Introduction to Context Design
Overview
Context design is the practice of deliberately selecting the information provided to an LLM for a given task. The quality and relevance of this selection directly determines output quality and stability.
The fundamental principle can be stated as a reframing:
The operative question is not "what can be included?" but "what should be included?"
Core Principle
Before including any information in context, the following test applies:
Is this information necessary for the model to complete the current task correctly?
If the answer is not an unambiguous yes, the information should be excluded. Potential relevance or speculative utility do not meet the inclusion threshold.
Key Takeaways
- Context is a design decision requiring deliberate selection
- Inclusion criteria are task-specific and immediate, not speculative
- Signal clarity outperforms information volume