Knowledge base

What is context engineering?

Managing what an AI model sees at any moment: which data, instructions and history enter the finite context window.

Definition

In one paragraph.

Context engineering is managing what an AI model sees at any moment: which data, instructions and history enter the finite context window, and what deliberately stays out. Writing prompts was yesterday’s skill; curating what the model sees is the skill that makes the difference now.

Why it matters.

The window is finite

Everything that goes in crowds out something else. Sharp context beats a bigger model with messy context.

Leaving out is half the work

Not sending everything, but sending the right thing: that determines whether the model’s attention stays sharp.

The foundation under agents

An agent that runs for months stands or falls with what it gets to see at any moment.

We wrote a full dossier on it, with worked examples. Read Context Engineering

All terms

Back to the knowledge base overview.

To the knowledge base