"Stencil One uses a new bundling algorithm inspired by “word embedding”, a popular Machine Learning technique. It allows us to pack components more efficiently, therefore reducing the network throughput and round-trips. In order to do this, we generate a N-dimensional vector space where each component represents a coordinate. These coordinates are generated based on the previous static analysis of how components depend on each other. Later, the components that are located close enough in the vectorial space are bundled together. A blog post explaining this technique will be coming soon!"