was creative collaboration. As one developer on X recently noted: "We've moved from 'AI as assistant' to 'AI as co-creator.' The implications are staggering."


# 3. The Speed of Iteration Is Game-Changing

Traditional development: weeks of planning, coding, testing.


AI-assisted development: 30 minutes from idea to working prototype. This isn't about replacing developers it's about empowering everyone to become builders.


# 4. Free Tools, Limitless Possibilities

Everything I used was free or very low cost:

  • Anthropic Workbench: A few cents to use the prompt generator or use any LLM as described below

  • Google AI Studio: Free

  • The computing power to run it all: Free


The only required investment? Time and curiosity.


Your Turn to Build

Here's my challenge to you, and I mean this sincerely - this week, build something. Anything. It doesn't have to be good. It doesn't have to be useful. It just has to exist.  Let’s look at how you can get going.


# Your 30-Minute Blueprint:

Before you begin, set up these accounts:

Anthropic Workbench

Sign up at console.anthropic.com (free account and you can start with $5 in credits)

Any LLM chat interface (ChatGPT etc)

And use the alternate LLM prompt below

Google AI Studio

Log in with any Google account at aistudio.google.com


That’s it! No credit cards needed (although I do recommend the Anthropic Workbench if its your first time), no downloads, no installations.


  • Pick a Problem or Passion - Something from work that bugs you - A game idea you've always had - A tool that would make your life easier. If you don't have an immediate inspiration type this into any LLM: "I want to vibe code a simple app. Work with me by asking questions and make suggestions to come up with an idea."

  • a. Write a Terrible First Prompt - Go to console.anthropic.com - Click “Generate a prompt” on that home page - Type out your idea, however messy - Click the"Generate" button to get AI to co-create with you.  Feel free to review and refine the idea from this output.

    b. Alternate LLM - Go to your favourite chat interface for your LLM and prompt with:

"I want to build an app using Google AI Studio, but my initial idea is rough and needs structure. Please help me transform my basic concept into a clear, detailed prompt that will give Google AI Studio the best chance of building a working application. My rough idea: [PASTE YOUR IDEA HERE]"

  • Head to Google AI Studio - Visit aistudio.google.com - Click on the “Build” icon on the sidebar (left hand side of the screen when I do this - it looks like a jigsaw piece).  Paste in your refined prompt from the Anthropic output and let it rip by hitting “Go” (the up arrow at the right of the prompt box).

  • Iterate Without Fear - You’ll then be presented with three panels - the chat interface on the left, the code base in the middle (don’t worry - you can essentially ignore this!) And the app preview on the right hand side which you can make full screen.  Don’t like the result? Modify and try again - Each attempt takes seconds, not hours - Embrace the weird - my best features came from AI suggestions

  • Share Your Creation - Screenshot it - Write about what surprised you - Tag me on LinkedIn I genuinely want to see what you build


Top tip:  If the AI starts describing changes rather than making them, here’s the magic phrase that got me out of a doom-loop: ‘Please update the actual code now, not just describe the changes.’  It’s like dealing with a brilliant but easily distracted creative - sometimes you need to be direct!


Where We Go From Here

As I write this, I'm already thinking about version 2.0. What if the animals could battle? What if we added sound effects? What if... But that's not the point. The point is that we're living in an extraordinary moment where the distance between imagination and creation has shrunk to almost nothing. The question isn't whether you can build something it's whether you will. Bruce Schneier recently wrote about the "democratisation of capability" in AI. He was talking about security risks, but there's a parallel: the democratisation of creativity. We're all builders now, if we choose to be.


I started Saturday morning pondering FMEA analyses and operational risk frameworks. I ended it with a game featuring a Butterlion and a Pandala (Panda + Koala, obviously). The journey between those two points? That's where the magic happens. Your turn. What will you build this week? If this is your first time building, I'm very excited for you (and slightly envious)!


Drop a comment below with your creation, or reach out on LinkedIn. I'm serious about wanting to see what you build. The weirder, the better. And remember if someone who spends their days thinking about non-financial risk and banking regulations can create a mythical animal fusion game, imagine what you can do.


Until next time, you’ll find me at 5.30am on a Sunday, turning spreadsheets into space adventures…



Resources & Further Reading


Tools to Get Started:


Essential Reading:


 
 
 

Comments

© 2023 by therealityof.ai. All rights reserved

bottom of page