07/30/2026
It's exciting to see CTOR (published as Torus) being explored in the context of collective intelligence and multi-agent systems.
Originally designed as a brain-bending strategy game, CTOR has drawn praise from Ernő Rubik (Rubik's Cube) and Alexey Pajitnov (Tetris), and it's fascinating to see new research directions emerging around its unique mechanics.
If you're curious about where modern board games, mathematics, and AI research intersect, this is well worth checking out.
This is an impressive story about the relationship between AI and the gaming industry, published in an article on LinkedIn. However, while physical worlds can be predicted through 3D games, and AI can be trained to play such worlds, logical worlds are not yet possible.
All AI bots playing abstract games are still unable to solve two problems:
1. Team-based logic games.
2. Modeling collective thinking, such as intuition.
It is these AI training models that our platform,
STOR, covers.
https://ctorgame.com/
2015 - Atari. Google DeepMind built a system that learned to play dozens of Atari games from raw pixels. No rules, no instructions, just the screen and a score to chase. First real proof an AI could learn a game the way a person does: by looking at it. The game was the opponent.
2016 - Go. AlphaGo beat Lee Sedol, one of the best players alive. It's a board game, but it's the moment the world paid attention, partly for Move 37, a play so strange no human would have made it. AI wasn't matching us anymore. It was out-thinking us.
2019 - StarCraft II and Dota 2. DeepMind's AlphaStar reached Grandmaster in StarCraft II, though let's not forget it first got its a$$ beaten by Polish pro Grzegorz "MaNa" Komincz. It went on to master all three races in SC2 without superhuman reflexes, playing under the same visual and action constraints as human professionals. Then OpenAI Five beat the Dota 2 world champions. The hardest real-time games we had, hidden information and thousands of decisions a minute, solved. To me, this was peak "AI vs game."
2025 - Pokémon. And then Anthropic 's Claude playing Pokémon Red live on Twitch, which wasn't about winning anything. It was a way to watch a model reason, plan, get stuck on a wall for an hour, and think its way out in real time. The game stopped being the opponent and became a window into how an AI actually thinks.
2024 to now - SIMA and General Intuition. And here the flip completes. DeepMind's SIMA 2 plays across 3D worlds it has never seen, learning to navigate and reason, aimed openly at robotics. General Intuition raised $454M to turn gameplay into world models that teach AI motion and space. The game isn't the opponent, or even the test. It's the classroom.
Look at that arc: the entire relationship inverted.
For ten years, games were the adversary. The question was always "can AI beat this?" Chess, Go, StarCraft, Dota, all have win conditions to conquer.
AI against the game.
Now games are the teacher. The value isn't in beating them, it's in what's inside them: physics, motion, space, cause and effect. Millions of hours of how things move and behave, the exact stuff text can't teach.
AI and games are on the same side now.
The game world is where AI goes to learn about the real one.
We spent a decade teaching AI to win games.
Now, games are teaching AI how the world works.
Over the past few years, we've developed several AI algorithms for gaming...the time has come to expand our presence in the gaming and AI industries. If you're ready to help with this process, register on the website and select your status. https://ctorgame.com/
