Context engineering is becoming more important than prompt engineering. A prompt tells the model what to do. Context determines what the model has available to do it well. That includes retrieved documents, conversation history, tool outputs, system instructions, user preferences and application state. A strong AI engineer learns to control this context deliberately. Keep irrelevant information out. Prioritize the most useful evidence. Separate trusted data from generated content. Handle stale information. Compress long histories without losing important details. Make tool outputs structured and predictable. Many AI failures that look like model problems are actually context problems. Better context often improves an application without changing the model at all.