Dynamic Corpus Architecture (DCA): A Knowledge Delivery Architecture for Connected Intelligence

The rapid advancement of Large Language Models (LLMs) has transformed the field of Artificial Intelligence by enabling systems capable of understanding, reasoning, and generating human-like language. Despite these advances, contemporary LLMs remain fundamentally constrained by a static knowledge paradigm in which their responses are limited to information acquired during training or retrieved through external mechanisms such as Retrieval-Augmented Generation (RAG). This limitation creates a gap between the reasoning capabilities of the model and the dynamic, continuously evolving nature of real-world information and services. This white paper introduces the Dynamic Corpus Architecture (DCA), a capability-oriented knowledge delivery architecture designed to extend the operational intelligence of LLMs through dynamic acquisition of external knowledge and services. Rather than treating knowledge as a collection of static documents, embeddings, or databases, Dynamic Corpus redefines knowledge as a set of discoverable and invocable capabilities exposed through external tools. These capabilities are described through a Knowledge Delivery Network (KDN), a structured capability manifest that enables an LLM to identify, request, and utilize external sources of information and computation on demand.