Encyclopedia · 176 concepts

NLP & Language · intermediate · concept 88 of 176

Context Engineering

The practice of deciding what occupies a model's context window on each call: system instructions, tool definitions, retrieved documents, and prior turns. Once an agent runs for many steps the binding constraint stops being prompt wording and becomes the budget of useful tokens, so compaction, summarising older turns, keeping notes outside the window, and retrieving just in time matter more than phrasing.

Key terms

CompactionJust-in-time retrievalToken budgetSub-agentsContext rot

Where you meet it in the real world

Long-running coding agents, research assistants, customer support over long histories