[krapp]

I'm sorry, but I've seen "finished products" from AI game devs that don't look much better.

This is a simple prototype (so simple I doubt anyone would bother) but the more complex it becomes, the less fine control you'll have, the more you'll have to just hope the LLM gives you something approximating what you want.

And then what? Use what's likely to be complex, buggy and inefficient code that you didn't write and don't understand? Build on top of that? Just keep prompting the AI and have it build the final product as well?

I really don't think this is better than existing methods even for prototyping because AI by its very nature just approximates and that takes creative control away from the developer and the artist. I don't think that's worth a fast turnaround.

[fkhasiyev]

People were talking like this about Cursor, Windsurf and any Code agents 2 years ago. It's improving day by day. It's just a sneak peek now, next year this day, there will be 10x better versions. keep an eye on AI Game Dev in 2026.