MyTechDailyNews General Guide: What a twelve-day iPhone game build can teach everyday makers

MyTechDailyNews General keeps returning to one practical question: when an app appears to have been produced by AI, what can readers do with that signal rather than worry about it? The Maker Lab story gives a concrete example. It is an iPhone game, free to download on the App Store, and the build moved from an empty folder to a finished version in twelve days during a recent staycation. The maker did not write the code by hand. The method was to describe the desired experience, play what came back, and point out what was wrong, again and again, until it was right.

For anyone scanning AI headlines at the moment, the more useful lesson may be less about jobs and more about play: the fun that first drew many readers into computing, tools, and making.

The makerspace idea that players actually touch

If players know Overcooked, they can picture the shape of Maker Lab quickly. Instead of a kitchen, the player is running a makerspace. At the hatch, customers turn up wanting things made: a keyring, an engraved coaster, or a drone frame. The maker has 3D printers, a laser cutter, a CNC mill, a couple of hand benches, and exactly one pair of hands. Stock must be fetched, machines loaded, other tasks handled, parts collected, and orders delivered through the hatch before the customer gives up.

The Maker Lab main menu

The first playable space is intentionally rough. Level one opens as a scruffy garage, a bare bulb, and one 3D printer, but the roughness is the point: the room improves as the player improves.

Level one: a scruffy garage, a bare bulb, and one 3D printer

The room as reward

The room is the reward loop. It starts as a lock-up garage with breeze-block walls, an oil stain, and a rusty bike against the wall. Each level cleared makes the room better: junk goes out, shelving arrives, and the bike is restored and lit as the centrepiece of the studio. Ten levels, ten rooms, and the lab is the reward.

Then problems enter the loop. From level five, machines begin to jam, and a jam left too long catches fire. From level nine, the power starts to trip, but the breaker is in the far corner. The hand benches keep working, though, because hands never need electricity.

Fetch the extinguisher to put out the fire

If players clear all ten levels, something changes on the character screen. The project does not reveal what happens.

The AI build loop behind the project

Through work, the project author had access to Claude Fable, a top-tier model from Anthropic, and used it as a coding agent in Claude Code. Because the agent had the project folder, it could build the app, run the tests, launch it in the iOS Simulator, and inspect screenshots of what it had made. The human sat beside that process like an opinionated producer.

The project notes say most of the 144 commits came from the author playing the game on a phone and saying what failed, including the simple complaint, “I can’t tell which printer is free”.

That is the describe, play, complain loop at its clearest: the first prompt was a couple of paragraphs, the first commit produced a playable Level one, and the next rounds came from play rather than code.

Guardrails that make a chat feel like a build

The project had a test system: 416 automated tests. On every run, a bot plays all ten levels with no screen, verifies that each one can be cleared, and checks that three stars stay out of reach on the last three, leaving something for a human to chase.

It also had memory: a short file called AGENT_CONTEXT.md holds decisions, and every session reads it first. That matters because an agent starts every session knowing nothing about yesterday, so a memory file separates a collaborator from a very clever goldfish.

Audio followed the same loop. Every click, bell, and motor hum is synthesised by a Python script, with not one recorded sample. The notes say that, on its own, the twelve days would not have produced that result.

Why this is not only a story about expensive models

The core claim is: you do not need the expensive model. Maker Lab has 3D graphics, physics, audio, and a test suite, putting it at the hard end of what these tools do well. Even so, the approach works at every level, including on a Raspberry Pi on a desk.

Using Ollama or llama.cpp, a Raspberry Pi 5 with 8GB or 16GB of RAM can run a small open-weight model locally. It is slower than the cloud and nowhere near as clever, but it is entirely yours: no account, no subscription, nothing to pay per prompt, and it even works with the Wi-Fi switched off. That makes it a good fit for a classroom, a code club, or a kitchen table.

For small games, that local setup can be plenty. One example from the project is asking for a Pygame game where a cat catches falling fish, then running it and asking the model to make the fish faster every ten points. That is the same describe, play, complain loop used for Maker Lab, only smaller. Watching a credit-card-sized computer write a working game while tea is made is the joyful part, especially for readers feeling glum about where all this is going.

For more AI power, Raspberry Pi makes a range of AI HATs to give it. Vision tasks like spotting people and objects in a camera feed are handled by the AI HAT+. The new AI HAT+ 2 is the one for projects like this: its Hailo-10H accelerator brings 40 TOPS of inferencing performance, and with 8GB of its own memory, the AI HAT+ 2 runs small language models, including a coding model called Qwen2.5-Coder, while taking the load off the Raspberry Pi itself.

Why the game is free and what it supports

Maker Lab is free, with no adverts and no in-app purchases. Just one button in the settings opens the JustGiving page for Alzheimer’s Society.

Maker Lab is raising money for Alzheimer’s Society

Makerspace games ask a lot of the head. A player has to keep three orders in mind while crossing the room, planning a route through the machines, and spotting a jam out of the corner of the eye. Working memory, planning, and attention are required, and all are things dementia takes away from people, and from those who love them. A phone game is not a cure for anything, but if a reader enjoys it, donating a few pounds to the people trying to understand dementia is a good trade.

The project also says the game is loosely based on its author’s own story, from tinkering in a garage to becoming Maker in Chief at Raspberry Pi, one cleared-out corner at a time. The author’s mother has Alzheimer’s, so that little button in the settings carries a lot more weight than its size suggests.

Common mistakes to avoid when copying the method

• Starting with code instead of description. The first prompt was about the game, not the implementation.
• Skipping the play step. Most useful feedback came after every change, when the human noticed what felt wrong.
• Assuming every session has context. Without AGENT_CONTEXT.md, the agent still starts from a blank slate each day, knowing nothing about yesterday.
• Talking about failures as bugs. The loop works better when “I can’t tell which printer is free” becomes a conversation, not a bug report.
• Waiting for a premium model. A small local model may be slower, but it can still run useful projects.

Checklist for a first afternoon

Start small: One screen, one mechanic, one afternoon.

• Describe the feel you want, rather than the code you think you need.
• Play it after every change.
• Write down what you learn in a file the agent reads first.
• When it goes wrong — and it will — treat it as a conversation, not a bug report.
• Have fun, build something, and give something back: three things AI does not tend to get credit for, and three things that are always the point.

FAQ: quick answers for MyTechDailyNews General readers

Is Maker Lab only possible with Claude Fable?

No. The project used Claude Fable, but the same describe, play, complain loop can be used with a local model on Raspberry Pi. The author’s point is that the expensive model is not required.

Can a Raspberry Pi 5 really run a coding model?

Yes. With Ollama or llama.cpp, a Raspberry Pi 5 with 8GB or 16GB of RAM can run a small open-weight model locally. With the AI HAT+ 2 and Hailo-10H, it can run small language models, including Qwen2.5-Coder.

Do I need Android or iOS skills first?

The article does not require readers to know either platform; the loop is described in plain language. If trying it, the recommended first step is a small Pygame game, not a polished app.

Where can readers see or try the game?

Maker Lab is out now, free on the App Store. Readers can try it, donate through one button if they want, and share what they build on a Raspberry Pi or anything else, or show it at a Maker Faire.

Now, the project’s own closing note is light: level eight still needs play-testing.