Devnexus 2026 - Connecting the Dots with Context Graphs - Stephen Chin

This talk explores the necessity of durable context in AI systems, highlighting how Context Graphs can serve as an essential architectural pattern for agentic AI. The speaker discusses context engineering, which involves designing and managing a persistent context layer that captures the influence of context on actions through various constraints, policies, and outcomes. By integrating structured, multi-hop context assembly inspired by GraphRAG-style hierarchical summaries, the talk reveals how these approaches enhance explainability, evaluation, provenance, and long-term reasoning in AI applications. Viewers will gain practical insights into building context pipelines that merge contextual retrieval with persistent memory and decision-making, emphasizing the growing importance of context graphs in enterprise-ready AI architectures.