Coding agent with algebraic memory (VSA) instead of RAG

This project replaces traditional Retrieval-Augmented Generation (RAG) in coding agents with a Vector Symbolic Architecture (VSA) for algebraic memory, enabling the agent to perform precise, compositional reasoning over code structures without external database lookups. It is designed for AI/ML researchers and developers building autonomous coding assistants who need more robust, interpretable, and computationally efficient memory than RAG provides. The approach is interesting because it leverages hyperdimensional computing to directly encode and manipulate code semantics as high-dimensional vectors, potentially eliminating retrieval errors and enabling novel forms of in-context learning.