AI & ML Efficiency Breakthrough

Proposes an agentic architecture that achieves O(1) token complexity relative to dataset size by strictly separating intent parsing from deterministic data execution.

March 25, 2026

Original Paper

Reasoner-Executor-Synthesizer: Scalable Agentic Architecture with Static O(1) Context Window

Ivan Dobrovolskyi

arXiv · 2603.22367

The Takeaway

Standard RAG systems suffer from costs that scale with retrieved context; this architecture ensures a static context window regardless of whether searching 100 or 100 million records. It also eliminates data-driven hallucinations by construction because the LLM never processes raw, noisy retrieved records.

From the abstract

Large Language Models (LLMs) deployed as autonomous agents commonly use Retrieval-Augmented Generation (RAG), feeding retrieved documents into the context window, which creates two problems: the risk of hallucination grows with context length, and token cost scales linearly with dataset size. We propose the Reasoner-Executor-Synthesizer (RES) architecture, a three-layer design that strictly separates intent parsing (Reasoner), deterministic data retrieval and aggregation (Executor), and narrativ