This was a weekend project for my daughter.

She is six year old add sometimes, she get bored and ask «what can we play»‘ So I think maybe an application can help to decide.

This was my initial prompt for GPT-4 and from this prompt I was able to build almost the entire application without any other help than GPT-4.

At the end of the article you’ll have the full conversation with the AI in a very-long PDF. I assure you I worth the read.

My conclusions about this very first experiment with an AI (I have never before done anything like this) are:

• GPT-4 was a tremendous help considering I’ve never build an APP in the past 8 years (I recognize I lied a bit to the AI, as I developed some applications in the past, but never using Swift).
• GPT-4 is able to follow a long conversation and keep track of many things I tell them. During the conversation I told several times to rename variables or functions and GPT-4 remembered that kind of orders on the following iterations without problem.
• GPT-4 is useful to solve specific problems but SW problems are sometimes a bit tricky, and you’ll need some experience to be able to solve it.
• GPT-4 does not make beautiful code or follow best practices by default. For example I needed to ask the AI to make a separated file for a new class instead of declaring it on the same file I was working. Once I instructed the AI to do so, it did it without problem and guide me successfully to do so.
• GPT-4 trust you when you tell them about a problem and focus solving that problem. But if you identified the problem wrongly, the AI is not able to identify what the actual problem is and solving the issue may be a nightmare. During the process I had a problem with a text that disappeared from screen. I wrongly thought the text was hidden behind another element in the screen and told that to the AI. The AI focused on properly stack the elements on Z-index and give me up to 3 differet solutions but no one worked. The problem was the text were pushed of the screen limits instead and not being hidden behind anything else. I need to figure this out by myself.
• GPT-4 sometimes make big refactor of the code that wasn’t needed. For example I asked to modify a function to never return the same result two times in a row, and the AI refactors the entire function with a completely different approach instead of adding a simple additional check to the current function.
• GPT-4 is able to understand and modify code. During the process I was able to pass the AI a function and tell it to modify it to do some extra things.

Download the full conversation: creating-an-app-with-gpt-4

See the full code: https://github.com/danidnm/Random-Games-Chooser