These Russian Mathematicians Taught AI Models How to Talk to Each Other Without Using Words

A Russian startup, Mostik, has developed a method for AI models to exchange information through their weights rather than text, enabling a smaller model to inherit capabilities from a much larger one at a fraction of the cost. For CIOs and technology leaders, the strategic implication is significant: this could make open-weight models far more competitive with proprietary frontier systems, accelerating a shift toward modular, specialized AI architectures that lower inference costs and improve deployment flexibility. For IT organizations, the approach suggests a future where AI capability may be assembled from interoperable components rather than a single monolithic model, changing how teams evaluate vendors, design AI stacks, and manage performance and spend.