Every model has a context window, and even when it's large, more isn't always better. Irrelevant or redundant information dilutes the signal and can confuse the model. Before adding something to context, ask: "Does the model actually need this to answer well?" Trim aggressively. A focused, relevant context almost always beats a sprawling one.