Beyond the Prompt
In June 2025, Andrej Karpathy posted a message that crystallized what production AI engineers had been feeling for months: "People associate prompts with short task descriptions you'd give an LLM in your day-to-day use. When in every industrial-strength LLM app, context engineering is the delicate art and science of filling the context window with just the right information for the next step." [1]
The distinction matters because AI agents are not chatbots. They run in loops, invoke tools, accumulate state, and make decisions with the residue of previous steps still in their context window. Prompt engineering — choosing the right words — is necessary but insufficient. By mid-2025, the real challenge was feeding an agent the right files, the right tool definitions, the right slice of conversation history, and the right retrieved facts at every turn, without the context window collapsing under its own weight. [2]
Shopify CEO Tobi Lütke put it concisely: "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." Within a year, Gartner declared 2026 the "Year of Context," identifying context engineering as the breakout AI capability of the year. Cognizant responded by announcing a deployment of 1,000 context engineers. The discipline now has its own job title, its own academic survey literature, and its own production infrastructure stack. [3]
What Context Engineering Actually Is
Context engineering is the discipline of designing dynamic systems that provide the right information and tools, in the right format, at the right time, to give an LLM everything it needs to accomplish a task. It encompasses everything beyond the prompt: system instructions, tool schemas, retrieved documents, conversation history, application state, and memory stores. [2]
A useful mental model from Karpathy: think of the LLM as a CPU, and its context window as RAM. The model can only reason over what is currently in the window. Everything else — your codebase, your documentation, your user's history — is invisible until you load it in. Context engineering is memory management for AI.
This differs from prompt engineering in a structural way. Prompt engineering optimizes a single instruction for a one-off task. Context engineering optimizes the entire information environment surrounding that instruction — across multiple turns, tool calls, and decision steps. If you are changing what data the agent retrieves, in what order, with what re-ranking, and what gets evicted when the window fills, you are doing context engineering.