Choosing what the model sees
If you ask an AI assistant to summarise a project, the app might send your question, recent messages, relevant project files, and instructions for the summary. Choosing and arranging that material is context engineering.
The app may also describe tools the model can use. The request and reply must fit within the model’s context window, the amount of information it can handle at once.
Files fetched for this request help shape the answer; they do not permanently retrain the model.
When extra context gets in the way
Unrelated or conflicting material can make answers worse, slower, and more expensive. Including unnecessary files can also expose private information. Teams choose relevant, trustworthy sources, respect access permissions, and test whether the added material helps.