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

Mostik’s bridge technique lets AI models exchange information through their weight values, enabling a 753-billion-parameter GLM-5.2 to share capabilities with a 4-billion-parameter Qwen-3.5, producing a hybrid that costs one-twentieth the full GLM and performs midway between the two. The method, praised by experts, bypasses text output, potentially allowing smaller models to match large-model quality and fostering specialized, domain-specific AI systems.