← sources

AI game development from someone doing AI game design (ie. cloning)

source: reddit (reddit:gamedev)published 2026-10-29 05:39ingested 2026-10-30 02:37

retained text

So, some context. I do game design, have for about six years mostly as a hobby, but recently as a paid game designer for a couple of AI startups. My job, for about four or five months now, has been designing games NONSTOP. Working on five or six games at a time, with multiple Max accounts, generating games using some unreleased tools. I've had a real behind the scenes look at these tools and the process of making games with multiple coding agents, probably hundreds of times at this point.

Where to begin. Honestly, I think the entire industry is cooked. Some of the things I've worked on personally, and some of the stuff I've seen other designers working on that hasn't been released yet, pure cooked.

Every traditional discipline is going to get pushed toward design. That's the distinction I want to make here. When I say design and writing survive, I'm saying that a lot of those other positions BECOME designers. Artists directing the art. Programmers directing the implementation. Animators directing the animation. You're still using that knowledge, but more of your job becomes telling the tools what you want, knowing when the result is wrong, and getting it where it needs to be.  The more experience you have in your discipline, the better your outcome will be with your chosen tool. 

The person behind the prompt has the most direct impact on the outcome. Put someone who understands games behind the seat, who knows how something should play and feel, and you can get an honestly GOOD GAME out of it.

The reason I'm bringing this up isn't necessarily to poo poo and cry doom. It's to give everybody a heads up about things coming down the pipe.

The companies I worked for specifically wanted clones of popular games as prototypes. Things to show investors. I've made playable versions of recognizable games, in hours, that LOOKED GOOD. We're talking PS3 level graphics in roughly three hours from prompt.

Before I quit, I was able to make a really good looking GTA-style prototype. Nowhere near the gameplay detail, of course, but I'd honestly put it somewhere between the graphical fidelity of GTA4 and 5. I wasn't given much time per project, and I wanted to see how much further I could take it, but wasnt given permission because the output was good enough to use a showcase. I eventually left when they started tasking me with some REALLY dumb games.

I've seen Titanfall-style games with absolutely incredible gunplay and movement built in... eight hours?

One of the last really cool projects I saw was a multiplayer game like MechWarrior, where you drop planetside like Titanfall, except you could take over any vehicle on the battlefield and go to war. Really impressive detailed interactive cockpits for multiple vehicles, really smooth gunplay. Ran at 100fps on an iPhone. Thirty-two person multiplayer, big lasers, super cool looking.

That guy was given something like two weeks to work on it, and it absolutely showed. I think it cost something like $2,000 in Claude Max subs. I wasn't privy to the API costs for all the specialized agents working alongside it, generating images, music, models, etc., but there was more on top of that.

My point is, in about one or two months, some things are going to get released, and I think those things are going to make it possible to prompt something like Uncharted 2 into existence in a few hours. Cutscenes, voice acting, gameplay, multiple levels. And it will be “good enough” for a lot of people.

For the record, the reason I took this job was because those prototypes weren't being sold. It was essentially flexing their in-house tools with known games as benchmarks. If Uncharted 2 looks this good, how close can we get with our tools, etc.

The games we found difficult to create or clone were games with a lot of technical depth and interconnected systems.

If your game is about driving, you're cooked. If your game is about driving and doing deliveries, you're less cooked. If your game is Cooking Mama, you're cooked. If your game is Harvest Moon, you're relatively safe, for a bit.

One of the hardest tasks I was given was a functional clone of Stardew Valley. It came out terribly. Stardew has a TON of interconnected systems. Farming, relationships, schedules, progression, dialogue, all needing to work together. Every connection was another thread Claude or Astra had to follow and implement without losing track of everything else.

And it would get lost. ALL THE TIME.

Now, I'm absolutely convinced I could have fixed it, but I was assigned tasks and given X amount of time to work on them. What ended up coming out of that Stardew attempt was actually an incredibly addictive farming and upgrade game. Plant, water, wait, harvest, sell, repeat, faster. Basically a mobile game. Nothing like Stardew, but an excellent game in its own right, and that resulting game actually has a ton of people playing it today.

All of this is to say, the games that are hardest to clone are going to be the ones whose design is complex, whose scope is enormous, or whose story is meaningful.

If your game relies on a simple mechanic, someone is going to reproduce that mechanic in hours. The harder part is reproducing everything around it that makes your particular game worth playing.

I've started thinking about that as the token cost of cloning my game. How much work does someone actually have to do to get close enough? At what point is it no longer worth the squeeze?

GTA6 has all three things going for it. Stardew has a ridiculous amount going on beneath the surface. With the Titanfall-style project, getting movement and combat that felt close enough happened FAST. And for a lot of people, close enough is going to be good enough.

My honest feeling is that in about a year, the entire landscape of indie game development will be irrevocably changed. After an initial tidal wave, tsunami, MEGA TSUNAMI of bullshit clones for everything under the sun, hopefully we'll start seeing games people have only ever dreamed of.

Because a AAA publisher was never going to take a chance on your quirky urban-noir take on being a werewolf in space. Or Spore 2. Or a successor to Black & White. Or whatever incredibly specific game you've been waiting fifteen years for somebody to make.

But somebody out there will want that game made badly enough to make it themselves. And we'll finally get to play some of these games that should exist, but don't.

So the tl;dr is, after working on AI game development from the lab side for a few months, I drastically adjusted the trajectory of my own projects and my future commercial release. I started thinking about my game from a cloner's perspective, and I think everybody else should too, especially before you show anything to the public.

What can someone copy just from watching your trailer? And how much of your game would they still be missing?

We're all still relatively good for now. But come January-February? COOKED. Get ready.

document fetched 2026-10-30 00:26 · canonical https://reddit.com/r/gamedev/comments/1xiblje/ai_game_development_from_someone_doing_ai_game