Building an AI game studio: what we've learned so far

Making an excellent game isn't even enough these days, that's just table stakes. With so many games out there it's a market of lemons. You need to have a way of rising above the noise. For AAA, it's having budgets to make games on a bigger scale than everyone else. For mobile games, it's generally about leveraging a large war chest to fuel an advertising-driven margin game. For prestige/indie games you need something that people will want to talk/write about, or build a dedicated fanbase over time. For tiny games like flappy bird / wordle you need massive built-in virality and lots of luck.

And I think making an excellent game is a lot like reaching product-market fit.

Making an excellent game takes quite different skills than creating a commercially successful game; just compare "pop culture" (optimized to be commercially successful) vs "high-brow culture".

Even if you are not the "extremely high-brow taste" kind of person, there exist quite some video game genres in which excellent games are produced, but the audience of these games is a quite small (but sophisticated) niche audience, so that just by volume, games of these genres likely won't make a lot of money.

You won't find them.

This is subject to survivorship bias. Balatro is a great game that deserves its success. But it's easy to imagine a universe where it simply never got any traction, anywhere, and died in obscurity. Conversely, this suggests that in our current universe, many games just as great as Balatro have quietly passed without ever getting their 15 minutes of fame.

At the end of the day, much of success just hinges on luck. There is no law of the universe that says that great art must be appreciated.

By definition you can't pinpoint (but I think it's a fair assumption) the many great games that don't gain traction on Reddit and Twitch, and silently sink without notice. You've just picked a single anecdote.

"Word spreading fast" means there is marketing going on. Thats kind of the genius of marketing to reddit/youtube/twitch as a "normal guy dev", it doesn't feel like marketing, but thats why its so powerful (see social media influencers). You could be even more cynical and look into guerilla marketing tactics they might use like fake views and upvotes/likes.

And again, survivorship bias.

What exactly do you mean? Reddit and Twitch are marketing-channels too. Marketing does not mean someone has to pay for it, just that people are talking about it and are aware of it.

If you are a beginner composer creating a hit, it is an extreme fluke.

You're punching too high. Music is a passion for many people. Same as game dev. I would expect >95% of game devs to be perfectly happy if they can make a living creating indie games. Of course it's nice to make a hit that makes you tens of millions of dollars, but most game devs would be happy to make enough to pay themselves a salary to live on.

Now, if we're not looking to make a hit, but we're looking to make a living, there are a lot of composers out there who can consistently achieve that (e.g. beatmaking for rappers advertising on youtube and such).

I'd honestly struggle to find any good games (right for their time and audience) that failed because of a lack of suitable marketing.

I'll offer Flappy Bird as an example.

Hate to break it to you, but the pop star success formula is actually meticulously detailed and very repeatable. The producers and song writers know exactly how to do it.

It’s not about challenges in finding success, so much as, at least in music, artists don’t want to follow a formula. They want to create songs and melodies that are personal and meaningful to them and their fans.

I'd honestly struggle to find any good games (right for their time and audience) that failed because of a lack of suitable marketing. In the same vein, I'd struggle to find a hit game that succeeded because of marketing.

only if you're looking at the most general audiences. That's the advantadge of targeting a niche. Not everyone will buy furry art, but the ones that do have high demand (and apparently, deep wallets). Indies can't afford to be "for everyone", and even AAA games are struggling.

But if you want an obvious example of the latter: Raid Shadow legends. Literally spends millions upon millions on advertising a year, and I can't imagine many would organically find and talk about such a game themselves. But even after 6 years they seem to have the funding to get every youtuber under the sun to talk about it. You very much can outadvertise a mediocre game as long as you have the right monetization scheme (which spoilers: is not "buy one copy of $10 game).

95 % of games are boring, and unoriginal. Why? Because most of these people, who create these games are entrepreneurs. They only see the possible money from these. These games have no heart whatsoever, why? Because good games need passionate people, not blue collar suits.

There are 1 million subscribers at the gamedev reddit. It's ridiculous, and when people ask something, they get total balooney, ridiculous botlike answers.

Making good games is easy

This is entirely you hallucinating an opinion they never expressed. The op said:

Making a game

not a _good_ game.

which shows how bad puritan culture is, most people don't care about your puritanism in their room, they want to play a game, see a hot char, and be able to do stuff with them in the game. it's the harsh truth that people seem to get mad at.

