A game created with AI in just 30 minutes is generating more than $50,000 a month, its creator claims—and could be a peek at how AI can, and can’t, change game design in the future.
fly.pieter.com, an in-browser “fun free-to-play MMO flight sim made with AI” was made by Pieter Levels, who amassed a huge online following for pioneering the practice of quickly developing and launching software and startups with the help of AI. As he explains in his X bio, “All my websites/apps/startups/projects are built by just me with vanilla HTML, JS with jQuery, PHP and SQLite. I'm very fast with my own little stack. I don't collaborate with other people and prefer shipping fast by myself.”
This is sometimes referred to as “vibe coding,” which generally means being less methodical and detail oriented, telling the AI tool what you want, and getting it to work without worrying about the code base being messy.
Levels’ X bio also lists how much revenue he claims to currently make from each of his projects, including $132,000 a month from PhotoAI, a service that says you can “fire your photographer” and “generate photo and video content for your social media with AI” instead. He was also on Lex Fridman’s six months ago if you want to listen to him talk about how he lives and works for almost four hours.
“Today I thought what if I ask Cursor to build a flight simulator,” Levels posted to X on February 22, referring to AI coding software. “So I asked ‘make a 3d flying game in browser with skyscrapers.’ And after many questions and comments from me I now have the official [fly.pieter.com] Flight Simulator in vanilla HTML and JS.”
About two weeks later, Levels said on X that he is on track to make $52,360 a month from the project. That’s an impressive milestone to hit in such a short time, and with a game that was essentially started with a prompt for an AI tool, but the reality of the situation is a bit more complicated.
To start, as Levels said on X, $360 of that figure comes from in-game purchases, specifically 12 jets that players in the game bought for $29.99 each. In-game purchases is how free-to-play games traditionally make money, and while $360 in such a short time is also not nothing, it’s not nearly as impressive as the $52,000 a month the game is making from 22 in-game ads “sold at various prices.”
One of the biggest problems in video games today, as it is with most other forms of media and entertainment, is discoverability. There’s so much stuff online, it’s hard for one thing to stand out and find an audience. Over 19,000 games were released on Steam last year, many of them free-to-play, and very few of them could get enough eyeballs that can be converted into any real revenue.
Fly.pieter was able to do this because Levels has a large and specific audience: 623,000 followers on X, many of whom are in the AI industry or interested in AI. One of the “bigger” sponsors according to Levels is Bolt, which, not surprisingly, makes an AI tool for developing web and mobile apps. Other ads I saw in the game are also for AI companies. It also probably didn’t hurt that the game was shared by Elon Musk, who has almost 220 million followers on X, a boost in reach that would probably do wonders for the business of any one of those 19,000 games that hit Steam last year.
Looking only at how quickly fly.pieter was made and generated significant revenue is also misleading because it ignores the talent he brings to the project and the rapid iterative work he’s put into it and documented on X. It’s wild that AI tools are at a place now that Levels was able to get a working game from just a prompt, and that he was able to do things like add controller support with just a prompt as well, but Levels is putting real work into it also. He had to fix a vulnerability that allowed people to hack the game to promote a pornographic site, add blimps and planets that serve as the ad space he’s selling, and has added a bunch of other features and bug fixes. All of this in order to say: it’s not as easy as typing a prompt or just feeling the vibes. It takes some actual experience and skill.
Whether the game is good or not is obviously subjective and also seems beside the point, but I think it’s fine. You have a little plane, you fly it around, you can shoot balloons and other planes though they appear to be lagging too much to really do that, and mostly you can look at ads for various AI companies. It still feels more like a graybox prototype than a real game. It’s tempting to see a short clip from fly.pieter and dismiss it as a simple looking multiplayer toy that allows players to bash low poly 3D models together, but that’s also kind of what Roblox is and it’s one of the biggest games in the world. It’s impossible to say how a project like this will develop and whether it can gain and hold an audience, which goes back to the problem of discoverability.
Will more people be able to create more games in the future because of AI tools? Fly.pieter makes a strong case that the answer is yes. But it is also a sign that AI tools will do to games what they are already doing to images, music, text, and other creative forms online: they are flooding the zone with low quality, AI-generated stuff that will only make discovering things harder.
