A flight simulator built from one AI prompt was making $87,000 a month.
The guy behind it is Pieter Levels, and the game, fly.pieter.com, is a browser toy where you fly a blocky plane past billboards. In early 2025 it turned into one of the strangest money printers the indie world has seen.
Pat Walls and Gus from Starter Story did a full breakdown of the whole thing: how the game got built, where the $87K actually came from, and why the "he only made it because he has an audience" crowd is missing the point.
The episode is from March 2025, right in the middle of the hype, which means we now know how the story ended. Spoiler: the lesson held up better than the revenue. Here's the video, then the whole story with numbers.
Who is Pieter Levels?
If you've spent any time around indie hackers, you know the name. Pieter Levels (@levelsio) is the Dutch founder behind Nomad List, Remote OK, PhotoAI and Interior AI. No employees, no funding, revenue dashboards public since forever.
He's probably the most copied solo founder alive, mostly because of the volume: he has shipped dozens of projects over a decade, and most of them failed in public.
That track record matters for this story. Keep it in mind, because the critics brought it up and Pat Walls had a good answer.
The flight simulator that hit $87K a month
In late February 2025, Levels typed a prompt along the lines of "make a 3D flying game" into Cursor and let the AI build it. That's vibe coding: you describe the thing, the model writes the code, you keep prompting until it works.
The result was fly.pieter.com, a free-to-play MMO flight sim you run in the browser. The graphics are honestly bad. Gus says it best in the episode: some of these vibe-coded games look like you're about to catch a virus.
Nobody cared. The game exploded on X, his posts about it pulled over a million views, and coverage at the time reported a peak of 17,000 people flying around at once.
Then came the business model, and this is the part most people skip:
- The game is free
- Startups pay to put their logo in the game, on blimps and billboards you fly past
- Price: $5,000 a month per advertiser
A few weeks after launch, Levels posted that the game was doing $87,000 a month. A one-prompt toy, monetized like a Times Square billboard.
He never charged a single player. The product he sold was the crowd watching.
"He only made it because he has an audience"
The moment the revenue screenshots went up, the quote tweets started. One guy wrote something like: good for Pieter, but I guess the lesson is just have an audience. If I had his followers, I'd be making money too.
Pat Walls has zero patience for this one, and his rebuttal is the best five minutes of the episode.
Everybody starts at zero. Levels built his audience by shipping things, most of which flopped. Pat pulls up Airline List, a directory of airlines Levels once built that almost certainly never made real money. There are dozens of projects like that in his graveyard.
→ Would he ever have had the idea for a flight sim if he hadn't built Airline List and 80 other things first?
Probably not. The audience and the idea both came from the same habit: shipping, in public, for years. Complaining that someone else's distribution is unfair is just a way to avoid starting your own. Even an audience of 100 people is enough to get your first real feedback loop going.
I've seen the same pattern in founders I interview. The million dollar solo founder stories almost never start with an audience. They start with reps.
The "shipping trash" argument
The other critique was louder: vibe coders are shipping insecure garbage they don't understand.
There was a Reddit post going around at the time titled something like "Cursor messed up 4 months of my work." Engineers piled on: an entire cohort is entering programming through AI and missing the basics. This can't last.
Pat, who has a software engineering background himself, calls it cope. His sites have been attacked plenty of times and none of them were vibe coded. Security is a problem in every codebase, not an AI-specific one.
His actual framework is more useful than the argument:
- Building a plane game, a directory or a job board? Security barely matters until you have scale. Ship it.
- Storing bank details or health data? Don't vibe code that part. Bring in someone who knows what they're doing.
And if you don't know what you don't know, you can literally ask the AI what would make your app more secure and work through the list. The gatekeeping, in his view, came mostly from engineers worried about their jobs, contemplating one line of code for 16 hours while the vibe coders ship a thousand and fix problems when they show up.
Toys turn into businesses
Is a vibe-coded flight sim a real business? Wrong question, and the episode nails why.
