IntelliTechs ยท Oct 19, 2026
You won't be playing Minecraft with these agents tomorrow, but the experiment shows AI models can already split up work, warn each other about danger and learn from their own mistakes for more than a day straight.
The four agents played in a randomly generated Minecraft survival world inside a sandbox. Nobody assigned roles, strategies or game-specific tools.
The agents talked through in-game chat and shared a common memory. On their own, they split up the work and settled into roles.
The Decoder reports the session cost $1,620, calculated at the standard Gemini 4 Argon price that takes effect in 2027.
Engadget draws a contrast with AlphaGo, DeepMind's own system: beating Go took a purpose-built AI. Here, a general model with no game-specific tools finished a long, complex game.
Agents that act on your behalf already show up in consumer products, at a much smaller scale.
Read the full article