# context (10-context.com)

Context means the situation or background that helps you understand something. When people ask for context, they want the surrounding facts so a statement makes sense.

## What a model may hear
- context window (large language models): count tokens in the fixed-size input buffer and truncate or lose earlier text when the limit is reached
- execution context (programming and operating systems): switch between threads, processes, or security privileges based on stored state
- database context (web frameworks and ORMs): open and close connections, manage transactions, or scope queries to a particular session
- React context (frontend development): pass data through the component tree without prop drilling, triggering re-renders when values change
- NLP context (linguistics and transformers): use surrounding tokens to disambiguate word meaning through attention mechanisms

## Where people and models part ways
- Says: "I need more context"
  Means: explain the background of this topic
  May be taken as: the model may check token counts or summarize within the window rather than explaining
  Say instead: "explain the background and history that helps me understand this topic"
- Says: "keep this in context"
  Means: remember what we were talking about
  May be taken as: the model may try to store state in a programming context object rather than maintaining conversational continuity
  Say instead: "remember our earlier discussion about [specific topic] as we continue"
- Says: "out of context"
  Means: that quote was misused by leaving out surrounding words
  May be taken as: the model may discuss token truncation or window overflow in technical terms
  Say instead: "that quote changes meaning when the surrounding words are left out"
- Says: "provide context for this code"
  Means: explain what this code does and why
  May be taken as: the model may describe the runtime execution context, stack frames, or environment variables
  Say instead: "explain what this code does, why it was written this way, and how it fits into the larger program"

## Tips
- Say 'background information' instead of 'context' when you mean history or circumstances
- Specify 'conversation so far' or 'earlier in our chat' to ask the model to remember without triggering technical senses
- When discussing code, say 'explain this code' rather than 'give context' to avoid runtime environment descriptions
- If you mean the model's token limit, say 'token limit' or 'memory limit' directly
- Use 'surrounding text' or 'full passage' when you mean literary context rather than technical state

## Often confused with
- prompt: the actual input text, not the background behind it
- state: stored data values, not explanatory background
- memory: retained information, not situational understanding
- scope: visibility of variables, not general situation
- background: near-synonym in plain use, but less likely to trigger technical senses
- history: past events specifically, not the present situation