Pat brings up Chris Dixon's 2010 essay, "The next big thing will start out looking like a toy." Skype looked like a toy. So did the first AI image generators. Levels built a goofy AI avatar app when Stable Diffusion was barely usable, and a couple of years later professional AI headshot services were a real category.
The pattern repeats:
- No-code looked like a joke to engineers. It evolved into the AI coding tools everyone uses now.
- Dropshipping and NFTs were trends that burned out. The underlying skill (spotting attention early) didn't.
- Vibe-coded games looked stupid in March 2025. The capability behind them is now how a huge share of indie apps get built.
The trend got so big that Levels ran a game jam for it, with one rule: your game had to be mostly written by AI. Hundreds of people entered with games built in hours.
Even Gus, who doesn't code, one-shot prompted a writing app that critiques his copy in the style of Gary Halbert, styled like Hemingway Editor. Built in an evening, for fun. That's the tell. When non-programmers start building software because it's fun, something structural has changed.
One more line from the episode worth keeping: cringe is a bull signal. If a trend makes you roll your eyes because everyone is suddenly doing it, it usually has a lot longer to run than you think.
What happened after the hype
This is where watching the episode 20 months later pays off.
The vibe-coded games gold rush cooled, exactly like the hosts predicted it would. Most of the copycat games are gone. Later coverage pegged the flight sim's ad income at a small fraction of its peak once the novelty faded.
But fly.pieter.com is still online today. I checked before writing this. And Levels went right back to his portfolio: public trackers put PhotoAI alone above $100K a month in 2026.
So was the game a failure because the $87K didn't last? No. It was a seven-figure-pace side quest that took him weeks, taught him (and everyone watching) what the new tools could do, and cost basically nothing. The game was never the business. What he proved was how fast one person can now run an experiment like this.
What I'd steal from Pieter Levels
Four things from this story worth applying this week:
1. Ship the toy. Your silly weekend project is not beneath you. Low-ego builders win because they ship things that might flop. The precious founders polishing their "baby" for 18 months lose to the person launching 15 experiments.
2. Sell attention, not just the product. The genius of fly.pieter.com wasn't the game, it was charging startups $5,000 a month for blimps. When something of yours gets eyeballs, ask who would pay to be in front of them. GeoSports did the same thing with a viral web game and turned it into $40K a month.
3. Volume beats perfection. The audience critique dies the moment you look at the failed-project graveyard behind every "lucky" launch. Your 15th project benefits from the 14 corpses before it.
4. The barrier is now $20 a month. Cursor cost $20/month in 2025. The tools are better and cheaper now. If you've been waiting for permission to build, this is the cheapest it has ever been to find out if you like it.
FAQ
What is fly.pieter.com?
Fly.pieter.com is a free browser-based MMO flight simulator that Pieter Levels built in early 2025 using AI coding tools, starting from a prompt asking for a 3D flying game. It monetizes through in-game advertising: startups paid $5,000 a month to put their logos on blimps and billboards inside the game. It's still playable today.
How much money did the flight simulator make?
A few weeks after launch, Levels publicly posted that the game had reached $87,000 a month in ad revenue, with coverage at the time reporting around 17,000 peak concurrent players. The revenue dropped to a fraction of that once the hype cooled, but the game remains online and the experiment cost him almost nothing to run.
Did Pieter Levels really build it with one prompt?
The first version, essentially yes: he asked an AI coding tool to make a 3D flying game and iterated from there. Worth being honest about the asterisk: Levels is an experienced programmer, so he could steer the AI and fix what broke. The tools removed the specialized 3D knowledge he didn't have, not the judgment he did.
Is vibe coding still worth learning in 2026?
More than in 2025. The episode's core prediction held: the games were a fad, the capability wasn't. Non-technical founders now routinely ship real apps with AI coding tools, and the founders I interview increasingly treat them as a default part of the stack. The people who started playing with these tools early are now years ahead.
Every week I sit down with bootstrapped founders doing $100K to $10M a year and dig out exactly how they did it, failed projects and all. If you'd rather hear these stories straight from the founders: