How It Works
Weight-space programming
Having an AI model read or write another neural network's weights directly — the numbers inside the network — instead of training that network on data. Researchers report that frontier models can tell what a small network was trained to do from its weights alone, and can write small working networks by hand.
Origin
Named in a November 2, 2026 research post by MIT CSAIL researchers, "Sparks of Weight-Space Programming"; the findings are early and not peer-reviewed.