Context Engineering
> Prompt Engineering

Prompt engineering asks: "How do I word this?" Context engineering asks: "What does the model need to know?"

What is Context Engineering?

Context engineering is the discipline of designing and managing the entire information environment that an LLM operates within. It goes far beyond the instruction (the prompt) to include:

The Iceberg Analogy

Think of every LLM call as an iceberg:

• Above water (10%): The user's message — "Summarize this meeting."
• Below water (90%): System prompt + retrieved docs + conversation history + tool schemas + guardrails.

Prompt engineering focuses on the tip. Context engineering manages the entire iceberg.

Why It Matters for Production AI

The Same Prompt, Different Context = Different Results

// Same instruction, completely different outcomes:

// Context A: Customer support bot with company FAQ retrieved
"How do I cancel my subscription?"
→ "Go to Settings > Billing > Cancel. Need help?"

// Context B: General assistant with no company docs
"How do I cancel my subscription?"
→ "It depends on the service. Usually check your account settings..."

// The PROMPT is identical. The CONTEXT made the difference.

Context Engineering = Managing the 90%

In practice, this means:

1. Versioning your system prompts separately from your retrieval logic.
2. Managing your RAG pipeline — what gets retrieved, how it's formatted, what gets filtered.
3. Curating tool definitions — which tools are available in which contexts.
4. Designing memory systems — what to remember, what to forget, how to summarize.

How PromptOps Fits In

PromptOps manages the most critical layer of context: the system prompt and its versions. By externalizing your prompts from code, you can iterate on instructions independently of your retrieval pipeline, your tool definitions, and your application logic.