A Brief Timeline of Pieter Levels’ Flight Simulator (or the Ten Thousand One-Shots Rule)
If you’re on X at all, you’ve likely seen Pieter Levels’ new flight simulator at fly.pieter.com He started it about 10 days ago. Since then, he has posted almost 100 times about the project on his timeline, and likely much more in the comments. According to his bio on X, he’s already made $38,000 selling in-game ads and fighter jets.
I got inspired by this and decided to comb through Pieter’s timeline to map out the trajectory of the project. In this article I will go through a brief timeline of the game’s development and then discuss AI-driven game development in general.
The Beginnings of a Flight Sim
On February 22nd he started building his flight simulator and made his first post about it.
https://x.com/levelsio/status/1893350391158292550
He had never used Three.js before. He made it live to play the same day.
On February 24, Anthropic added the ability to create missiles because of Levels saying that Claude was giving him problems creating missiles due to AI safety measures.
https://x.com/levelsio/status/1894103554496831909
By February 25th he had 200 players online
https://x.com/levelsio/status/1894468597205856413
Then 500 players.
https://x.com/levelsio/status/1894471126228623848
On February 26 Elon Musk quoted it in a post.
https://x.com/elonmusk/status/1894744125938819481
To which Levels responded by adding Mars to the game.
https://x.com/levelsio/status/1894750792147460499
On the same day, he made it possible to buy an F-16 for $29.99
https://x.com/levelsio/status/1894762158031827437
Next, he gets the brilliant idea to sell ads on blimps from an X user @thekitze.
What if we sell blimps, what price should we charge though? https://t.co/WavO7DP23e pic.twitter.com/8G7jtmAvSF
— @levelsio (@levelsio) February 26, 2025
By February 27th he had already earned $1270 with the game and 89,000 people had tried it.
https://x.com/levelsio/status/1895126942233350340
He added dogfighting so you can shoot planes down. https://x.com/levelsio/status/1895169895027089471
On March 1 he added joystick / gamepad support.
https://x.com/levelsio/status/1895892285575086376
By March 2 he had made $17000 with ads and $360 by selling planes and someone bought the Moon and Mars for $5000.
Someone just bought the moon/mars 🤯
— @levelsio (@levelsio) March 2, 2025
Thanks @lewisbuildsai https://t.co/46XpT7WRsl pic.twitter.com/cCaShRix0H
On the same day someone also bought Jupiter for $5000.
https://x.com/levelsio/status/1896351936179556860
As of March 3, as many as 16,000 people in one day have flown in the game and he has raised in-game ads to $5,000 per month.
The mechanics are simple and need refinement, but the seeds of a game are all there. Right now, you can fly up to Mars and Jupiter. Personally, I think it would be a cool feature if you could enter the different planets’ atmosphere and then it brings you to a new level with different terrain.
A bright future for AI games, or the Ten Thousand One-Shots rule.
One-shotting trivial games with bouncing balls, such as Pong clones, has become a common metric for evaluating state of the art LLMs. These examples are good beginnings, but the true differentiation emerges when you combine ten thousand one-shots together. In honor of the Ten Thousand Hour Rule by Malcolm Gladwell, let’s dub it The Ten Thousand One-Shots Rule.
First, you need to ensure the game is playable by developing basic mechanics and simple models. Once the fundamental idea is working, then you start fine-tuning everything and getting user feedback. Make it fun and interactive. Refine the models. Then start going into detail on the things that work.
It’s inspiring that Levels immediately started getting results with modern LLMs, despite never having used Three.js before. This example shows that one person can leverage AI to create things for users while learning in the process.
Perhaps a new age of venture funding is upon us. A future in which you can build things, get users, and learn, all at lightning speed. If you broadcast your work on social media, many people are happy to help fund development by pitching in. Of course, having hundreds of thousands of followers doesn’t hurt!
He has his detractors no doubt. But results are results. And honestly, with a name like “Levels,” it makes sense that his destiny might be to create video games. But the most inspiring part is that anyone can do this. The only limiting factor is imagination.
This may be the opening salvo in the race of AI-driven game development, demonstrating how small and agile teams can get a large advantage over more established but slower-moving game development companies. While anyone can use these tools to generate code, breakthroughs demand relentless effort over a sustained period of time.