Context engineering, a term coined by Tobi Lutke, has replaced prompt engineering as the dominant framework for building AI products. The shift is simple: prompts alone are not enough. What produces reliable, user-aligned responses is an orchestrated combination of instructions, retrieved knowledge, narrow tools, selected memory, and system state, all targeted to the specific task at hand.
Nielsen Norman Group's piece on context architecture extends this further, applying information architecture principles directly to AI systems. The argument is that how context is structured, not just what it contains, determines whether an agent interprets information correctly and responds in ways users actually need.
The full article is worth reading for how it operationalizes these principles into a design framework, not just a conceptual shift. If you are building agents or AI-assisted interfaces, the architecture decisions described here are the ones that will define output quality before a single prompt is written.
[READ ORIGINAL →]