Context Window | behavior.engineering

The context window is the model’s working memory for a given conversation: everything that happened before the current message, plus the system prompt, retrieved documents, and tool outputs, all has to fit within this limit. Modern models have dramatically expanded context windows — from a few thousand tokens a few years ago to hundreds of thousands today — but longer contexts aren’t free. Models can lose track of information buried in the middle of very long contexts, and longer inputs cost more to process. For behavior architects, context window design is a key architectural decision: what information to include, how to order it, and what to leave out when space is limited all significantly affect the quality and cost of model responses.

Context EngineeringThe broader practice of designing what information a model has access to at inference time — including instructions, memory, tools, and retrieved content.
Context StrategyA deliberate plan for what information to include in a model's context window, how to structure it, and what to exclude given space and quality constraints.
Few-Shot PromptingProviding a model with a small number of examples of the desired input-output pattern before asking it to complete a new task.
In-Context LearningA model's ability to adapt its behavior or improve at a task based on examples and information provided in the prompt, without any change to its underlying weights.
Prompt EngineeringThe practice of crafting and refining the text given to a model — instructions, examples, context — to reliably produce desired outputs.