Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines
(1 edit) (+1)

You will find that AI involvement in game design is troublesome at best. There isn't an Agent out there that can keep more than 400 or 500 lines of code straight, and even then it is almost inevitable that AI will lose the plot.
As a tool to work out designs, assist in typing and explaining function it does well with limited, small tasks.
It cannot comprehend your vision.. It cannot and will not act as a watchdog, keeping the designer on track.
Over several iterations, the available resources (memory) for the AI is depleted and data loss occurs.
Prompt language is crucial or the AI will assume (they assume anyway) and almost always drift off track.

The previous posters argument that AI is simply stealing other's work is not really accurate. All programmers normally use others' code either deliberately or inadvertently.
For many methods covering common and repeated tasks, it is highly likely that every programmer is going to write it the same way. There are only so many ways to  do many things in code. So where is the line?
I don't agree with simply posting a prompt and assuming what you get is correct and valid, then selling it. I don't see it as theft, but laziness. Low hanging fruit. If you aren't ready to settle with "close", then you will have a lot of granular code adjustments to do, that AI simply can't comprehend.
It isn't the golden ticket that many suggest. It is actually more work explaining and rewriting over and over until you comprehend the problem sufficiently to simply tell AI what to change and let it do the typing. This has been my experience using AI to generate code for several varied projects. I am also learning C# in a trial by fire type of scenario. You have to. AI can't be trusted to see what you see.

The reality of coding with AI is a frustrating and convoluted disaster. If you can get past the AI deciding what you asked for is not what you want and flying off on some tangent repeatedly, you may be able to get something that works as required, but most times you will have to just settle with what you get, and it is likely not what you want.


If you want to test this, simply have Deepseek, Grok, ChatGPT or any other build an interactive tic tac toe game in python. About as simple as it gets these days.
See if you can get good results with a prompt less than  2000 characters and a single generation.
You will quickly understand what I am referring to.