Especially now with ever increasing rates of ADHD and other stuff, having an old school genre like an RPG have that type of stuff as a "reward" for beating a level or whatever reinvigorates a genre that felt outdated due to a lack of a proper reward structure.

Fact of the matter is, if you as a game dev want to make money, better lose your inhibitions real fast, because the consumers, want more choices in their game, even if it's just romance for the most part.

Look at stardew valley, I would never have played it if I hadn't found out you could literally date and marry a girl in the game, it just added a layer of reward to the gameplay.

Does that actually matter? You see the same sort of embarrassingly low numbers in all artistic mediums. Writer, painter, musician, sculptor, interpretive dancer, comedian, indy game dev - on average none of these pay enough to count as a career and you only hear about stuff on the successful end of the bell curve. Even if you improve your odds by making commercial art. In many cases even the artists considered successful still need a day job.

Except that’ll just be a fluke.

The next thing isn’t a never-ending list of AI generated games stored somewhere that we’ll have to sift through somehow.

The next big thing will be ephemeral games, created on the fly based on the context and the preferences expressed, both explicitly and implicitly, by the player’s (or players’) past behaviour (and, for multiplayer games, that of their social graph).

That and perhaps, if we manage to preserve it and keep it alive, an industry of hand-crafted games made by people trying to beat the odds.

This theme of comment comes up for sooo many types of posts, and it is unfortunate that it usually dominates the conversation. If Facebook or Airbnb-type companies share a new JavaScript lib, the comment sections will be about how Facebook is evil, or an anecdote about someone's horrible Airbnb experience when they cheaped out in Paris.

Don't want to be rude, but most people here probably aren't game devs. So bikeshedding taking over is inevitable.

Probably the same effect that VCs invest in many startups, knowing only a tiny fraction will make a large profit?

Perhaps a better formulation would be: "Making games is hard. Marketing games is even harder."

When in a gold rush, sell shovels.

You're right, devs aren't as picky with tools as gamers are with games. As long as it works and is proven to save time, it'll sell. But the downside is that you're audience is much less.

Something that I never see addressed for this: So, let's just grant that someday, the tech will be mature enough that this is possible, and let's even say it goes beyond videogames to movies, to visual art, to graphic design, to writing, etc. Let's say that AI gets to a place where any joe blow can put in a prompt, and get a competent, and even let's be generous and say good product out of it. A solid 8/10.

So... who the hell is going to buy it? Because videogames as an industry is already entirely saturated with products that range a whole spectrum from utter dogshit to amazing works of technical expertise, writing, design, etc. There are over 70,000 games on Steam alone now, with 9,000 added in the last 9 months. If this tech actually got to this place, there will be exponentially more games, because all you have to do is tell an AI what you want to play.

And you can take that further: Movies are also highly saturated as an industry, especially as larger studios move ever further into less making "movies" or "series" and just making "content" endlessly remixing their intellectual properties. So now, all of those companies (and all the people who like their stuff) can now just make their own Iron Man movie? Their own Wandavision? Just endlessly making and remaking and remaking, as though tons of people aren't already sick to death of all the television programs and movies that are being made?

And again, you can just keep extending this to any media: print, music, art... we have more of everything now than we ever have before and the goal of companies like Adobe, like OpenAI, etc. is to put even more powerful creative tools into even more hands, broadening the group of people who can create stuff but like... even if you take it as granted that this can be done...

Who the hell is watching all of this stuff? Who is playing all of these games? And why in the world would you pay to watch someone else's AI movie when you can pay to generate your own with whatever you want in it? Why would you ever buy a game off Steam again if you can just ask your game making AI to make you the exact game you want, even just copying the damn description out of steam?

All I see this doing is potentially killing off dozens of creative industries and funneling shit tons of creative control and platform-style power to a handful of massive corporations, running warehouses full of fucking graphics cards, to generate the same games, the same movies, the same music, over, and over, and over, to suit everyone's personal taste, and absolutely destroying entire rainforest's worth of electricity to accomplish it. And like... why do we want that?

If the AI doesn't get the mechanics right, and you have subtle game-breaking bugs that make it impossible to complete, you will never be able to fix it.

Precisely specifying desired functionality of software in ‘plain English’ is hard. Especially if you can’t provide diagrams, sketches, mocked-up screen layouts, etc.

Also, the low-code/no-code tools allow precise editing. With current prompt-based AI party tricks, a tiny change to the prompt can produce completely different output.

Emphasis on "if you could simply type in English" (or insert native language here).

