Building a game with AI: from a notebook to production
Typerush is an experimental project created with a very simple goal:
To see how far you can go building a real product using AI, free tools, and an idea that had been sitting in a notebook for a long time.
_
This project aimed to answer one question:
Is it possible to turn these ideas into functional products without knowing how to code?
Typerush is the result of that exploration. It’s not just a game.
It’s an experiment about product, AI, and process.
Case Study
Typerush
Design and development of a casual game
to discover and improve typing speed
The starting point
In my day-to-day life, the train is a regular means of transport, and one of the most curious things I’ve noticed is how many people are typing messages on their phones at an astonishing speed, using only their thumbs.
At first glance, it seems like an ordinary action, but in reality it’s a skill that almost no one measures.
After doing a bit of research, I discovered some pretty surprising things:
- The YouTuber Mythicalrocket reached 305 words per minute in 15 seconds (2024)
- The historical record by Stella Pajunas is 216 PPM
- There are modern competitions where this skill is pushed to the extreme
That’s when the idea was born:
Turn typing speed into a short, simple, and addictive game.
The challenge I wanted to explore
We live in an era where writing is constant:
- Chats
- Emails
- AI prompts
- notes
- Social media
But:
- Almost no one knows their real typing speed
- There aren’t many casual experiences to discover it
- Existing tools are either too educational or not very fun
This opened up a clear opportunity:
Create a quick game that turns an invisible skill into a personal challenge.
Project goals
This project had three very clear goals:
- Learn how to build products with AI
- Understand how far you can go using free tools
- Turn an idea from a notebook into a real experience
There was also a product hypothesis:
If users discover their typing speed, they’ll probably want to improve it.
The process — the really interesting part
The process wasn’t linear.
It was more like:
test → fail → learn → try again → break things again → repeat.
And honestly, I think that’s where the real value of the project is.
Phase 1 — Designing with AI
Once I had the product architecture clear, I started working with Figma Make.
The idea was to test something:
How far can AI help build the foundation of an interface?
I started by generating the main screen of the game.
- A lot of prompts
- Many experiments
- Some YouTube videos
- A few articles to learn how to write better prompts
And quite a bit of trial and error.
When I achieved a look & feel I liked, I went back to Figma Design to do what designers do best:
Manually refine the details.
There I took the opportunity to:
- Give personality to the components
- Adjust typography
- Define spacing
- Create the foundation of the design system
I also prepared the library so it could be exported as .json, since I knew I wanted to work with AI on top of that structure later..
Phase 2 — Talking to Claude (a lot)
When I already had:
- Homepage
- Profile page
- Copy Structure
- Tone of voice
I started working with Claude, and that’s when I learned something important.
Running out of credits on the free plan teaches you to think better before writing.
Interestingly, that makes you more efficient. I also have to say: Claude has something different that’s quite addictive.
Building the way a developer would.
_
I decided to follow a fairly classic logic:
functionality first, aesthetics later.
This means the first versions of Typerush weren’t particularly beautiful, but they worked.
And that was what mattered.
The rules of the game (self-imposed)
_
I set a clear premise:
I wanted to see how far I could go investing €0.
Yes, my time has a cost.
But I wanted to test the real limits of free tools.
Claude suggested this stack:
- GitHub → to host the code
- Vercel → to deploy the frontend
- Supabase → authentication and database
And honestly… it worked perfectly.
The initial product structure
The homepage needed to have:
- Language selector
- Login and signup
- A header that challenges you to play
- Quick instructions
- Share results
- A ranking
- And a footer (websites need to end somewhere, of course)
The concept was simple:
Enter, play, discover your speed.
Classic MVP mistake
Intentar fer massa coses massa aviat.
Trying to do too many things too early.
I wanted to include:
- User profiles
- Game history
- Full authentication
- Rankings
And here… I broke the project several times.
Literally. Until I stopped, grabbed a notebook and a pen, and thought:
“How would I make this simpler?”
The answer was pretty obvious: Magic Link.
It’s not the sexiest system in the world. Email isn’t particularly beautiful. But it works, and for an MVP it’s perfect.
Phase 3 — Languages and gameplay
There were two things I was very clear about. The game had to be:
- Extremely easy to understand
- Addictive enough to want to play again
The mechanic is simple:
60 seconds to type as many words as possible, score points, and build combos.
I also wanted to allow playing in multiple languages:
- Catalan
- Spanish
- English
An important decision in the game system
nitially, I wanted to do something quite ambitious: different themes for every day of the year. But that meant generating close to 5,500 words organized by topics.
The problem?
Files that were too heavy, more complexity, less performance. Claude honestly didn’t recommend it. So I had to pivot, and in the end I created a database with more than 1,500 words per language.
The system does:
- Random selection
- Different games every time
- Impossible to memorize
And on top of that, I asked Claude to connect the words through semantic context so the sequence would make sense. This gives the game a pretty natural feeling.
Result
After 5 pretty intense days:
Typerush was working.
All that was left was:
Connect the domain, publish it, and see what would happen.
What did I learn from this project?
An idea without structure will never be a good idea.
It’s a bit like cinema. You can have lots of special effects,
lots of technology, lots of production but if the story is mediocre… it will only be entertainment.
On the other hand, a good story can survive with fewer resources.
With AI, something similar happens.
Before, many ideas never left the notebook because we didn’t know how to code, we didn’t have resources, we didn’t have a team. Now that has changed. AI allows you to prototype, build, validate, and understand whether an idea makes sense before investing too much in it.