About AI and Tech Demos
It occurred to me recently while writing my post about AI decompiling that a lot of folks out there actually lack significant context when it comes to looking at AI generated “games” and thinking that we’re super close…

It occurred to me recently while writing my post about AI decompiling that a lot of folks out there actually lack significant context when it comes to looking at AI generated “games” and thinking that we’re super close to just being able to prompt generate your own game. When they see an AI generated mashup of Spider-Man in Arkham City, they immediately think “Wow, that looks amazing! I can’t wait to play it!” and it’s understandable - they’re imagining a fully realized game full of cool and fun content like you’d get from a Spider-Man or Batman Arkham game. The issue is that the AI doesn’t actually do that - it does exactly what it was asked for, and the broader the terminology used, the more difficult it becomes for the AI to give the prompter what she wants.
This should immediately become obvious to anyone who actually tries to play these AI mashups for a few seconds. Any player will play with the locomotion for a few seconds and be impressed, and then they’ll go look for a quest giver or a fight to see how the combat plays and immediately realize that there’s no content to play with, because such things weren’t in the prompt. It’s a good rule of thumb to remember that things don’t exist because they should, they exist because something made them.
This is actually very similar to an experience many of us have had playing many large world games. Think about entering a zone or area where the environment is basically finished - the area has landmarks, environmental geometry, lighting, and so on already. But the zone is also extremely sparsely populated with content - there’s close to no quests, enemies, items, rewards, and general things to do. You can go exploring for its own sake and find some cool vistas, but there’s no treasure at the end of this rainbow aside from seeing something cool. This is interesting to a small subset of players, but most players would quickly get bored from lack of things to do. This is why we must build huge amounts of content to populate these environments - it’s things for players to do.
Now… somebody could try to prompt out a bunch of content and maybe they could get that to work, but that takes a significant amount of time to iterate and build all of this content. That’s why games are expensive - it takes a lot of developers to create the content that populates all of the environments. Somebody has to come up with the idea for the quest to kill ten rats. Somebody has to set up the spawners for the rats you need to kill. Somebody has to create the rat monster itself, somebody has to create the rat-killer badge item reward for completing the quest. All of this takes thought and a coherent plan. What level of player is this aimed for? How hard should it be? Is it part of a longer quest chain? If so, how does that fit together narratively? And so on and so on. All of this requires context and expertise in the skill that you’re asking the AI to do. Without that expertise, you’re going to end up with the kind of shitty gameplay you see in all of those Steam shovelware games.
There’s real potential for AI to be a helpful tool for people who actually have the expertise needed to make their ideas reality, but there’s a very real skill gap that must be closed between a junior developer with big dreams and ideas, and a real experienced dev who knows all the little details that must come together for a cohesive gameplay experience. The next time you see an impressive-looking vibe-coded AI game, I invite you to think about it critically. Look at as if you were evaluating whether you would actually want to spend your money on it. Not what it could be, but what it currently is. Don’t mentally fill in any gaps that aren’t shown explicitly - if they don’t show it, it probably doesn’t exist. What does it show you it plays like? Does it actually look like fun to play? Consider it through this lens - we do this with every real game trailer video that we post. We have to show the game’s features supported so that players know what’s there. You should apply this critical lens to any game, AI-generated or not.
[Join us on Discord] and/or [Support us on Patreon]
Got a burning question you want answered?
- Short questions: Ask a Game Dev on Twitter
- Short questions: Ask a Game Dev on BlueSky
- Long questions: Ask a Game Dev on Tumblr
- Frequent Questions: The FAQ