Sometimes the AI misinterprets what you want. Sometimes this is due to lack of sophistication on the model's part, but I think the bigger issue at hand are the shortcomings of natural language. Even when working with an intelligent human being, sometimes you have to communicate your thoughts/ideas through pictographic means.

It's something I've thought a lot about lately. No-code/low-code is about as old as GUIs themselves, yet it's never become the de-facto way of creating software. However, for everything else we do we use a GUI. No one creates illustrations on the computer programmatically (unless you're into generative art like me). You probably aren't doing your taxes in a terminal. You get the idea.

This is a segue into my other critique of prompting as a UI. A lot of UX people think it is the ultimate UI, and natural language the premier way of communicating jobs to be done. I disagree. I think symbols and diagrams communicate more with less. Concerning language, I think that for whatever reason programming languages are the most natural way (that we've currently found) to communicate intent in a programmatic context, markup languages are the most natural way to communicate intent in a layout context, and that's the reason why despite 50+ years of GUI innovation programmers still use text-editors. The old guard has all aged out of the profession, the new guard still uses text-editors.

So that all being said, I think the most incredible AI demo I've seen so far is the tldraw stuff done by Steve Ruiz, namely Make Real, and that's because you're communicating with the AI in a pictographic / symbolic context, and because it's a whiteboard, there's no limit to the symbols at your disposal to communicate your intent. The limit is AI's ability to interpret what those symbols mean.

You are comparing an apple to an apple pie. The “easy” website is going to be something that looks like static html. The “easy” game will look similarly low tech: maybe a platformer that looks like it was written for an atari. 3d modern games are not the only way to make games although thats been a popular way for some time.

I've been using godot with chatgpt as a side project to see how much I can coax a game out of it, with minimal effort. Its a manual form of systems like this but definitely shows how generative tools enable independent devs or small teams to tighten their dev loops and rapid prototype ideas.

When you can rapid prototype ideas in a game search space, its more likely to get a high quality output, or at least games that hone in on fun more effectively than a AAA made with a waterfall model and a huge team with poor crosstalk abilitiy.

Yea you might want to update your blog post with some original and creatively inspired new content. Demoing Star Wars is just asking to be picked apart by people.

We haven't looked into exactly how this will be handled yet

That's how we know you're the real deal! Move fast and break things, laws and such are mere details. Capture the market first, work around problems later.

This is an unacceptably lazy and reckless comment: your very first example blatantly plagiarized from Star Wars. You even called the bombs "BB-8 bombs" to make them look like the robot! It's just a dumb demo, but it means your product has no safeguards against copyright infringement, and you don't even care.

I suspect you might say "users of our technology assume full responsibility for copyright violations" or whatever, but what are your customers supposed to do if your product plagiarizes something more obscure than Star Wars? They can't be expected to check every generative AI output against every IP. You and your team need to take responsibility for the copyright problems your tool is guaranteed to create as currently implemented.

I really hate this "better to seek forgiveness" approach to copyright OpenAI has encouraged.

Please look into it. You are already breaking Disney's copyright just with this post, and if you ever hope this product Disney's lawyers would have a very easy case here.

That seems like a ridiculous stance. The safe harbor laws that those platforms depend on are for user-generated content. Here it's your platform that's generating the infringing content. (And obviously it's not by accident, you're practically reveling in it with your choice of demo.)

Even a failed attempt at this would be great - imagine your crew frequently misunderstanding your orders and the ensuing chaos.

It's been done. The AI plays it rather well.

https://ai.meta.com/research/cicero/diplomacy/

It actually worked fairly well, but obviously wasn't freeform since you couldn't order units to move to any arbitrary point on the map with a voice command.

I would bet that more than 0% of that perception, at least from developers, comes from how _fucking awful_ the Unity editor is. It's one of the worst pieces of software I've ever used. The games you can make with Unity are absolutely very impressive, but the tools Unity provides are just awful, and give a lot of newbie developers the perception that Unity is a hacked together pile of garbage (which...it is, but so are all game engines). Godot, Unreal, and GameMaker are, I would say, the main competitors to Unity in the Indie space, and all three of them have much higher quality editors than Unity's.

none of them seemed to face any issues with being made in Unity.

I'm sure they did. But I wager there is no such thing as "smooth development". Every engine has its quirks, every team has its odd dynamics and clashes.

engines are there to make development smoother, not smooth.

it's bound to have a bunch of shovelware made on it with how many resources there are for learning it (and Unity being free), but it's popular for a reason.

