AI Concept Atlas
FIELD NOTES / 01
Guided learning · 6 lessons

What the model can see

Distinguish current context, retrieved evidence, and retained memory.

Lesson 2 of 6

Context Window

Tokens are the units used to measure context capacity.

A context window is the token capacity available for a model’s current interaction. It limits how much supplied information and generated content can fit in that interaction, according to the model and service’s accounting rules.

[Anthropic]

In practice

An assistant loads a relevant page from a long manual instead of placing the entire manual in an already crowded conversation.

[Anthropic]

A little deeper

Instructions, conversation history, and tool results can consume context. Information stored elsewhere is not automatically visible to a model just because the application retained it. [Anthropic]

A common mix-up

A larger context window is permanent memory.

Context capacity and storage across sessions serve different purposes. [Anthropic]

Helpful to know: Token

Explore Context Window and its connections ↗

← Previous lesson