A June 2025 tweet by Tobi Lütke proposed "context engineering" as a replacement for "prompt engineering," a term Andrej Karpathy endorsed a week later as "the delicate art and science of filling the context window with just the right information for the next step." Within months the term appeared on an O'Reilly book cover and was structured into a taxonomy by LangChain. The article frames a single-document RAG pipeline through this lens, where four bricks (document parsing, question parsing, retrieval, generation) each emit typed pieces that converge on one LLM call. Parsing produces relational tables; question parsing produces a typed ParsedQuestion; retrieval produces a filtered subset of lines plus an audit; generation produces a Pydantic answer with cited evidence. The system prompt stays fixed for caching. Context engineering covers everything in the model's context window for one call: the system prompt, retrieved documents or rows, conversation history, tool definitions and outputs, memory, scratchpads, agent state, structured metadata, and the actual user input. The discipline shifts from focusing on one block of text to managing multiple upstream sources.