Your desired setup is just a single prompt away...

sure let's keep piling slop over slop, they're not very good to de-spaghettify code, they're good at piling further slop

LLMs only as good at software architecture as you are.

Isn't that the issue though?

If you are good, like, no?

I mean I crafted complete complex game prototype using Gemini 2.5 Pro with nearly zero coding. I done it in a week: with client-server architecture, robust networking, AI, acceptance test coverage, replays.

It just different way to build software. You just spend 30% of time on specification, 30% on testing and 30% on refactoring also using AI.

Actual slop generarion take like 10% of time and rest of the time you turn it into maintainable code.

Of course you can do it manually, but then it will take 5-10 times the time and you wont be as flexible in changing things because with AI you can do major refactoring in a day, but manually it could take weeks and kill the project.

Share or you're just lying

You're free not to believe me, but I do shipped 3 commercial games (links in profile). I know that what would usually take to build 2-3 months of developer time can now be built in a week.

PS: I might eventually post here about rapid game prototyping with LLM.

If you actually have some credibility feel free to reach me via Linkedin or non-anonymous email and I'd happily share my experience or the code.

BTW if you need to refactor on day one it already smells like good slop

This is a game prototype. If you ever built one you know how iterate on one trying things,

Bold statement not true

Why bother doing that when a non-engineer can just change the prompt and output a different result? :shrug:

right and keep piling slop over slop, software will collapse with this mentality. And more importantly the more the code is convoluted the more even the llm will bail out and won't be able to make further adjustments because of bad code and context rot