New term: context engineering
2025-06-19 · Publisher : Tobi Lütke · Domain : Artificial Intelligence
Context engineering means giving a model everything it needs to succeed at a task: instructions, documents, history, tools, memory. The term extends and partly replaces prompt engineering.
What changes
- The focus moves from wording the prompt to selecting the information.
- Retrieved documents, memory, tool results and history share one budget: the context window.
- Summarizing, pruning and isolating clutter in sub-agents become common practice.
- Prompt engineering gives way to an architecture skill.
### Definition **Context engineering** is the craft of filling a model's context window with the right information, in the right format, at the right time: instructions, examples, retrieved documents, history, tool descriptions and tool results. ### Origin On June 19, 2025, Shopify CEO Tobi Lütke wrote on X: "I really like the term "context engineering" over prompt engineering. It describes the core skill better: the art of providing all the context for the task to be plausibly solvable by the LLM." On June 25, Andrej Karpathy followed up: "+1 for "context engineering" over "prompt engineering"", and the term spread within days. ### What the term covers - For Karpathy, people associate prompts with short task descriptions, whereas in any serious application the job is to fill the context window "with just the right information". - Choosing what to retrieve (RAG), what to summarize, what to drop and what to hand off to sub-agents. - Describing tools and formatting their results so they stay usable. ### Criticism and limits Some see it as just a new name for prompt engineering. Yet the term usefully reminds us that an overloaded context degrades answers, an effect documented shortly afterwards as "context rot". ### Key takeaway Answer quality depends less on how the prompt is phrased than on the information you put into, or remove from, the model's window.
Links
Tags : Vocabulaire, Contexte, Agents, Prompt