When we talk about generative AI, the focus tends to land on the model. But a model without context is a conversation without memory: it answers, yet it does not understand where it is.
Context engineering is the deliberate choice of what surrounds each decision the system makes: which data, which history, which rules, which limits. It is a form of design, not configuration.
In a well-designed intervention, context is not accumulated — it is curated. The question is not how much the system can know, but what it needs to know to behave meaningfully at this precise point in the service.