- Updated: July 17, 2026
- 2 min read
The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality
The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality
Sharp et al. (2025) introduced the concept of agentic inequality to describe disparities in AI agent access across availability, quality, and quantity. Building on this framework, we identify a finer‑grained divide that operates at the level of individual interactions: the Context Access Divide (CAD). The CAD captures the difference between systems that can autonomously retrieve relevant context from a user’s knowledge corpus (Dynamic Context Retrieval) and those that require users to manually attach documents for each query (Manual Attachment).
For knowledge‑intensive professionals whose intellectual capital spans tens of thousands of files, the CAD represents a critical threshold in AI usefulness. When context must be manually curated, the cognitive burden shifts back to the human, negating the efficiency gains that AI promises. We formalize the CAD with a probabilistic model grounded in the fan‑effect literature, demonstrating that manual attachment leads to a combinatorial collapse in task‑success probability as corpus size and task conjunctivity increase, whereas dynamic retrieval architectures remain structurally insulated from this collapse.
We examine the technical foundations of the CAD in the Model Context Protocol (MCP) and Retrieval‑Augmented Generation (RAG) architectures, and discuss its implications for knowledge‑work stratification and AI platform governance. By introducing contextuality—the degree to which an AI system autonomously accesses a user’s accumulated knowledge capital—as a new dimension of AI‑mediated inequality, we complement the existing Sharp et al. framework and highlight a previously overlooked source of disparity.
Read more about our research and related resources on ubos.tech.
Andrii Bidochko
CTO UBOS
Andrii Bidochko is an AI entrepreneur and researcher focused on AI agents, reinforcement learning, and autonomous systems. He writes about the technologies shaping the future of machine intelligence, from frontier models and agent architectures to real-world AI applications.