Building in public
On 8 September 2026, MiniMe Squad was a single AI image of me surrounded by tiny versions of myself. About two months later it was an iPhone game with 100 levels, 10 bosses and a squad of characters made from photos. I built it with the newest AI tools, and I wrote down everything that went wrong.
Where it started
I asked an image model to turn a photo of me into a crowd of little figures. It worked far better than I expected, and I immediately wondered whether people would play a game with their own face in it.
Two days later I had five game ideas side by side. The throw won, because it is the thing you do hundreds of times, and the thing people film. From there it was a browser prototype, then an iPhone build, then a lot of evenings.
The game
Take a photo and you, your family, your dog or your favourite mug become a MiniMe character. Bugs and bosses come for the squad, and you swipe to throw your minis at them.
This is the moment the whole game is built around. Players can choose a cartoon mini or a more lifelike one, and objects work too.
What surprised me
Getting a first version took days. Getting it to behave the same way twice, on a real phone, for a stranger, is where the time went.
Everything worked in the browser. On an iPhone the shop froze. Finding out why changed how I test everything now.
People, pets and a muddy boot all had to become characters. Each one broke the model in a different way first.
A mini is not finished until it is funny in the air. That one rule did more for the game than any feature.
The full answers, and the fixes, are in the course.
Every mini has its own way of flying when you throw it. Barry rides a rubber duck in. Mavis refuses to participate. Your own mini gets one too.
The course, coming soon
A short, self-paced course behind the scenes of the whole build. Not theory: the tools, prompts, mistakes and fixes, so your next build starts where mine finished.
The waitlist hears about the course before anyone else.
Building something yourself?
I built MiniMe as a deliberate test of the newest AI tools. A game felt like the right playground: fun enough to stay curious, hard enough to show what the tools can and cannot do. The point was to learn how they work, how to get more out of them, and how to use them well on other projects. Those learnings are becoming the course.
The same approach is what I bring to founders as a fractional CTO: find out fast what AI can really do for your idea, ship a working version, then make it solid enough to grow.
FAQ
Yes, with a caveat. AI tools wrote most of the code and made most of the art for MiniMe Squad, but they did not decide what to build, notice when something felt wrong, or test it on a real phone. That part was still me. The course is about how those two halves fit together.
Claude for building, testing and releasing the game, OpenAI image models for turning a photo into a character, and Higgsfield for creatures, bosses and the marketing films. Around them sit ordinary tools: Capacitor and Xcode for iOS, Firebase, PostHog, AdMob and SendGrid.
The first image was made on 8 September 2026. About two months later the game had 100 levels, 10 rooms, 10 bosses and was in TestFlight testing on iPhone.
No, but it helps to be comfortable with a computer and curious about how things work. The course is aimed at founders and builders who want to ship something real with AI, not just a demo.
It is being made now. Price and date are not set yet. People on the waitlist hear first.
Not yet. It is in testing and coming soon to the App Store.
