Context Engineering
The practice of curating exactly what information enters a model's context window at each step, so the task is solvable.
- term
- Context Engineering
- category
- knowledge-memory
- short_def
- The practice of curating exactly what information enters a model's context window at each step, so the task is solvable.
- long_def
- Popularized in June 2025 when Shopify CEO Tobi Lütke and AI researcher Andrej Karpathy endorsed the term over 'prompt engineering'. Karpathy called it 'the delicate art and science of filling the context window with just the right information for the next step'. In production agents the prompt is a tiny fraction of context; the rest is retrieved documents, tool outputs, history and state — all of which must be engineered.
- see_also
agent-skillsagentic-ragprompt-caching- etymology_origin
- Popularized in June 2025: Shopify CEO Tobi Lütke endorsed the framing on 18 June 2025 and Andrej Karpathy publicly preferred 'context engineering' over 'prompt engineering' shortly after.
- related_to
agent-skillsagentic-ragprompt-cachingtoken-economicsrag- contrast_with
- Unlike prompt engineering, which crafts a single instruction, context engineering manages the whole context window — history, retrieved docs, tool results and state — of which the prompt is only a small part.
- example
- On 18 June 2025 Shopify CEO Tobi Lütke endorsed 'context engineering', and Andrej Karpathy followed, calling it the art of filling the context window with just the right information.
- source
- https://simonwillison.net/2025/Jun/27/context-engineering/
- status
- active
- why_it_matters
- Context engineering explains why lean, well-structured, retrievable content wins: it is cheaper and more reliable to fit into an agent's context window than sprawling HTML.
- sameAs
https://en.wikipedia.org/wiki/Prompt_engineering- bridge_entity
- agent-readiness
- last_verified
- 2026-06-15
- md_twin
- /glossary/context-engineering.md