These Russian Mathematicians Taught AI Models How to Talk to Each Other Without Using Words
Russian startup Mostik cracks inter-model communication without natural language — a new frontier for multi-model systems.
Why it matters
A novel approach to model-to-model interaction could reshape how practitioners compose multiple AI capabilities in production. If replicable, this moves beyond ensemble methods and touches on agent orchestration, capability routing, and compute efficiency.
The key facts
8 to knowMostik startup developing non-linguistic model-to-model communication
Russian mathematics team behind the approach
Potential application to multi-model capability composition
Published September 2, 2026 in Wired
Framed as 'wild new approach' — technical novelty unclear from headline alone
Mostik (Russian startup) develops non-linguistic communication protocol between AI models
Eliminates natural-language bottleneck in model-to-model coordination
Potential application: agent-to-agent systems operating at higher efficiency
Go to the source
Wired AIwired.com
Publisher excerpt: A startup called Mostik has a wild new approach to combining the capabilities of AI models.