To be fair, that one is more on Unity. You had to pay to remove the splashscreen, so that meant low effort titles would have "MADE WITH UNITY" front and center, where higher budget titles could remove the logo. It was great for awareness, but it completely reversed the perception of the engine quality as a result.

(disclaimer: I used to work at unity)

I don't want to be asked 20 follow up questions to exactly specify the nature of the Orc

This is the tell that these folks don't know enough about game design to be opining so strongly on "the future of AI driven game development". Game development is almost exclusively about the 20 follow up questions. Any moron can say "I want minecraft but in space with guns", but making that and, more importantly, making it feel good, are where all the juice lies in games. The details are the most important part, which is why I think all these attempts to put AI into the game development flow are ill advised at best. Copilot is okay at helping out with game code, but less so than for normal e.g. REST API or UI code, because game code tends to follow rules that don't exist anywhere else and every game is architected completely differently. Games are just...different.

(I'm the in-house AI engineer at Braindump, but I come from a gamedev background.)

Yes, we agree that doesn't scale. Braindump's designed to let you change those assumptions at any time; the AI works over the state of the entire game (for the most part), so it can change past assumptions to better meet the user's current requirements.

Our rationale was that we wanted to go beyond the existing engine integrations (i.e. Unity already has an AI assistant) and reimagine the entire game creation experience with AI and multiplayer in mind.

How does that pan out?

There's hundreds of millions of dollars of product development in Unity and Unreal specifically for game design. How will you ever catch up?

Godot got $8m, for the people who authored it, what does it have to show for it? It seems like hopelessly little money.

Maybe you said to the investors, "You don't want to be locked in" or whatever. The browser still an engine. Mobile Safari and Chrome are also moving targets. You're still playing in someone else's back yard. There are no games that are designed to be played in Mobile Safari and also make money. I mean the browser also has AI assistants.

Obviously I want you to succeed but this sounds more like a good story than a good strategy. It doesn't hold up to an ounce of scrutiny. Right now it sounds like it's not interesting to develop the toolset for Unity because they are already doing it.

I'm guessing they mean with some type of procedural generation or another randomization approach in mind that creates a bunch of new maps, bosses, vehicles even!

Advance Wars, if I recall correctly, has a hand-crafted set of maps and battles structured in a campaign story. So once you're done, you're done. Replay-ability is a bit limited.

I absolutely loved Advance Wars and my brother and I played the heck out of it growing up!

So far we've focused 100% on the creation side of things, but we're absolutely interested in helping people on the marketing side too.

Something we've discussed here is; since it's an online platform, could it be possible for creators to form communities around their games as they are being built (even more than now), with perhaps features to engage your community around what you create (imagine that your players can comment on things in the game as they play test it, for instance). This is quite speculative for now, but it's something we're interested in exploring.

Marketing is hardly something that can be done via an "add-on" or some mechanical approach.

It's more of a mindset, getting yourself and what you're working on known is something you need to do throughout the entire process, before, during and after the release of a game.

I'd say it's something in that direction (GameMaker) yeah. One big difference (other than the AI part) is that both building and playing is multiplayer by default. (Though of course, you can choose to make things single player and to build solo). With biting the hand that feeds it; I sure hope not! :) I've built my fare share of obscure indie games, and we in the company have friends and family members who do now as well, so the aim is to build something that can help people like this.

I like how the first one has two faces, one on the front and one on hair at the top

"They" would be Meshy. Again, I'm not a lawyer, but my understanding is that you can't grant a license to some IP unless you own it.

I think they're taking the piss.

Palworld was a big one. Tribes 3 is another.

Which games like that used AI?

They are baiting the mouse with the star wars items

Sounds like a nightmare tbh. Half the fun is the actual coding (aka the journey). I guess if you want to be a prompt engineer that sounds like fun but it sucks all the art of programming out of it.

The callback style in createOrUpdateRule appears to create isolated snippets of code that can't invoke other rules.

Perhaps I've misunderstood what is happening, but if this is how it is, the agent can never write a function and compose other functions out of it.

That is functional programming where functions cannot call each other.

Composition of code units (ie. code A calls code B) is the basic building block that allows software to scale.

Without it, I don't believe you can get above trivial complexity.

looking forward to everything either way :)!

Very cool stuff Mr. Noren. Thanks for sharing!